Abstracts of papers presented at the ISLAR (International Symposium on Laboratory Automation and Robotics) 2001

s of papers presented at the ISLAR (International Symposium on Laboratory Automation and Robotics) 2001 The 19th International Symposium on Laboratory Automation and Robotics provided presentations on state-of-the-art developments in laboratory automation and robotics. The symposium programme included papers and posters on all aspects of the technology. These comprised: managing laboratory automation (drug discovery); bioanalytical analysis; managing laboratory automation in drug discovery development and QC laboratory; functional genomics strategies and high throughput screening; advanced integration strategies; method development and global methods transfer; compound handling and logistics; combinatorial chemistry and automated synthesis ; high throughput LC-MS-MS; increasing eæ ciency in dissolution testing; lead optimization; strategies for UHTS; increasing throughput for ADME toxicology; data management/data handling and bioinformatics; using contract laboratories to increase productivity; assay miniaturization; process optimization; compliance and automation Ð the regulatory perspective; novel high throughput screening technologies; compliance and automation Ð the industry perspective. Several discussion sessions were included and activated, and provided interactive communication on a wider range of subjects. Although the programme was very comprehensive, the Symposium was designed to provide time for both formal and informal exchange of information. The technical presentations were organized into concurrent sessions with grouped papers on related topics. Abstracts for each paper and each poster are included here. Full presentations of several of these papers will appear in later editions of this journal.s for each paper and each poster are included here. Full presentations of several of these papers will appear in later editions of this journal. Designing custom automation systems to  t the needs of early development Adam M. Fermier, The R. W. Johnson Pharmaceutical Research Institute, Raritan, NJ, USA Co-authors: John Troisi, Ramon L. Rodriguez, and James V. Weber, Michael P. Graham and Erin C. Herita The pharmaceutical industry is under extreme pressure to reduce the costs of bringing drugs to market. Thus, to achieve this goal the entire drug development process requires continued ® ne tuning and the introduction of new technologies. Drug discovery has been moving through this change over the past decade. Drug development is now looking at that same daunting task of increasing throughput while maintaining compliance and headcount restraints. Custom automation systems provides one opportunity to attain this goal. We have been investigating and building custom automation into our drug development process. For example, we have modi® ed a Zymark Benchmate for weighing small quantities (1± 50 mg) of dry powders. The addition of weighing dry powders from a stock tube enables many advantages and increases the  ̄ exibility of the Benchmate. One application we have been using this robot for is to weigh individual vials of drug substance into HPLC vials for subsequent accelerated solid state degradation studies. This typically requires about 20± 40 weighings for a single study. A typical degradation program would have a dozen studies for a single compound. The weighing robot not only saves time but also oå ers some intrinsic safety because the analyst is not directly exposed to the material. We have also designed a custom automated system to conduct accelerated degradation studies. These studies require precise control of sampling and temperature. A typical degradation study is usually conducted over a period of weeks. To compare studies in a relative manner for excipient compatibility studies, these times and temperatures must be maintained. It is an ideal application for automation because of its repetitive nature. The system allows for 16 reactions to be conducted in parallel. The software keeps track of temperature and times of sampling. In this presentation we will review criteria we have found to be successful in designing the instrumentation and identifying processes that would be ideal to automate. Bringing Caco-2 studies from moderateto highthroughput screening Dawn Alderman, Neurogen Corporation, Branford, CT, USA High throughput screening (HTS) technologies have been widely used in many areas of drug discovery. Human Caco-2 cells provide an in vitro model of drug bioavailability. They are derived from a colorectal adenocarcinoma cell line and exhibit functional characteristics of the lower small intestinal tract. They diå erentiate as a monolayer with apical and basal directionality that mimics the intestinal barrier. Involvement of P-glycoprotein and other active transport systems expressed endogenously in ̄ uence drug absorption. Currently, most Caco-2 studies have low to moderate throughput due to assay complexity. Many diå erent assay formats exist and are commercially available. Becton Dickinson provides a 3-day assay system and a 10-day assay protocol. The 3-day system includes a Journal of Automated Methods & Management in Chemistry Vol. 24, No. 3 (May–June 2002) pp. 61–98 Journal of Automated Methods & Management in Chemistry ISSN 1463± 9246 print/ISSN 1464± 5068 online # 2002 ISLAR http://www.tandf.co.uk/journals DOI: 10.1080/14639240210141787 61 media kit, vital to the growth and diå erentiation of the cells used within a 3-day period. The 10-day protocol suggests that the cells can be seeded and used after a 10day growth period without the requirement of the enhanced media necessary for the 3-day system. The 10-day assay media that we currently use has an antibiotic in it to protect against the potential bacterial contamination that may occur with the cells seeded in the plate for 10 days. The antibiotic condition may aå ect the data when comparing with the 3-day system. Neurogen is currently evaluating these assays for any potential diå erences. The Caco-2 model does not take into consideration ® rstpass metabolism. The detection method also creates a signi® cant bottleneck. Currently LC/MS/MS is being used. By using HTS technologies, such as laboratory automation and miniaturization, we have shown that the throughput of the Caco-2 assay can be signi® cantly increased. The introduction of the TECAN GENESIS has allowed the automation of the drug addition to the Caco-2 assay plates. It has also allowed for automated sampling. The TECAN GENESIS that is currently in our facility has 1 ml Tē on-coated syringes, and a 21plate deck design. The programs being used have been written in-house using the GENESIS software. We have evaluated the 3-day assay system and are currently evaluating the 10-day protocol, in 24and 96-well formats. We have found that our data are of good quality and are reproducible in-house. We are still evaluating many aspects of the 96-well assay format. Eå ective cell density and the volumes used in each assay format have been assessed. We began by investigating a couple of cell densities. We have evaluated 12.5 K/well (Becton Dickinson protocol) and 25 K/well. Well volumes have also been assessed. Initially 100 ml in the apical ports and 250 ml in the basal ports was used. However, this condition is associated with a suspected wicking problem. The sample under the plate lid as well as the volume moving up the interior wall of the basal compartments has been observed. This is a possible suspect for interwell contamination. We are further developing this assay and will evaluate this phenomenon. We have attempted to cut the volumes from 100 ml in the apical compartment to 80 ml. The basal well we have cut the volume from 250 to 200 ml. There are many assay conditions that have been published that vary between laboratories. Many publications suggest that a pH gradient from 6.5 (apical compartments) to 7.4 (basal compartments) is the best model, representing the conditions of the lower small intestine. Another assay condition that varies between laboratories is what concentration of drug to use. High concentrations of drug create the risk of saturating the P-glycoprotein transporter as well as the other active transport mechanisms. We are currently evaluating our Caco-2 studies in modi® ed HBSS buå er at 5 mm drug concentration using the pH gradient. Again, cell density is a factor that varies between laboratories. Becton Dickinson recommends a cell density in a 24-well plate to be 200 K/well and in a 96-well plate to be 12.5 K/well. We are currently using this recommendation for both assay formats discussed. Currently, samples are evaluated for Papp (apparent permeability) using this published equation: Papp...10cm s¡1† ˆ ...Vd=A†...dC=dT†: This is a rate distance measurement, where Vd is the volume of the donor compartment. A is the surface area of the monolayer, dC is the change in concentration and dT is the change in time. The Caco-2 cell model mimics the dynamics of the human lower small intestinal tract. The potential for this model to move towards a high-throughpu t format oå ers the high-throughput compound screening in the future, which is essential to modern pharmaceutical development. Optimizing the current cell density and volume speci® cations serve to strengthen this assay format, and is currently a major hindrance in its development. Rational approach for selecting extraction parameters for homogenizer-base d tablet assay

The pharmaceutical industry is under extreme pressure to reduce the costs of bringing drugs to market. Thus, to achieve this goal the entire drug development process requires continued ® ne tuning and the introduction of new technologies. Drug discovery has been moving through this change over the past decade. Drug development is now looking at that same daunting task of increasing throughput while maintaining compliance and headcount restraints. Custom automation systems provides one opportunity to attain this goal.
We have been investigating and building custom automation into our drug development process. For example, we have modi® ed a Zymark Benchmate for weighing small quantities (1± 50 mg) of dry powders. The addition of weighing dry powders from a stock tube enables many advantages and increases the¯exibility of the Benchmate. One application we have been using this robot for is to weigh individual vials of drug substance into HPLC vials for subsequent accelerated solid state degradation studies. This typically requires about 20± 40 weighings for a single study. A typical degradation program would have a dozen studies for a single compound. The weighing robot not only saves time but also oå ers some intrinsic safety because the analyst is not directly exposed to the material.
We have also designed a custom automated system to conduct accelerated degradation studies. These studies require precise control of sampling and temperature. A typical degradation study is usually conducted over a period of weeks. To compare studies in a relative manner for excipient compatibility studies, these times and temperatures must be maintained. It is an ideal application for automation because of its repetitive nature. The system allows for 16 reactions to be conducted in parallel. The software keeps track of temperature and times of sampling.
In this presentation we will review criteria we have found to be successful in designing the instrumentation and identifying processes that would be ideal to automate.
Bringing Caco-2 studies from moderate-to highthroughput screenin g

Dawn Alderman, Neurogen Corporation, Branford, CT, USA
High throughput screening (HTS) technologies have been widely used in many areas of drug discovery. Human Caco-2 cells provide an in vitro model of drug bioavailability. They are derived from a colorectal adenocarcinoma cell line and exhibit functional characteristics of the lower small intestinal tract. They diå erentiate as a monolayer with apical and basal directionality that mimics the intestinal barrier. Involvement of P-glycoprotein and other active transport systems expressed endogenousl y in¯uence drug absorption.
Currently, most Caco-2 studies have low to moderate throughput due to assay complexity. Many diå erent assay formats exist and are commercially available. Becton Dickinson provides a 3-day assay system and a 10-day assay protocol. The 3-day system includes a media kit, vital to the growth and diå erentiation of the cells used within a 3-day period. The 10-day protocol suggests that the cells can be seeded and used after a 10day growth period without the requirement of the enhanced media necessary for the 3-day system. The 10-day assay media that we currently use has an antibiotic in it to protect against the potential bacterial contamination that may occur with the cells seeded in the plate for 10 days. The antibiotic condition may aå ect the data when comparing with the 3-day system. Neurogen is currently evaluating these assays for any potential diå erences.
The Caco-2 model does not take into consideration ® rstpass metabolism. The detection method also creates a signi® cant bottleneck. Currently LC/MS/MS is being used. By using HTS technologies, such as laboratory automation and miniaturization, we have shown that the throughput of the Caco-2 assay can be signi® cantly increased. The introduction of the TECAN GENESIS has allowed the automation of the drug addition to the Caco-2 assay plates. It has also allowed for automated sampling. The TECAN GENESIS that is currently in our facility has 1 ml Te¯on-coated syringes, and a 21plate deck design. The programs being used have been written in-house using the GENESIS software.
We have evaluated the 3-day assay system and are currently evaluating the 10-day protocol, in 24-and 96-well formats. We have found that our data are of good quality and are reproducible in-house. We are still evaluating many aspects of the 96-well assay format. Eå ective cell density and the volumes used in each assay format have been assessed. We began by investigating a couple of cell densities. We have evaluated 12.5 K/well (Becton Dickinson protocol) and 25 K/well. Well volumes have also been assessed. Initially 100 ml in the apical ports and 250 ml in the basal ports was used. However, this condition is associated with a suspected wicking problem. The sample under the plate lid as well as the volume moving up the interior wall of the basal compartments has been observed. This is a possible suspect for interwell contamination. We are further developing this assay and will evaluate this phenomenon. We have attempted to cut the volumes from 100 ml in the apical compartment to 80 ml. The basal well we have cut the volume from 250 to 200 ml.
There are many assay conditions that have been published that vary between laboratories. Many publications suggest that a pH gradient from 6.5 (apical compartments) to 7.4 (basal compartments) is the best model, representing the conditions of the lower small intestine. Another assay condition that varies between laboratories is what concentration of drug to use. High concentrations of drug create the risk of saturating the P-glycoprotei n transporte r as well as the other active transport mechanisms. We are currently evaluating our Caco-2 studies in modi® ed HBSS buå er at 5 mm drug concentration using the pH gradient. Again, cell density is a factor that varies between laboratories. Becton Dickinson recommends a cell density in a 24-well plate to be 200 K/well and in a 96-well plate to be 12.5 K/well. We are currently using this recommendation for both assay formats discussed.
Currently, samples are evaluated for P app (apparent permeability) using this published equation: This is a rate distance measurement, where V d is the volume of the donor compartment. A is the surface area of the monolayer, dC is the change in concentration and dT is the change in time.
The Caco-2 cell model mimics the dynamics of the human lower small intestinal tract. The potential for this model to move towards a high-throughpu t format oå ers the high-throughpu t compound screening in the future, which is essential to modern pharmaceutical development. Optimizing the current cell density and volume speci® cations serve to strengthen this assay format, and is currently a major hindrance in its development.

Alger Salt, GlaxoSmithKline, Research Triangle Park, NC, USA
This poster will present a rational approach for choosing extraction parameters, speci® cally homogenizer times and speeds for tablet test methods that employ the TPW II workstation or similar homogenization-base d sample extraction procedures. This approach is based on the assumption that extractions can be modelled using a ® rst-order rate equation. The extraction pro® le, a curve showing the amount of drug dissolved versus time, can be characterized by a single number k, the ® rst-order rate constant.
The amount of drug extracted at a given time can be estimated or predicted if the rate constant is known. Conversely, the amount of time required to reach a given amount of drug dissolved can be estimated. These assumptions allow one to model the process and predict a combination of parameters to ensure that the tablet is completely dispersed and that the analyte is extracted quantitatively. A formal protocol can then be designed and performed to validate the extraction process. The poster will describe this approach and illustrate it with a case study. This presentation will discuss the pressures being placed on quality control and the bene® ts and concerns of introducing new technologies.
It is easy to support the introduction of new technology when the bene® ts can be expressed in terms of dollars saved, but what is the true value of improved regulatory compliance?

High-throughput analysis of free amino acids in biological matrices
Arthur Rugg, Cereon Genomics LLC, Cambridge, MA, USA Co-authors: Shaoxia Yu and Lily Li Traditionally, chromogenic or¯uorescent derivatization (FD) coupled with high-pressure chromatograph y (usually 20± 30 min/separation ) is used for amino acid analysis due the amino acids' lack of UV absorbance. To detect an alternated change in the levels of free amino acids through the screening of thousands of biological samples, the analysis of 20 amino acids using an HPLC/ FD approach is tedious and time consuming. Therefore, a high-speed 1.5 min per sample LC/MS/MS method has been developed, validated and implemented to analyse 20 free amino acidsÐ in biological matrices.

Matt Boeckeler, Neurogen Corporation, Branford, CT, USA
With drug-discovery technology advancing and changing as rapidly as it is, there is a growing need for automated equipment to have the capability to change with these changes in technology. The use of singleworkstation approaches to liquid handling is decreasing due to the increasing need for a system that can perform various liquid-handling tasks on the same platform.
The Staccato TM automated workstation is an extremely versatile liquid handler with the capabilities of handling plate formats ranging from 96-to 1536-well plates. Thè Drag and Drop' architecture of the software allows for exibility and ease in process development. With the use of CLARA TM 2001 software, method development and system recon® guration become less intensive. Its scalability lends itself to supporting changes in platforms as emerging technologies arise. The Staccato TM system can be equipped or con® gured with many components and options to ® t your liquidhandling needs. Using the SciClone ALH as the base liquid handler, the system can come equipped with a disposable tip Rapid head, or a small volume ® xed tip head with 96 or 384 cannula arrays. To augment the liquid-handling capabilities of the SciClone, up to four separate Cavro TM bulk-dispensin g manifolds can be mounted on the head for reagent additions. Liquid handling can be further enhanced by the addition of an eight tip, independent Z-axis probe used for various liquid-handling applications. Integration of the Autostack with the SciClone gives an exceptionally large storage capacity for unattended operation. The use of on-board barcode readers in conjunction with the CLARA TM Data Manager, gives the system the ability to track all compound and assay plates entering the system.
Reformatting chemical libraries with the Staccato TM system is very eae cient. With its large storage capacity and¯exible con® gurations, reformatting can be done through a variety of methods. Using the Rapid head con® guration, 300 96-well plates can be reformatted to 75 384-well plates in one 8-h day. Using the small volume ® xed-tip con® guration, one can convert 200 96-well plates to 50 384-well plates in one 8-h day. Although ® xed tip con® gurations yield less throughput due to implementation of wash protocols, there are no consumable costs. It also gives you the ability to reformat further from 384-to 1536-well plates, with an even smaller volume 384-cannula array designed for 1536 applications.
The independent Z-axis probes provide the ability to cherry pick select compounds, assemble serial dilution plates, and manipulate compounds and reagents for ADMET studies. Performing parallel processing of plate types using the`on the¯y' array swap capabilities, the Staccato TM is capable of reformatting from 96-to 384well plates using the 96-cannul a array. After dropping oå the 96 array, it can then pick up the 384 array and further dilute or deliver compounds directly to assay plates for execution with no human intervention.
The presented pipetting specs are based on ® ve trays run with three transfers performed per tray. A 0.5 ml wet transfer produces excellent precision with an average %CV of 6.4, and an accuracy range of 1.06± 7.3%. A 0.5-ml dry transfer yields an average precision %CV of 8.6, and an accuracy range of 0.03± 10.1%. Dry transfers of 1 ml yield a precision average %CV of 5.8 and an accuracy range of ± 1.5 to 4.4%. Volumes > 1 ml yield similar results to the 1-ml numbers.
System use at Neurogen Corporation entails taking in chemical entities from our existing archive or from our high-speed synthesis group and performing compound dilution procedures, or reformatting processes on the plates and transferring the compounds directly into assay plates, all within the same method. On this platform, the system can also build dose± response plates, prepare compounds for interdepartmenta l support or execute a variety of assays.
After compounds are in the proper format and concentration, with the use of the Cavro TM dispensers, multiple reagent addition to assay plates can be accomplished with the proper scheduling of processes. In essence, the system can take compounds from their starting format and concentrations, manipulate the plates to accommodate assay-speci® c plate formats and conditions, then proceed to add the appropriate reagents via the Cavro TM dispensers, bringing the assay from compound delivery, through reagent addition. Dispensing by the Cavro TM syringes for volumes of 10± 2500 ml all yield similar results. Across plate precision yields an average%CV of 3.8 with an accuracy range of 1.1± 5.5%. Non-contact dispensing of reagents eliminates cross-well contamination and with multiple manifolds, various reagents can be used.
Replication of compound data generated by the 384cannula array at a dry transfer volume of 0.5 ml is quite Abstracts of papers presented at the 2001 ISLAR good. Dry dispensing of such a small volume is not easily achieved, but obtainable with acceptable %CV using particular pipetting methods. Using the same pipetting methods and comparing data to proven control methods for drug delivery yields very acceptable %CVs and correlation.
CLARA TM 2001 Software is used for scheduling methods and integrating peripherals into the existing system. It communicates with adapter modules that translate commands between CLARA TM and the ICP of the peripheral module, triggering program execution of the instrument. This method of integration makes the scalability of the system much easier and more eae cient than previous conventional methods. Upon process completion, the ICP sends a message back to CLARA TM through the adapter to con® rm completion, or send notice of an error in execution. The system' s open architecture, in conjunction with the CLARA TM software, allows for¯exibility and ease when modifying or implementing new assay platforms and technologies.
The CLARA TM Data Manager gives you the ability to manipulate and ® lter incoming data with the use of the Data Manager' s ZyMap. Data can be con® gured into text ® les, Excel sheets or directly transported to a database via ODBC connections. However, the current version of the Data Manager is unable to catch critical eventsÐ events that would trigger the data to be sent to the database, such as an`end of run' event. This inability does not yet allow the direct triggers to export data to a database. A temporary remedy supplied by Zymark is a developer' s kit (a collection of APIs) that gives you the ability to build your own Data Manager using prede® ned templates. Using this method could require extensive time and programming on the user' s end depending on the process and data being captured. However, an updated version of the Data Manager is anticipated to be released at the end of October that would alleviate this problem.
Although the Staccato TM system has many bene® ts and capabilities for liquid-handling needs, there were some implementation issues along the way. The biggest setback experienced was with limitations of the Data Manager. One of the top reasons for the selection of the Zymark system over other vendors' was the ease of database connectivity and data export. Another issue was in the design of the ® xed tip arrays. Great care must be taken not to damage their Te¯on coating. Any scarring or scraping of the material will cause variation in pipetting procedures especially at small volumes. Service for the most part is good; however, the redesign and complexity of the system left the need for more time in training of technicians to come up to speed in the overall workings of the system.
Overall, Neurogens experience with the Staccato TM system has been a good one. All systems have their bene® ts and drawbacks. The most important thing is the cooperation and speed of assistance received when problems do arise and eae ciency in coming up with solutions. In this aspect, Zymark has shown dedication and commitment to their products and customers.
Update from PQRI's Blend Uniformity Working Group: balancing workload with batch homogeneity assurance Garth Boehm, Purepac Pharmaceutical Co., Elizabeth, NJ, USA The Product Quality Research Institute (PQRI) is a consortium between industry, academia, and FDA, which aims to provide a scienti® c basis for the development of regulatory policy. One of PQRI' s ® rst initiatives was to form a Blend Uniformity Working Group (BUWG) charged with providing a scienti® c basis for continued development of FDA policy on BU testing. The BUWG sought wider industry input through a survey of current BU testing practices, and a public Workshop. Using this input together with the experience of the BUWG members and colleagues, a draft proposal has been developed and re® ned. The proposal provides a guide for appropriate testing during scale-up and validation, and for subsequent routine manufacture. The proposal is based around initially establishing the relationship between BU testing and strati® ed testing of the resulting dosage units. The type and amount of testing recommended during routine manufacture depends on the outcome and robustness determined from the validation testing. The BUWG is currently seeking data from industry to challenge the proposal with real data to establish the validity of the suggested approach. The next step will be to seek public comment on the BUWG proposal.
Views from the past: HTS in the Molecular Biochemistry Department, Glaxo Wellcome, Inc.

Brent Butler, GlaxoSmithKline, Research Triangle Park, NC, USA Co-authors: Cole Harris and Steve Blanchard
The pharmaceutical industry saw a tremendous increase in screening throughput in the 1990s with the introduction of advanced robotics and liquid-handling systems, as well as homogeneous assay formats conducive to higher density assay plates. During this time, there have been numerous debates on the advantages and disadvantage s of centralized versus decentralized screening sites. There have also been discussions about the use of large, integrated robotic systems compared with small bench-top robotics.
Glaxo Wellcome Research Triangle Park has screened in a decentralized manner using both large automation platforms and small bench-top systems. In addition, the screens have migrated from 96-to 384-well plates. This talk will focus on the lessons learned from the screening eå ort in the Molecular Biochemistry Department of Glaxo Wellcome.

Fully automated screening of intracellula r calcium using a novel detector
Brian Rasnow, Amgen, Inc., Thousand Oaks, CA, USA Co-authors: Jack Lile, Karen Kearns, James Treanor and Peter Grandsard We have constructed a fully automated high-throughput , cell-based screening system that measures aequorin lumi-Abstracts of papers presented at the 2001 ISLAR nescence kinetics as an indicator of intracellular calcium concentration. The system uses a modi® ed Zymark Rapidplate96 pipetting station to add drug candidates and known modulators to aequorin-expressin g cells grown in microtitre plates. Simultaneous with compound addition, custom optics and a high-performance chilled CCD camera added to the Rapidplate96 take an image sequence of the plate bottom. Compounds that aå ect calcium entry or release from intracellular stores can be analysed using this system. Agonists and antagonists can be detected in the same screen by imaging both during compound addition and during a subsequent addition of known agonists.
The 96-channel imaging¯ash luminometre is integrated with a Beckman CO 2 incubator, Zymark Twister, tip store, ORCA robot, and other peripherals to perform fully automated screening of up to 72 microtitre plates with a cycle time of several minutes per plate (about 10 000 high-content, cell-based assays/system per day). The system is controlled with Amgen' s automation software (`Synchronicity' ). Analogue and digital signal processing are performed in real time to remove artefacts from cosmic rays and bad pixels, and normalize any systematic eå ects from geometrical and biological heterogeneity across the plates. This reduces the about 1.5 MB/ plate of raw data to about 10 kB of time series, which can be further compressed and the compounds classi® ed using principal component analysis and other methods.
We have validated the system for ion channel and GPCR HTS and dose± response characterization, and it has screened thousands of plates. We will discuss and contrast its design and performance with other devices that have comparable capabilities, and explore some of the data and information processing challenges presented by highthroughput, high-content screening systems.

Automated production of medicinal agents from plant matrices
Bruce E. Richter, Dionex Corporation, Salt Lake City, UT, USA Co-author: Richard E. Carlson Interest in the pharmacologically active compounds found in plant tissues is growing. The extraction techniques normally used to remove these compounds from plant tissues require long periods and copious amounts of solvents. Accelerated solvent extraction (ASE) has been proven eå ective in removing target compounds from a variety of plant tissues. Using ASE, the extraction of compounds from medicinal plants is completed in about 15 min using only 20 ml solvent. Extraction of the target compounds may be only a part of the isolation process. Interfering compounds such as waxes, pigments and tannins can be coextracted.
These interferences must be removed before the extracts can be subjected to screening techniques. Solid-phase extraction (SPE), liquid± liquid extraction (LLE) and preparative liquid chromatograph y are the techniques most often used for removing interferences from plant extracts. However, these steps are separate and not coupled with the extraction process. This presentation will discuss the coupling of ASE with liquid-handling apparatus to produce plant extracts free from interfering coextracted compounds. This combined system produces natural product extracts that are ready for biological or chromatographi c assay. The comparison of results from this procedure to those obtained without automation will be discussed. In vitro screening of active compounds for toxic eå ects earlier in the drug-discovery process will help identify problems and reduce the high failure rate of current clinical candidates. Cell-based toxicity assays are crucial in this eå ort, but are limited by the result generated, usually cell death. Thus, we have developed a more sensitive multiparameter cell-based cytotoxicity assay that quanti® es early changes to key aspects of cellular physiology that can lead to cellular toxicity. This assay uses our ArrayScan 1 HCS System High Content Screening (HCS) platform that simultaneousl y measures changes in nuclear morphology, cell permeability and lysosomal physiology for individual cells, and also changes in cell density in microplate wells. Dose± response and time-course data demonstrating the assay' s multiparametric nature will be presented for sample compounds that aå ect the cellular targets in various cultured cell lines and also rat primary hepatocytes. Our data show the assay' s ability to capture cellular reactions to compounds and correlate multiple toxicological indicators at a single-cell level. This assay represents a distinct advantage for HCS as a screening tool early in the drug-discovery process.

Development of a high-throughput, homogeneous, cell-based assay for screening inhibitors of multiple drug-resistanc e pumps on FLIPR or FLEX-Station systems
Christopher Silva, Molecular Devices Corporation, Sunnyvale, CA, USA Co-authors: Kelly J. Cassutt, Anne T. Ferguson and Jesse Twu Multiple drug resistance (MDR) pumps expressed on the surface of normal cells are involved in eliminating toxic compounds generated by or exogenously introduced to the cells. In cancer chemotherapy, these pumps may become over-expresse d by the tumour cells and render the drug treatment ineå ective following relapse. MDR pumps are also normally expressed at the blood-brain barrier (BBB) and as such give the brain`sanctuary ' from chemotherapeuti c drugs.
Our goal at Molecular Devices is to provide a tool to screen libraries for compounds that inhibit the MDR pumps. Such compound leads could be important for inhibiting MDR, which might otherwise lead to failure of the therapeutic regimen. The compound leads identi® ed might also help facilitate the absorption or retention of other useful therapeutics for disease treatment.
The assay is designed primarily for screening inhibitors against the two most common MDR pumps, P-glycoproteins (P-gps) and multidrug resistance-associate d proteins (MRPs). Both classes of pumps are known to expel a variety of mechanistically and structurally unrelated cytotoxic drugs such as anthracyclines, tacanes, vinca alkaloids, and epipodophyllotoxin s as well as some¯uorescence substrates or indicators from the cells. For a given assay the compound library is prepared in a 96-or 384well plate and allowed to incubate with cells at room temperature for 15 min or longer. The microplate is placed in the FLIPR or FLEXStation System followed by addition of the¯uorescence indicator by the instrument. The kinetic readout of the results can then be collected in a real-time mode. The results are expressed as the change in relative¯uorescence units (RFU) over time, usually 5-8 min after the start of the experiment. Inhibition of the MDR pumps is indicated by the increase in RFU above the baseline control.
Using the cell line MCF-7/ADR* (resistant to adriamycin), we showed that the baseline change in counts without inhibitor is nearly zero while 30 mm cyclosporin A (a known inhibitor to P-gp) increased the RFU to > 5000. These results indicate that the P-gp expression level is high enough to keep dye completely out of the cells and the MDR pumps were inhibited eå ectively by cyclosporin A. Various tumour and non-tumour cell lines including the parental MCF-7, Caco-2, and T-cell line CEM were also tested and showed parallel results. Very little or no DMSO interference to the assay was observed.
We have developed a robust and reproducible homogeneous cell-based assay for high-throughpu t screening of compound libraries for inhibitors of MDR pumps. The ease of use inherent in the format with no wash step involved along with the rich kinetic data previously unavailable to the large-scale MDR studies indicate that the FLIPR MDR Assay Kit will be a time-saving and cost-eå ective tool in HTS screening of inhibitors of MDR pumps.

Custom hit-picking robotic system
Claude Dufresne, Merck Research Laboratories, Rahway, NJ, USA Co-authors: Christopher Napolitano and Keith Silverman An in-house integrated automated`cherry picking' system will be described. It is designed to provide backup samples of natural products'`hits' to biologists performing primary screening. The system uses a Tomtec MegaStor for high-capacit y source plate storage. The system is controlled by custom software, written in Visual Basic 6.0, and makes use of ActiveX controls for each of the system components. We have focused on online extraction techniques for high-throughpu t analysis. In a previous study we investigated the potential of online extraction for the analysis of basic and acidic drugs in rat plasma. Recently, we have turned our attention to other types of matrices, e.g. the study of toxicity of drugs in cell cultures used in preand post-discovery phases. Several chemical and physical lysate methods were evaluated. The extracts were injected onto an extraction column at high¯ow rate (i.e. 4 ml min ¡1 ) [1± 3] to remove macromolecular compounds such as proteins, but trap smaller analytes on the head of the column. Several con® gurations for direct injection are possible. In the simplest con® guration, the extraction column is connected directly to the MS/MS system. Other versions are con® gured with a single or a dual extraction column coupled to an analytical column. It is often necessary to split the¯ow. However, in cases where sensitivity is low, this option is not recommended. For eae cient high-speed analysis, the use of a second pump and a 10-port valve is also a good choice. One line (high ow rate) can be dedicated to the extraction column, while the other (low¯ow rate) drives the analytical column and the mass spectrometer.
A three-valve con® guration using two extraction columns was used for the analysis of basic drugs in cell culture lysate. The online analysis was performed on an Oasis HLB extraction column (2:1 £ 30 mm, 25 mm) using a Waters Alliance 2790 TM in the gradient mode and a 515 stand-alone pump in the isocratic mode. The extracted analytes were forward¯ushed into an XTerra column (2:1 £ 30 mm, 3.5 mm), which was added to provide additional separation power. The drugs were quanti® ed using a MicroMass Ultima TM triple-quadrupole mass spectrometer equipped with an electrospray source and set in multiple reaction-monitorin g mode (MRM).
A look to the future: screenin g in the Systems Biology Department, GlaxoSmithKline , Inc.

Cole O. Harris, GlaxoSmithKline, Research Triangle Park, NC, USA Co-authors: Brent T. Butler, Steven G. Blanchard and David C. Morris
Screening at the newly formed GlaxoSmithKline Research Triangle Park site will be heavily dependent on optimizing the successful processes incorporated over the previous 5 years in the Molecular Biochemistry Department at Glaxo Wellcome. This will require expansion of the developing approaches to research aided by successful utilization of existing automation equipment. GlaxoS-mithKline will compartmentalize these processes to exploit the bene® ts of speci® c strategies, further de® ning the process of moving chemistry towards hit identi® cation. This discussion will focus on accommodating a larger infrastructure that provides greater target numbers and larger libraries, as well as complete follow-up screening using the automation resources available.

Courtney T. Ward, MDS Pharma Services, Bothell, WA, USA
Reliable and consistent distribution of cells into microplates is a key aspect of the performance of cell-based screening assays. Typically, this has been done using a hand-held multichannel pipettor in a biosafety cabinet, a task that is both labor-intensive and slow. We have found that this step can be accelerated signi® cantly using a 96channel pipettor such as the Hydra-96 Microdispenser whose physical pro® le permits the use of the instrument in a standard tissue-culture hood. This device consists of an array of 96 glass syringes (290 ml) with ® xed needles where the plungers move in unison under microprocessor control. Using four mammalian cell lines, we compared the pipetting performance of the Hydra-96, a hand-held pipettor, and a disposable tip-based 96-channel pipetting workstation. The results indicate that the well-to-well dispense precision and cell viability are equivalent for all three devices.

Automated o² ine solid-phase extraction for polar new chemical entities using the Zymark Rapid Trace TM systems
Daksha Desai-Krieger, R. W. Johnson PRI, Springhouse, PA, USA A generic automated o¥ine SPE method using the Zymark Rapid Trace TM systems was developed for three very polar new chemical entities (NCEs) undergoing development at RWJPRI, namely a neuroaminidase inhibitor, a cephalosporin and a thrombin inhibitor. The polar characteristics of these compounds made it impossible to carry out liquid± liquid extraction as a sample clean-up step for these compounds.
Our goal was to obtain simple automated SPE methods as sample clean-up for these polar compounds from biological matrices and develop quantitative LC-MS/ MS assays for these compounds. Challenges encountered during assay development and details of the method will be presented. Transfer of these automated o¥ine assays to online automated SPE and/or o¥ine automated 96well SPE will also be discussed. Blood-borne viruses such as HCV, for example, pose a danger of being transmitted from aå ected donor(s) onto the recipients of the blood/plasma or thereof derived pharmaceuticals. The use of nucleic acid testing/nucleotide-ampli® cation testing (NAT) was shown to shorten the`window period' , i.e. a period when donors are in a state of active but serologically undetectable infection. When applied to the plasma pools assembled from the donations destined for the manufacture of the plasmabased therapeutics, HCV-speci® c NAT would decrease the risks of HCV transmission onto the recipients of such therapeutics if virus-contaminate d donation(s) are iden-ti® ed and removed from the fractionation pool.
Accordingly, an automated plasma-pooling process and pooling-driven back-tracing algorithm have been developed and validated at Cangene to support the production of WinRho and VariZig therapeutics bearing lesser risk for HCV transmission. The validated Microlab ATplus 2 automated liquid handler has been used for the preparation of the plasma mega (master) pools in a three-stage process where ® nal mega pools were assembled from the intermediary secondary pools and these, in turn, from the subpools. The automated barcode-reading function of the pooling instrument, in a combination with a suitable barcode-labelling system of the individual plasma donation-and pool-holding containers, allowed for the positive identi® cation of both the individual donations and assembled pools. The pipetting reports generated by the instrument facilitated a full traceability of all pipetting events whereas gravimetric in-process veri® cation assured pipetting accuracy.
A pooling scheme-driven back-tracing algorithm has also been developed and validated for singling out the individual contaminate d donation(s). This algorithm involved four-stage back-testing and was based on the assumption that all NAT-positive test results are truepositives. It allowed for identi® cation of viral contamination of a single donation by testing 26 samples, i.e. one ® nal pool, ® ve secondary pools, 12 subpools and eight individual donations. The developed and validated processes allow for cGMP-compliant preparation of the plasma pools and regulatory required back-tracing of the viral contamination to individual plasma donation(s). Assay miniaturization and the implementation of highdensity 1536-microwell screening is vital to our need to increase the eae ciency of primary screening and early phase lead discovery. To serve this need, we have developed a 1536 uHTS platform that employs microwell plates, novel¯uid handling and optical detection technologies. Full use of this platform requires the ability to develop assays eae ciently for a variety of biological targets in a straightforward and expeditious manner. Having a portfolio of miniaturizable assay technologies that can be easily developed into ultrahigh-throughpu t screens for any member of a class of biological targets would enhance and extend the utility of this technology. We have evaluated a variety of assay technologies that are suitable for uHTS of kinase and GPCR targets. This presentation will discuss the relative merits of these technologies in assembling a portfolio of uHTS assays for screening these target classes.

Prequali cation of pharmaceutica l leads
David Casebier, ArQule, Inc., Woburn, MA, USA Reliable predictive models and increased throughput of ADME screens currently enable the design, construction and pro® ling of chemical libraries, providing con® dence for series selection in hit-to-lead discovery. Predictive models for metabolism, absorption, BBB as well as diversity measurements assist in library design and reagent selection. Robotic syntheses then generate spatially addressed, high-quality libraries, which in turn are pro® led for ADME characteristics. Integration of these technologies allows for the prioritization of lead series using data indicative of developmental survivability.

Automation of identi cation and screening of genomic targets related to psychiatric disease
David M. Evans, Psychiatric Genomics, Inc., Gaithersburg, MD, USA Psychiatric Genomics, Inc., is interested in discovering the underlying genetic causes of psychiatric diseases and identifying therapeutics for them. Many of these diseases are multigeneic in nature and have not previously been well characterized owing to the diae culty in obtaining the appropriate tissue samples and the lack of technologies that could detect the changes in gene expression pro® les within the brain regions.
In recent years, several tools and technologies have been developed that allow multiple parameters to be tested in a single experiment. Micro-array technology allows the discrimination of changes in many genes at one time, and by comparing`normal' versus diseased tissue, it is possible to identify gene expression patterns quickly that are altered due to the disease. Having identi® ed the disease gene pro® le, it is important to be able to develop screens to examine whether small chemical compounds can revert the gene pro® le of the diseased state back towards normal' . This presentation will describe some of the approache s being taken by Psychiatric Genomics, Inc., to automate each of the steps in the drug-discovery process using the new paradigm of automated drug discoveryÐ`Multi-Parameter High Throughput Screening' (MPHTS TM ).

Comparison of ELISA between automation and manual testing for measuring ARANESPTM in rat serum
De Chen, Amgen, Inc., Thousand Oaks, CA, USA

Co-authors: Yan Wang, Monica Zordich, Han Gunn and Sharon Baughman
This presentation compares the results of measuring ARANESPTM in rat serum, obtained manually or via a fully automated immunoassay system. ARANESPTM is a new erythropoietic protein currently used in clinical trials for the treatment of anaemia related to chronic renal failure.
The Tecan Genesis workstation 200 was developed to perform the ARANESPTM quantitative ELISA. It was equipped with eight pipetting ceramic needles capable of pipetting 5± 1000 ml, a shaking incubator, a Columbus microplate washer, microplate hotels and a microplate reader. A robotic arm transported microplates and lids between the peripherals on the workstation deck.
Seven independent studies were performed on various days by the TECAN and manually. Standards, quality controls and known samples were loaded in triplicate on the plates in each assay. The data generated showed no signi® cant diå erence. ANOVA analysis con® rmed this conclusion.
Fully automated systems can be introduced to quantitative immunoassay for protein measurement in serum. Technical advances in methodologies, robotics and computerization will lead to signi® cant enhancements in the capacity of immunoassay, resulting in signi® cant cost and eae ciency gains. Extraction of plasma samples was performed using IST Phenyl SPE cartridges (200 mg 3 ml ¡1 ) conditioned with methanol and water. Sample clean-up was achieved by rinsing cartridges with methyl-t-butyl ether. The compounds of interest were eluted using ethanol:water (90:10 Abstracts of papers presented at the 2001 ISLAR v/v) and subsequently evaporated under nitrogen to dryness. The residue obtained was reconstituted before injection on an LC-MS/MS in the positive-ion APCI mode.

LC-MS/MS assay for the quantitation of RWJ-270201 using the Zymark Rapid Trace SPE workstation
The developed assay demonstrate d good precision and accuracy and was linear over a broad curve range. The method developed for these analogues was routinely used to support preclinical pharmacokinetic studies and was later adapted for use with 96-well plate technology in support of clinical studies. The need arose in our department to develop a faster and safer yet user-friendly means of sample preparation before sample analysis. With shorter mass spectrometer method run times, the rate-limiting factor was the sample-processing procedure. To expedite this procedure, automation through robotics and 96-well plate technology was implemented. Not only is this robotic protein precipitation assay process faster than manual preparation, but also the precision and accuracy are comparable. In addition, our staå has less exposure to potential pathogens in the samples, therefore increasing safety. This robotic system is operated by very userfriendly software that enables the processing of samples either in a fully or semi-automated manner. The system can manage 96 samples in < 1:5 h.

Automated o² ine solid-phase extraction for polar new chemical entities using the Zymark Rapid Trace systems
Daksha Desai-Krieger, R. W. Johnson PRI, Springhouse, PA, USA A generic automated o¥ine SPE method using the Zymark Rapid Trace( systems was developed for three very polar new chemical entities (NCEs) undergoing development at RWJPRI, namely a neuroaminidase inhibitor, a cephalosporin and a thrombin inhibitor. The polar characteristics of these compounds made it impossible to carry out liquid± liquid extraction as a sample clean-up step for these compounds. Our goal was to obtain simple automated SPE methods as sample clean-up for these polar compounds from biological matrices and develop quantitativ e LC-MS/MS assays for these compounds. Challenges encountered during assay development and details of the method will be presented. Transfer of these automated o¥ine assays to online automated SPE and/or o¥ine automated 96-well SPE will also be discussed.

. Tsinman and A. Avdeef
We have successfully adapted the classical shake-¯ask method to the 96-well microtitre plate format to obtain near shake-¯ask-quality results at high-throughpu t speeds. The new instrument uses a robotic¯uidics system and a parallel detection system employing a microtitre plate UV (190-500 nm) spectrophotometer . Samples are introduced as DMSO stock solutions.
The essence of the approach is based on making UV concentration reference solutions from moleculesÐ especially ionizable moleculesÐ whose UV spectroscopic properties are unknown at the start of the assay. This can be achieved in either of two ways, with or without cosolvents.
A unique computational method was developed to extract the aqueous intrinsic solubilities of drug molecules from data distorted by DMSO drug binding or drug± drug aggregation reactions.
An improved method for determining concentration by UV spectrophotometr y was derived. It uses a novel peakshape algorithm for adjusting weights in a whole spectrum-weighted regression analysis, matching spectra of reference solutions (of known concentration, under conditions avoiding or suppressing precipitation) to solutions containing an analyte of unknown concentration (due to precipitation).
A quality-assessment scheme has been developed around the standard drugs diclofenac, indomethacin,¯urbiprofen, chlorpromazine, phenazopyridine, verapamil, piroxicam and griseofulvin.
The new solubility method, which allows determination at one pH or as many as 96 pHs, has a limit of detection of about 0.1 mg ml ¡1 , as demonstrated by measurements of the intrinsic solubilities of terfenadine and tamoxifen.
About 200 assays may be performed per day on the instrument.

Christoph M. Huwe, Schering AG, Berlin, Germany
At Schering we have established a¯exible, integrated solution-phase synthesis, puri® cation and reformatting work¯ow based on Chemspeed-and Zymark-automate d synthesizers, parallel normal-phase chromatograph y as well as reversed-phase HPLC-MS puri® cation equipment, and a highly automated Zymark reformatting robot. In this presentation, the special features of the chosen equipment, the ideas behind the system design and the respective work¯ows will be discussed.

Douglas R. Myers, Intelligent Laboratory Solutions, Naperville, IL, USA Co-author: Je¡rey D. DeCicco
Developing new HPLC methods is a time-consuming and knowledge-intensive task. We have developed an online software package called ChromSmart that automates HPLC method development for both normal and reversed-phase chromatograph y for chiral and non-chiral sample mixtures. Unlike other software packages that assist with o¥ine HPLC modelling, ChromSmart uses online arti® cial intelligence technology to capture and implement the method-developmen t knowledge of expert chromatographers . ChromSmart contains a real-time intelligent experiment planner and rule-based engine that execute the HPLC methods, monitor results and dynamically update the development plan. A complete audit trail of the decision-making and results is generated.
ChromSmart comes with a core set of chromatograph y knowledge that automates method development. Based on user preferences, such as retention times and resolution, ChromSmart develops a method-developmen t strategy. This strategy is executed and can dynamically change based on information from online, real-time experimental results. Features include automated equilibration, detection and cleaning of retained components, gradient and isocratic optimization, peak purity analysis, and peak tracking. When ChromSmart ® nds a method that meets or exceeds the user' s criteria, it noti® es the user and continues to develop another method for the next sample. ChromSmart operates 24 £ 7 £ 365 signi® cantly increasing productivity of equipment and personnel.

Je¡rey B. Medwid, Wyeth-Ayerst Research, Pearl River, NY, USA Co-authors: Jerry Edgar, Syed Rahman and Scott Keenan
In the fast-paced pharmaceutical business, quicker turnaround times and larger sample loads face every analytical laboratory manager. Unfortunately, this must be accomplished with no signi® cant increase in analyst resources. Laboratory automation is the only answer. Our research automation laboratory contains several Zymark robots including the Multidose Dissolution Workstation s and Tablet Processing Workstations . The latest addition to our laboratory is a UV Fiber Optic Dissolution Workstation manufactured by Delphian Technology. This presentation will highlight some of our recent work accomplished using these workstations and some of our other initiatives in automation.

Phil Small, Tripos Receptor Research, Bude, UK
A process-integratin g design, synthesis and analysis of combinatorial libraries has been implemented at Tripos Receptor Research. Crucial to the success of this process has been the development of a proprietary informatics system. This has so far been developed to manage reagent inventory, track samples and record data from synthesis and analysis to provide a valuable database of compound information.
The combination of Tripos' proprietary design software with automated synthesis and informatics leads to the production of drug-like libraries with well-de® ned purity and full synthetic history. This presentation will cover what were considered the important aspects involved in setting up and managing an automated chemistry facility.

E. Michael August, Boehringer Ingelheim Pharmaceuticals, Inc., Ridge¢eld, CT, USA Co-authors: Lori Patnaude and Carol A. Homon
Screening throughput is limited by its slowest step, and thus a major goal of uHTS must be to identify and overcome these bottlenecks. With the advent of highdensity liquid-handling devices capable of delivering liquid to 384 or more wells at rates approaching 1 plate min ¡1 , the bottleneck has shifted from reagent delivery to plate reading. Conventional plate readers require several minutes to obtain data from an entire plate.
Imaging technology allows the detection of an entire plate simultaneously. Coupled with rapid, sensitive image-analysis software, such a system enables data capture to keep pace with reagent delivery. We present here screening results from a recent enzymatic Del® a uHTS campaign comparing the Wallac Viewlux Imaging system with the LJL Analyst. Several hundred plates in both 96-and 384-well formats were read on both readers and the data were analysed with a variety of statistical parameters.
Similar comparisons of other assay formats will be discussed. The inherent¯exibility of imaging systems is highly compatible with assay miniaturization and the trend to higher-density plate formats. However, we must also realize that the bottleneck in the overall screening process has not been removed, but surely shifted to another step.

Finally the paperless and compliant laboratory
Ed Halpin, VelQuest Corporation, Hopkinton, MA, USA Compliance and cost have long been at odds in the modern regulated laboratory. Focusing on one meant sacri® cing the other. Regulatory compliance today is a Abstracts of papers presented at the 2001 ISLAR labour-intensive, paper-based operation. The`paperless' laboratory has been a goal of the pharmaceutical industry for over the past decade. The con¯uence of technology and innovation has reached a point where the paperless laboratory is now economically and technologically practical. Good manufacturin g practices and 21CFR Part 11 de® ne the requirements to attain a paperless, compliant laboratory.
The solution would enable direct electronic data capture from human observations , instruments and networked PCs replacing paper notebooks, ® les and control sheets. The solution would enhance the collection of data and metadata by dynamically linking it to the procedure by which it was collected. Reviews, audits, investigations and approvals would be electronic processes, replacing the current paper systems. Redundant checking would be replaced by`review by exception' and resources reallocated towards the real goal of bringing new human therapies to the market more rapidly. The solution must also complement and exchange data with other laboratory applications such as Laboratory Information Management Systems (LIMS) and Chromatograph y Data Systems (CDS).

Dragos Horvath, 128 rue Danton, Rueil Malmaison, France Co-author: Michael Entzeroth
Current drug development is characterized by a strong belief in high-throughpu t methods, both in combinatorial chemistry and screening. The trigger for constantly increased performance is the need the industry to adjust to the race against time in order to secure shares on the market place. Over many years the industry has tried to keep pace by the constant introduction of new technologies. The associated costs for drug development increased over the last years. Aggregated R&D expenditures for new chemical entities (NCE) peaked in 1996 close to US$500m. 1 Recently, Lehman Brothers estimated the total costs for a launched product to account up to US$635m, 2 while for drugs entering clinical trials between 1972 and 1982 these costs were estimated to add up to US$312m. 3 Research-based companies will invest $26.4 billion in R&D in 2000, a 10.1 percent increase over 1999. 4 The number New Chemical Entities approved by the FDA has not changed signi® cantly over the last years (1997: 39, 1998: 30, 1999: 35), indicating that the success rate is not related to the initial eå orts.
The portion of IND applications that fail has been estimated to approximatel y 20% in phase II, 60% in phase II and most importantly 87% in phase I. 5 83% of the drugs for which INDs were ® led between 1964 and 1989 were dropped before reaching NDA status. 6;7 For the pharmaceutical companies each dropout during the development is associated with huge costs that reduce the pro® t derived from NCEs that reached the market. The failure of the candidates can be primarily attributed to eae cacy, safety and economic concerns. The predominant reason for failure, however, are inappropriate pharmacokinetic properties of the drug candidate (39.4%), followed by adverse reactions and toxicity that add up to 21% of the causes for withdrawal from development.
In most of the cases the lack of a complete picture of the drug' s properties is not available at the time the decision to select a respective candidate is taken. In part this is due to the current process of drug discovery where information is accumulated stepwise rather than in a de® ned parallel manner. The choice of the candidate taken forward is often dominated by emphasizing the aspect of potency rather than addressing early on selectivity and drug property issues. On the other hand, the companies have accumulated vast amounts of in-vitro and in-vivo data that may be used to re® ne the decision matrix either by retro-or prospective interpretation of existing results. As a consequence and in order to take the serendipity out of such fundamental decisions, many pharmaceutical companies have started broad range pro® ling (i.e. testing the drugs in parallel against a panel of diå erent in-vitro model, both in pharmacology and ADME/PK) in the preclinical phase and have put emphasis on the development and automation of secondary tests. High throughput pro® ling using robotics and automated workstations allows to accumulate today a broader knowledge on the hits identi® ed in the primary screening. Both pharmacologicalÐ speci® city and selectivity versus other targets, such as enzymes and recep-torsÐ and pharmaceutical propertiesÐ physicochemical parameters or metabolic and in-vitro safety aspectsÐ of the candidates are evaluated in depth. 8 Through the progress in linking of high-throughpu t pharmacological, physicochemical and`in-silico' methods with in-vitro approaches, progress is being made in predicting drug properties. 8 The goal to predict in-vitro and, ® nally, in-vivo properties of drugs has become reasonable by linking the chemical structure via molecular descriptors, area descriptors or ComPharm ® elds to certain in-vitro pro® les that again can be correlated to the corresponding in-vivo eå ects (® gure 1).
An example for pharmacophor e ® ngerprints are the Fuzzy Bipolar Pharmacophore Autocorrelograms (FBPA 9 ) monitoring the numbers of atom pairs within each of the 252ˆ21 £ 12 categories that can be de® ned in terms of the 21 combinations of six pharmacophoric features a,b2{Hydrophobicity , Aromaticity, Hydrogen Bond Acceptor & Donor, Cation, Anion} times 12 considered distance ranges ¢2{(3..4), (4..5), . . . , (14..15)} [A ® ]. For similarity searches and structural comparisons, these ® ngerprints are extremely powerful in combination with ComPharm ® eld descriptors 10 that are taken as the intensities of the empirically de® ned pharmacophor e ® elds' generated by the atoms of the considered compound at the space points occupied by a reference structure, in a con® guration corresponding to its optimal alignment with respect to this reference.
The data mining approach described in this report is part of the BioPrint TM project 11 that makes use of a broad spectrum pro® ling of marketed drugs and candidates in more than 90 diå erent in-vitro models. The marketed drugs, included in the test collection, comprise more than 1,500 compounds that are available in the US pharmacies today. Important are also the drug candidates Abstracts of papers presented at the 2001 ISLAR included which have failed to make it all way through the diå erent phases of preclinical and clinical drug development. The models and predictive tools generated open a new approach for candidate selection and early discovery activities, such as target identi® cation or library design and will result in the long term in substantial cost savings due to reduced failure rates and time spent on the research and development process. In-silico, linear models to describe solubility and permeability as well as predictions in pharmacology based on predictive neighborhood behavior are presented.

Linear ADME models
The interest in an in-silico logD prediction model resides both in its applicability in library design and its further use as a calculated molecular descriptor in other structure-activity relationships, if the activity under study is expected to relate to the overall lipophilicity of com-pounds. By contrast to logP models that mostly rely on incremental group contributions and/or molecular descriptors of the average polarity of the solvent-accessible surface (an aspect which is in our model successfully captured in terms of PTA descriptors), a logD model needs to account for proteolytic eå ects in the aqueous buå er of pHˆ7.4 used for experimental determinations. In our models, proteolytic ionization of acids/bases is implicitly accounted for by the pharmacophor e descriptors relying on rule-based assignments of a positive/ negative charge status to ionizable groups. Furthermore, bipolar pharmacophor e elements or ® eld overlap terms successfully account for potential pKa shifts due to neighboring groups of the ionizable center.

Linear apparent permeabilit y models
The main challenge in predicting the apparent permeability of compounds through a Caco-2 cell monolayer is

Adverse effects Therapeutic effects
In Vivo Effects ADME/PK Adverse effects Therapeutic effects

In Vitro Profile
In Vitro tests Pharmacophore Pharmacophore

In Vitro
In Vitro ADME ADME--T T properties properties

In Vitro
In Vitro ADME ADME--T T properties properties Figure 1. Predictive modelling of the properties of compounds correlates chemical structures via molecular descriptors with ADME/ ë K and therapeutic e¡ects. the in-silico recognition of molecules which are transported by one of the active in-or e¥ux mechanisms. This will enable the introduction of speci® c correction factors explaining the diå erence between the actual membrane crossing rate and the rate expected on the basis of purely passive transmembrane diå usion. While the latter aspect of predicting the passive diå usion rates has been tackled in various publications, little progress has been made to understand the problem of transported/ e¥uxed compounds, routinely discarded as`expectable' outliers in previous work. By contrast, the herein described modeling eå ort focused on pharmacophor e descriptors in order to search for potential pharmacophor e motifs that may characterize e¥ux and/or transportation, while descriptors of the molecular surface properties and the generic calculated logD descriptors were supposed to account for passive membrane crossing properties. The calculated logD (or, even better, the experimental logD values) are relevant variables of the linear log(P A± B ) apical-to-basa l apparent permeability model. A positive coeae cient for the linear contribution, together with a negative term in logD 2 characterize an optimal logD range maximizing the passive membrane crossing rate. This makes physical sense, since too hydrophilic compounds never leave the aqueous phase, while too hydrophobic ones tend to accumulate in the membranes.
In spite of the important number of bipolar pharmacophore elements entering the model, a series of heavily e¥uxed compounds are nevertheless mispredicted by the model and appear as`false permeable' . Most of the outliers in the marked area are shown (® gure 3, circled area) to be indeed e¥uxed molecules, since their measured basal-to-apica l apparent permeabilities were sig-ni® cantly higher that the corresponding apical-to-basa l values.
It can be therefore concluded that either (a) e¥ux cannot be understood in terms of bipolar pharmacophor e descriptors only or (b) e¥ux might in principle be characterized in terms of such descriptors, but the number of descriptors that would be required to enter the model is much larger than the number of examples of e¥uxed compounds currently included in the data setÐ note that multiple e¥ux/transportation mechanisms with potentially diå erent pharmacophor e characteristics coexist in Caco-2 cells, and that enough compounds exemplifying each one of them would be required to solve the problem.
The introduction of ComPharm descriptors overriding the limitations of bipolar pharmacophor e terms is indeed seen (® gure 3) to signi® cantly improve the permeability prediction of the outliers from the previous model. These ® eld descriptors were obtained on the basis of superimposition models of BioPrint TM compounds against some of the most heavily e¥uxed/transported substances encountered in this set.

Solubility model
A categorical solubility model based on a linear equation has been developed (® gure 4) in order to predict the solubility class of the compounds: (1) Low: S < 10 mM; (2) Medium: 10 mM < S < 100 mM; (3) High: S > 100 mM. This model (not shown here) included 12 explaining variables, such as logD and logD 2 and various Pharmacophore Type Areas (PTA) and bipolar pharmacophore elements that predicts drug solubility from its structural features. With encouraging results it has also been compared with other solubility prediction software packages and shown its advantages with 86% of the drugs predicted in the correct solubility class.

Predictive neighborhood behaviour (PNB) models
By contrast with linear approache s that require a speci® c equation for each of the properties that are to be predicted in terms of molecular descriptors, PNB models rely on the measured properties of related compounds (e.g. nearest neighbors in descriptors space) from the BioPrint set in order to provide an estimation of the property value of a new molecule. The PNB algorithm works in principle like a`fuzzy data base querying tool' in the sense that if molecule X is not included in the BioPrint database, the BioPrint compounds that are most similar to X are retrieved and used for the evaluation of the properties of X. The property of the unknown molecule is computed as a weighted average of the experimentally determined properties PN(x) of its nearest neighbors N(X) from the BioPrint set, where the weighing factors are decreasing functions of the molecular dissimilarity scores that de® ne how`near' a neighbor N(X) is with respect to X. These dissimilarity scores are Dice scores calculated on the basis of FBPA, PTA and EFO descriptors, the intervening empirical parameters being speci® cally ® ne-tuned in order to maximize the predictive power with respect to a particular property. Figure 5 shows the example in the case of SCH23390, a D1 receptor ligand, which was not in-cluded in the original data set, however, tested later for comparison.

Conclusion
During the last decade technologies such as ultrahighthroughput screening and combinatorial chemistry have signi® cantly contributed to the advances in drug discovery. The increase in R&D costs associated and the bottlenecks further down the development as well as the high dropout rates have increased the demand for an alternative approach. BioPrint TM combines molecular descriptors, high-throughpu t pro® ling with in-vivo drug pro® les. It takes advantage of data mining technologies to generate predictive models that link the three components. These models help to build a eå ective strategy that will foster drug discovery in the new millenium.   One of the most fundamental routines for a chemist is the preparation of samples from dry powdered samples.
There have been many expensive attempts to automate this process. We demonstrat e powder dispensing using an existing robot workstation and some common disposables.
Powder dispensing was accomplished by building a dispensing station on an integrated weighing robotÐ the Zymark Benchmate. The operating principle of the station is based on applying a vacuum to an airline connected to a pipette tip. The system is designed to transfer dry powder from four-dram vials to 16 £ 100 test tubes. The user can specify target weight and replications for each compound through a custom-designed Visual Basic program.
The sample vial is basically carried to the ® xed location of the pipette tip, placing the tip into the powder. The vacuum is applied to the airline and the vial removed. The vacuum holds a small amount of sample powder on the pipette tip. A test tube is then moved into position below the pipette tip, the vacuum is released and the powder is dispensed into the 16 £ 100 test tube. When the target weight is reached, sample preparation (e.g. dilution to concentration and mixing) can then be accomplished on the same workstation.

uHTS of phosphatases using 2D-FIDA anisotropy detection
Flora Tang, Sugen, Inc., South San Francisco, CA, USA Various parameters of¯uorescence measurement, such as translational diå usion, brightness and rotation, have been widely recognized as readout fully amenable to homogenous, miniaturized uHTS. A novel detection system using Evotec' s FCSplus reader, which combines the¯uorescence-intensity distribution analysis (FIDA) anisotropy detection with confocal microscopy, provides a very sensitive and ultrahigh-throughpu t assay format. The collaboration between Sugen and Evotec OAI is focused on phosphatase s and has demonstrated successful adaptation of 2D-FIDA anisotropy detection for phosphatase assays at the 1 ml/well level. A discussion of this assay format and the uHTS screening results on the automated EVOscreen Mark II system for several phosphatases will be covered.

G. Bryan Crist, VanKen Technology Group, Cary, NC, USA
Among the future challenges for the pharmaceutical laboratory exists the necessity to automate dissolution methods to allow maximum throughput with optimum compliance. This presentation will provide suggestions for determining the level of automation needed to obtain peak performance as well as providing insight into regulatory compliance issues. Associated topics to be discussed include: novel dosage forms today and providing analytical challenges in the future, alternate methods of analysis, the dissolution apparatus of the future and the involvement of new analytical approaches. Instrumenta-tion¯exibility will be required in the dissolution laboratory of the future to utilize fully pooled sampling techniques, online UV, online HPLC, capillary electrophoresis and even atomic absorption for nutritional analysis.
Semi-automatio n to total automation concepts including in situ methods of analysis will also be discussed. As novel concepts evolve each day, engineers working with scientists will lead to more eae cient, better-integrated dissolution instrumentation in the future.

Update from PQRI's Blend Uniformity Working Group: balancing workload with batch homogeneity assurance
Garth Boehm, Purepac Pharmaceutical Co., Elizabeth, NJ, USA The Product Quality Research Institute (PQRI) is a consortium between industry, academia, and FDA that aims to provide a scienti® c basis for the development of regulatory policy. One of PQRI' s ® rst initiatives was to form a Blend Uniformity Working Group (BUWG) charged with providing a scienti® c basis for continued development of FDA policy on BU testing. The BUWG sought wider industry input through a survey of current BU testing practices and a public workshop. Using this input together with the experience of the BUWG members and colleagues, a draft proposal was developed and re® ned. It provides a guide for appropriate testing during scale-up and validation, and for subsequent routine manufacture. The proposal is based around initially establishing the relationship between BU testing and strati® ed testing of the resulting dosage units. The type and amount of testing recommended during routine manufacture depends on the outcome and robustness determined from the validation testing. The BUWG is currently seeking data from industry to challenge the proposal with real data to establish the validity of the suggested approach. The next step will be to seek public comment on the BUWG proposal.
Abstracts of papers presented at the 2001 ISLAR

Potential impacts of implementing automated dissolution
Gerard Schneider, LEC Consulting, Blairstown, NJ, USA The Sotax AT-70 Smart oå ers the ability to automate fully both USP Apparatus 1 and 2 dissolution methods. This system, used in conjunction with other eae ciencydriven technologies such as premanufactured medium and online spectrophotometri c analysis of samples, can profoundly reduce the resources required for conventional dissolution testing while enhancing the consistency of results. Most importantly, these eae ciency gains can be achieved within the parameters of a compliance-conscious industry.
The eå ects of implementing such technology will be discussed. The discussion will address the eå ect that eae cient, automated dissolution testing has on the product development area, where companies are focussed on delivering products to market in increasingly less time, as well as marketed product stability and release laboratories where QC managers are asked to reduce sample turnaround times without increasing resources. Examples of eae ciency gains will be given in both cases.

Gladys Range, Human Genome Sciences, Rockville, MD, USA
In the automation of any process, there are basic steps that when carefully suited and applied to a particular laboratory assay or work¯ow, serve as a tool for eå ective automation. Biological validation, time and highthroughput screening assays are, by nature, very diå erent, and are designed with diå erent criteria in mind. A biological assay is a complex system designed to yield biological activity using qualitative measurements. Issues of cost, amount and quality of reagents, although important, do not play a paramount role in development. The biological assay is usually a manual assay that requires signi® cant change before it is automation ready.
High-throughpu t assays in contrast are speci® c, fast, reproducible, low standard deviations, low cost and usually can be automated as much as mechanically and humanly possible.
Key elements in assay automation are sample sourcing, screening design, robotics' hardware and data management. These elements need individual careful analysis, a master plan and follow-up. A challenge in automation is to enable individual laboratory personnel to contribute to the ® ne ongoing activity of building a database by doing their work without stopping to write in a laboratory book or even to enter information in a computer. These in return will prevent redundancy and will improve accuracy and completeness.
Second-generatio n laboratory robotics and liquid handlers are now designed to handle very high-throughputs , very dense formats and very small-volume applications. Likewise, the recent development of sophisticated¯uorescence systems operating at longer wavelengths, using time-resolve signals, and ® breoptics' technology reduce read time and facilitate moving away from less desired chemistries and separation steps. While some of the very large complex robotics systems could automate the complete sequence of operations comprising an assay, a higher level of productivity is achieved by paying careful attention to the work¯ow. Determining the optimum assay conditions, standardizing some of the operations and selectively introducing automation bring in the best results and bene® ts. The quantitative transfer of biological¯uids is the most critical aspect of automated sample preparation. The sampling of plasma, serum, blood or urine represents the most diae cult and time-consuming step in most bioanalytical extraction procedures. We will present two practical solutions that address these separate problems: . automating`diae cult' sample transfers with higher quality; . signi® cantly increasing the processing speed of most simple' sample transfers.
First, most laboratorie s automate sampling on a multiple-tip liquid-handling workstation such as a Packard MultiPROBE 1 or Tecan Genesis TM . Today, samples are most commonly transferred from various size tubes to a deep-well plate or a 96-well solid-phase extraction (SPE) block. Despite their widespread use, these systems cannot perform sampling with suae cient precision and accuracy for certain applications (e.g. samples containing clots or other insoluble matter; samples with low or insuae cient volumes).
We will show how the integration of an analytical balance on an x-y-z liquid-handling deck can easily improve the overall accuracy of assay data. The application also allows the workstation to process samples that would otherwise need to be transferred manually. Samples are aspirated in parallel then dispensed one at a time, allowing the balance to capture the exact weight (ˆ> volume) of each sample that was added to a deepwell or SPE plate. The data are used to correct automatically for any diå erences in sample weights, usually by pasting the results directly in the`dilution factor' column of an analytical system' s worklist.
Second, for most other applications in which reliable transfers are easily achieved without gravimetric con-® rmation, the speed of sample processing remains a ratelimiting step. Most systems require about 20± 45 min to transfer 96 samples. This time is signi® cant and limiting compared with the 1± 2 min required for other steps such as the addition of a reagent using a 96-channel pipetting station (e.g. Zymark RapidPlate TM -96, Tomtec Quadra 96 TM , Apricot PP550 TM ).
We will detail how custom sample racks can be used to speed up tube-to-plate sample transfers dramatically.
Abstracts of papers presented at the 2001 ISLAR These novel racks have a standard microplate footprint and can be used on any 96-channel pipetting station. The racks, which consist of complimentary pairs, each hold 48 sample tubes arranged in an alternating (staggered) sequence. With these racks, 48 samples are transferred simultaneousl y and 96 are transferred in as little as 2 min. The racks work with most brands of 13-mm outer diameter cryotubes and 11-mm outer diameter microcentrifuge tubes commonly used in bioanalytical laboratories.
With these two inexpensive workstation modi® cations we can now automate 100% of our applications and sig-ni® cantly speed up the majority of these assays. Although the sequence of the human genome has been determined, the regulation and function of many genes remain largely unknown. Cytokines are tightly regulated soluble proteins transiently produced by cells in response to immunological stimulation or disease manifestation. The expression of these proteins has been linked to many diseases including cancer, AIDS, obesity, autoimmunity, immunode® ciency and infertility, suggesting that cytokines, their receptors, as well as cytokine-induced genes are potential targets for new drug development. Microarrays such as gene chips and protein arrays are highthroughput technologies for drug discovery in the postgenomic era.

Identi cation of cytokine-regulate d genes associated with infertility and obesity by DNA gene-chip micro-arrays
We employed DNA transcriptiona l pro® ling micro-arrays to detect genes under control of cytokines in selected experimental models. Hence, in ovarian granulosa cells, TNF stimulates DNA synthesis but inhibits FSH-induced oestrogen production. These results suggest that identi®cation of TNF inducible`master control genes' may reveal new targets for infertility and possibly other diseases associated with oestrogen. TNF induces more than 70 genes in rat ovaries in vivo, including genes for cytokines, kinases, receptors, transcription factors and enzymes. In contrast to its inhibitory eå ect on oestrogen production in the ovary, TNF stimulates production of this sex steroid in human pre-adipocytes and adipocytes derived from more than 10 patients. Interestingly, human pre-adipocytes/adipocytes themselves produce TNF as well as other cytokines and our data suggest a positive autocrine/paracrine mechanism for control of oestrogen production by fat cells. In addition to TNF, 16 cytokines including LT-a, IL-1, IL-6, IL-11, LIF and oncostatin M all stimulate oestrogen production in human fat cells.
These results indicate that TNF has diå erential and cell-speci® c actions on key endocrine mechanisms. Iden-ti® cation of genes induced or inhibited by TNF and other cytokines may lead to the discovery of novel targets for drug development. The three-dimensional structure of any protein plays a key role in the process of understanding the exact function of these macromolecules. One proven approach to structural information of proteins is based on X-ray diå raction of single crystals. In the past, automated crystallization was restricted because a reliable hardware platform and well-suited microplates for high-throughpu t screening were missing.
The Max-Planck-Institut e for Molecular Genetics (MPIMG), the Protein Structure Factory (PSF) and Greiner Bio-One have collaborated to develop a unique 96-well protein crystallization microplate (Crystal-Screen TM ) with a standardized microplate footprint for high-throughput applications. Each of the 96 mother liquor wells corresponds to three crystallization wells. This allows checkerboard screening with up to 288 crystallization options per plate to investigate optimal crystal growth followed by 3D-structure analysis. In combination with a preformed lid, this microplate enables high-throughpu t multichannel screening of proteins in automated systems.
The system in use at MPIMG and PSF allows simultaneous sitting-drop and hanging-dro p vapour diå usion crystallization experiments at reduced costs. A huge storage system for 10 000 crystallization microplates and a pipetting device based on solenoid inkjet technology enables one to set up a complete plate with 96 crystallization conditions in < 3 min. All crystallization wells are inspected in regular intervals with an automated camera-based detection system to inspect the crystal growth. potential to result in signi® cant bene® ts including resource liberation and an electronic compliance platform for process management and data review.

Seamless integration of information in a pharma
This paper will describe the process of working with four vendors to create a complementary data-managemen t solution to Purdue Pharma' s needs. The process of logging in samples, creating worklists, executing analytical methods, capturing data electronically, storing electronic records, and submitting reports will be reviewed. After data creation, electronic data review will be discussed enabling electronic`instant replay' of laboratory information.
The primary bene® ts of this complementary solution approach hold the following possibilities.
. Increased capacity by liberating valuable resources from rework loops and the manual data review process. . Leveraging the approach to validation of the overall solution saving time associated with multiple computer validation plans and execution. . Promise of reduced time to market for new drugs by using a paperless environment where possible.
This presentation will describe the challenges, successes and experiences surrounding the task of integrating complementary information systems in a pharmaceutical development environment.

Using laboratory automation to prepare compounds precisely for HTS
Haissam Abdelhamid, Purdue Pharma LP, Cranbury, NJ, USA Pharmaceutical companies are constantly expanding their compound inventory and reorganizing their screening plates in an eå ort to ensure new chemical entities can be identi® ed as rapidly as possible. Whether these new compounds come from libraries (focussed or diverse) or medicinal chemistry eå orts, robotics has become an essential tool when preparing compounds for highthroughput screening (HTS). Using robotics to automate routine laboratory tasks not only increases the rate at which these activities can be completed, but also reduces incidents of human error. In our laboratories, we have integrated several robotic systems to aid in compound dissolution and HTS plate preparation. These enhancements have saved time and increased the reliability of our screening inventory, particularly when coupled with our in-house data-managemen t and inventory tracking software. Examples of laboratory automation used at Purdue over that past year will be presented. The con¯uence of robotics, biotechnology, computer sciences and the completion of genome sequencing eå orts for several organisms have resulted in revolutionary changes in how biomedical research is carried out. It is now possible to fabricate high-density arrays of speci® ed DNA sequences that include every known gene of an organism on a single glass slide. Labelled RNA or DNA targets (such as mRNAs obtained from cells, tissues or organisms under diå erent conditions) can be analysed by hybridization on the array. Diå erences in the levels of expression for thousands of genes can thus be assessed all at the same time in a single, simple experiment. Genomics, informatics and automation are playing increasingly important roles as discovery tools in the basic biological sciences, and as diagnostic and rational therapeutic aids in the clinical arena. We discuss the use of automation to increase productivity in micro-array fabrication and describe how automated procedures increase the quality of results in micro-array experimentation.

Use of the Bio-Tek Precision 2000 TM automated pipetting system for micro-arra y sample preparation
Paul Held, Bio-Tek Instruments, Winooski, VT, USA Co-author: Gavin Picket The production of micro-arrays requires the spotting of large numbers of unique DNA fragments onto several diå erent substrates. While several commercially available instruments have automated this spotting task, the sample preparation, culture propagation and maintenance of the DNA library are often performed manually with multichannel pipettes in 8 £ 12, 96-well, or 16 £ 24, 384-well formats. Manual multichannel pipettes, while more eae cient than single-channel pipettes, still represent a large amount of pipetting with many opportunities for pipetting errors. Here we describe the use of the Precision 2000 TM automated pipetting system to carry out many of the necessary pipetting tasks required for the preparation of samples for micro-array spotting. These steps include: the propagation of plasmid libraries, PCR reaction preparation, treatment of PCR products for agarose gel electrophoresis and the reconstitution of lyophilized samples before micro-array spotting. The Precision 2000 has a completely con® gurable six-station platform to hold the required pipette tips, reagent troughs and microplates (96 and 384 well) for¯uid transfer. The platform is removable, allowing for multi-user friendliness, easy cleaning and set-up of the instrument. The eight-channel pipette arm moves up and down as well as side to side, while the platform moves front to back to provide complete access to all locations on the work platform and complete con® gurability. The pipette arm uses a proprietary technology to pick up and seal reliably any standard tip with individual free-¯oating barrels that compensate for tips out of position. An optional rapid dispense eight-channel manifold, which uses a precise bidirectional syringe pump to dispense accurately and rapidly¯uids from a large unpressurized reservoir, is also available. Within this laboratory, analytical methods are developed to perform manual analysis. Consequently, we developed a procedure to transfer successfully a manual analytical method to this robotic system. First, the whole parameters of the TPWII that have an in¯uence on sample preparation have been determined and, second, the most important ones have been pointed out in order that they be validated. This has led to the establishment of a transfer protocol, which is in two parts: . development of the automatic method (study and optimization of all parameters); . validation of the automatic method.

Accelerating knowledge transfer for improved lead-candidat e selection
John P. Helfrich, NuGenesis Technologies Corporation, Westborough, MA, USA Today' s modern drug-discovery and development research groups are expected to improve the eae ciency of delivering NCEs to the clinic. The new high-throughpu t processes are dramatically increasing the raw number of data sets that must be interpreted for decision support across a global research team eå ort. The NuGenesis SDMS database serves as the centralized repository for analytical reports, compound presentation or summary documents, project reports, and instrumental raw data. This database can then be extracted to work in conjunction with your LIMS, Enterprise Data Management System and/or local specialized visualization and statistical data software products. If your data sources save or print, the NuGenesis SDMS platform can get it, automatically save it and allow fast eae cient utilization across the entire enterprise. After all, the data become information that when eå ectively communicated turns into critical-path knowledge for decision support in drug discovery.

Maximum return (strategic investments for highthroughput screening)
James LaRocque, Wyeth-Ayerst Research, Pearl River, NY, USA Advanced technologies theoretically enable modern HTS laboratories to screen expanding libraries quickly against a drastically longer list of targets while reducing reagent costs and maintaining more rigorous quality of results. Even in large pharmaceutical corporations, however, the collision of ® scal reality and the cost of state-of-the-ar t HTS technologies make getting the most out of capital investments a critical skill for success. Wyeth-Ayerst Research has made cost-eå ective investments in Packard CCS PlateTraks and Wallac CCD imagers, while also continually upgrading our Zymark-integrate d systems with new components and detectors.
This combination of equipment provides maximum throughput for 384-well formats and supports the transition to 1536-well formats with relatively modular units that can be programmed for multiple tasks. While the management of individual screening projects is delegated to individual scientists, the complexity of the automation infrastructure oå ers ample opportunity for the development of specialty skills, creating a well-rounded HTS staå . By practising both individual initiative and cooperativity, a relatively small group is empowered to Abstracts of papers presented at the 2001 ISLAR achieve eå ective sample plate replication, 384-and 1536well HTS, and rate-based hit characterization.

Automated LC/MS analysis of biomolecule s using ProMass
Je¡ Whitney, Novatia, Princeton, NJ, USA Co-authors: Mark E. Hail and David J. Detlefsen The recent increase in genomic and proteomic discovery has increased the need for high-throughpu t automated tools for biomolecule characterization. The¯ood of new protein targets will ultimately demand more eae cient tools for the evaluation of expressed proteins for drug discovery and structural biology studies. ProMass is an automated biomolecule deconvolution and reporting program used to process LC/ESI/MS data or single ESI mass spectra to derive molecular mass information. We recently integrated ProMass with the ThermoFinnigan Xcalibur data system to create a version of ProMass known as ProMassXcal i TM . ProMassXcal i processes entire Xcalibur sample sequences, deconvolutes mass spectra from LC/MS data and produces web-based reports.
ProMassXcali uses a novel cross-platform deconvolution algorithm known as ZNova TM . ZNova incorporates a unique charge-state scoring method that assigns the charge states of all signals in the ESI mass spectra and transforms the input ESI mass spectra to produce zerocharge mass spectra (i.e. molecular mass information). ZNova incorporates signal-processing techniques and unique logic that allow application to low charge state spectra and data of low signal-to-nois e ratio. As a result, ZNova can be used to process data from a wide variety of biomolecules including large proteins, oligonucleotides, peptides, etc. In this poster, an overview of the Pro-MassXcali/ZNova system will be presented along with selected applications which highlight its utility in a highthroughput environment.

Je¡ Whitney, Novatia, Princeton, NJ, USA
In recent years, drug-discovery researchers have placed greater emphasis on obtaining qualitative measures of drug candidate`quality' by measuring ADME/Tox tendencies earlier in the discovery process. This has led to the development of in-vitro assay methodologies for measuring metabolic stability, cell permeability, solubility, toxicity, etc. The challenge today is to automate fully all aspects of these assays by integrating intelligent data analysis and interpretation tools into one simple-touse solution.
This talk will focus on the analytical methods and intelligent data analysis approaches we have employed to assess rapidly metabolic stability and metabolite identi® cation. With the use of SmartLCMS TM technology, we will demonstrate the rapid assessment of metabolic stability of a compound followed by automatic detailed re-analysis using on-the-¯y intelligent dataanalysis techniques. In addition to up-front intelligent automation, we will illustrate the preliminary results of our MetLab TM software suite for back-end chemometricbased data processing techniques to con® rm automatically expected and unknown metabolites.

Je¡rey Murray, Paradigm Genetics, Inc., Research Triangle Park, NC, USA Co-authors: Ioana Popa-Burke and Chris Beecher
Paradigm Genetics, Inc., has industrialized the process of gene-function discovery for human health, nutrition, crop production and industrial products. The company has designed the GeneFunction Factory TM , an industrial-scale laboratory that explores the function of genes in organisms by integrating state-of-the-ar t sequencing technology with phenotype, metabolite and gene-expression pro® ling to collect hundreds of data points through a single technology platform.
The Metabolic Pro® ling group is responsible for monitoring changes in the biochemical pro® les of organisms that occur over the course of development in response to stress or induced genetic modi® cations. This is accomplished by using LC-TOF-MS, GC-TOF-MS and ICP-MS instruments. We describe here the innovative high-throughpu t processes developed for the cataloguing, storing in a dry environment, grinding, dispensing and extraction/derivatization of samples using custom robots, as well as proprietary instruments such as the`Mash-A-Matic and Buster' .

Kelly M. Jenkins, Bristol-Myers Squibb Co., San Diego, CA, USA Co-authors: Robyn A. Rourick, Reginald Angeles, Marianne Teopaco-Quintos and Daniel B. Kassel
Early determinations of pharmaceutical properties can serve as predictors of a compound' s likely developmental success. Our laboratory has implemented high-throughput ADME assays that address absorption, metabolism and physicochemical properties in an eå ort to minimize discovery to market attrition. While trying to meet the throughput demands of parallel synthesis, we established an integrated solution for ADME assays which incorporates a SAGIAN TM core system for the determination of both metabolic stability in human liver microsomes (HLM) and cytochrome P450 (CYP) inhibition. This automated solution has allowed an increase in capacity, throughput and reliability for both ADME assays.
The HLM assay uses a Multimek TM 96-channel pipettor for liquid handling. The analysis plates are transferred o¥ine for ® nal analysis using high-throughput parallel LC/MS. The CYP inhibition method uses a combination of liquid handlers and a¯uorescence plate reader to perform a single concentration pro® le assay for 88 compounds. CYP inhibition is measured for both CYP3A4 and 2D6 isozymes.
This system represents a fully integrated approach in support of high-throughpu t ADME evaluation in a drug-Abstracts of papers presented at the 2001 ISLAR discovery environment. The core system concept creates a plug-and-play approach that combines a series of modular stations to build a robotic system, which is exible, upgradable and easily recon® gurable when assays change or are newly developed. The application of these strategies as a means of assessing metabolic stability and CYP inhibition of our combinatorial libraries is discussed.

Joe Bradley, P¢zer Global R&D, Sandwich, UK Co-authors: Chris Chambers, Helen Boyd, Emma Faure, Joe Bradley, Simon Dales, Neil Benson, Wilma Keighley and Andreas Sewing
The growth in compound numbers and subsequent need for increases in throughput and reduction in cost are key drivers of developments in the ® eld of high-throughpu t screening. Increasingly, there is a move towards¯uorescence-based assay technologies, which are ideally suited for screening because they are versatile, homogeneous and amenable to both automation and miniaturization.
We developed a novel FRET-based assay to identify inhibitors of a metallo-endopepidase . This assay uses a novel substrate with a low K m , thus making it costeå ective for high-throughpu t screening. We have successfully transferred and validated this assay onto our Robolab linear track-screening platform. A full HTS campaign was subsequently conducted in 384-well format, achieving a maximum throughput of 75 000 data points within 24 h. This poster will illustrate the HTS process from assay development to high-throughpu t screening using state-ofthe-art technology.

John Helfrich, NuGenesis Technologies Corporation, Westborough, MA, USA
Managing and tracking data through the discovery process requires a compilation of many diå erent types of analytical, biological and image output. This includes the collection, storage and management of relevant scienti® c information about lead candidates, as well as immediate access to this information for complete compound document creation. Data collaboration across the discovery arena, even across the entire enterprise, is critical to making crucial go/no go decisions about subsequent lead candidate development.
The NuGenesis 1 Scienti® c Data Management System (SDMS) allows discovery scientists electronically to view, share, reuse and access data within the laboratory, as well as throughout the enterprise. You can easily capture laboratory and report data produced in any Windowsbased software application. These data can be catalogued automatically and accurately for easy retrieval. Immediate access to this electronic information is possible from anywhere around the world using Web technology. This presentation will focus on the principles of good data management that allow for pharmaceutical and biotech discovery facilities to leverage data as an asset, protect valuable intellectual property and ensure the accurate and easy creation of complete compound documents throughout the drug-discovery and development process. To increase sample throughput, many recent LC/MS methods have described techniques to shorten and/or simplify the chromatograph y step(s) before MS analysis. We describe a discrete solid-phase extraction step using Empore SPE membrane to capture the analyte, followed by an elution step that features the rapid quantitative transfer of the elutri zone (pure zone) into the MS. This application was designed primarily for bioanalytical support of high-throughpu t ADME screening.

Elutri-Zone
An SPE Card having the outer dimensions of a microtitre plate was developed by molding a plastic frame around a sandwich of 3M Empore sorbent (0.5 mm thickness, 8 mm particles) and micro® bre support material. Ninety-six discrete elution zones (7 mm diameter, 9 mm centres) were welded into the sheet. The SPE Card was designed to ® t in a modi® ed cell harvester (Tomtec) that facilitates SPE processing. The harvester performs sorbent activation, load and wash steps. Once the analyte has been retained, an interference wash is completed and the plate transferred to an elution device (Elutrix, Tomtec). Each well was eluted directly into the MS using¯ow rates ranging from 1.0 to 3.0 ml min ¡1 . A single HPLC pump delivered eluent composed of 5 mm Ammonium Acetate combined with 50± 90% methanol or acetonitrile.
The SPE Card and Harvester protocol allow rapid (2± 5 min) o¥ine sample clean-up followed by direct elution into the MS detector. This protocol results in the concentration of the sample and reduces liquid-handling steps as dry plates are transferred directly to the Elutrix for LC/MS processing. Owing to the small particle size of the Empore sorbent, the SPE Card can mimic the performance of discrete HPLC columns, similar to those used in high-throughpu t sample analysis or in online clean-up protocols. Relevant parameters and preliminary results will be described here.

John R. Alianti, GlaxoSmithKline, Research Triangle Park, NC, USA Co-author: Glenn Smith
Good laboratory practices (GLPs) require that appropriate calibration and performance testing be con-Abstracts of papers presented at the 2001 ISLAR ducted on all pipetting devices, including automated liquid-handling systems. Routine evaluation of the equipment must be performed to assess both the accuracy and reproducibility (precision) of liquid transfers. Additionally, the testing methods and acceptance criteria must adhere to established company-wide policies (SOPs). These methods must be sensitive as well as being robust and practical. Likewise, the data generated must be easily analysed, interpreted and documented.
Pipetting accuracy is determined by weighing 96-well plates (in replicates of three) before and after dispensing a speci® ed volume of liquid. System accuracy at various volumes is determined and the results used to calibrate the workstation as appropriate (e.g. Zymark RapidPlate Syringe Calibration Factors, Tecan GENESIS Liquid Classes, Packard MultiPROBE II Performance Files). The between-tip precision, individual-tip precision and individual-tip accuracies are all ascertained by subsequent colorimetric testing using a photometric microplate reader. A preformatted Microsoft Excel spreadsheet containing embedded formulas automatically analyses the data and generates a report.
The procedure developed combines the simplicity of gravimetric measurement with the speed of colorimetric microplate reading. The process is generic and can be used to assess or compare the accuracy and precision of any workstation. Automated data analysis and standardized report generation facilitate the process and ensure regulatory compliance.

John R. Kagel, Charles River Discovery and Development Services, Worcester, MA, USA Co-authors: Larry E. Elvebak, Brian E. Lilley, Jakal M. Amin and James A. Jersey
Client services for analysis and bioanalysis must: provide reliable results, meet increasingly aggressive timelines, meet increasingly lower unit costs and be performed under appropriate regulatory conditions. Automation was used successfully to satisfy these criteria regarding sample preparation for high-throughpu t LC/MS/MS GLP and non-GLP bioanalysis in a CRO. The ® rst phase in automating sample preparation involved processing samples in a 96-well format using 96-tip parallel pipetting workstations (e.g. Tomtec) in an open-access environment. The impact of this implementation was to increase productivity by at least twofold. The second phase in automating sample preparation used a liquid handler (e.g. Tecan Genesis) for the preparation of calibration standards and automated reformatting and dilution of samples from tubes or vials into 96-well plates. In addition, calibration and compliance issues related to automated sample preparation will be discussed. Surface plasmon resonance (SPR) technology is a powerful, label-free method for the analysis of molecular interactions. Prolinx 1 , Inc., has developed a new automated instrument, the Octave Molecular Interaction Analysis System, that allows this powerful analytical platform to become ubiquitous in life-science research and drugdiscovery laboratories. This technological breakthrough is the result of combining the miniaturized Texas Instruments Spreeta TM 2000 SPR sensor with the Prolinx Versalinx TM Chemical Aae nity Tools. These technologies have made possible the development of an instrument of moderate cost that incorporates eight independent sensors operating in parallel. The sensor surfaces can be readily and eae ciently modi® ed with molecular targets, and exhibit low non-speci® c binding. Samples are introduced to the sensors from standard microwell plates using an integrated liquid-handling robot. This new instrument will signi® cantly increase the throughput of SPRbased molecular interaction analysis in basic biological science and drug-discovery applications.

Kathryn B. O'Mara, GlaxoSmithKline, Research Triangle Park, NC, USA Co-authors: Lisa St John-Williams and John A. Dunn
Recent advances in automation have signi® cantly reduced the amount of time it takes to prepare clinical samples for LC/MS/MS analysis. As a result, the eaeciency bottleneck has shifted from sample preparation time to analytical run time on the mass spectrometer. To reduce the mass spectrometry analysis time, the use of dual-column technology was evaluated for three singlecolumn validated LC/MS/MS methods.
For each of the three methods, the HPLC system was enhanced with switching valves and the appropriate programming to allow the unit to function in the dualcolumn mode. A semi-automated liquid handler was used to perform either protein precipitation or solid-phase extraction using 96-well technology. A triple quadrupole mass spectrometers was used to acquire data using multiple reaction monitoring.
The three quantitative LC/MS/MS methods using dualcolumn technology are currently being used in a highthroughput laboratory. Each method has been validated to demonstrate acceptable accuracy and precision. Robustness has also been demonstrated by the analysis of thousands of samples in a GLP environment. For each method, the analytical run time has been reduced by 50% and instrument eae ciency and sample throughput have been increased twofold.

Kelly Johnson, Waters Corporation, Milford, MA, USA Co-authors: John Erve, Andre Dandeneau and Beverly Kenney
Modern drug discovery has been transformed by the automation of research. The resulting explosion of data in the discovery pipeline, combined with the pressure to reduce costs and speed up drug-discovery cycles, provides a strong demand for fast and selective analytical methods to produce quality data.
In vitro metabolic stability assays provide a rapid estimation of new chemical entities in the discovery phase of pharmaceutical development. These multistep assays routinely incorporate robotic liquid-handling systems to automate the incubation step of test compounds with human liver microsomes in 96-well plates. Metabolic stability is subsequently measured by LC/MS as the amount of substrate metabolized (expressed as the percentage remaining of the initial substrate). The resulting information is crucial in the selection process in determining a compound' s potential`drugability' .
Given the large number of samples generated by the liquid-handling systems in metabolic stability assays, employing automation and higher throughput in every step of the assay has become a necessity. This presentation will demonstrate the use of high-throughpu t LC/MS methods to analyse samples produced in a metabolic stability assay. We will develop LC/MS methods that signi® cantly increase overall throughpu t without sacri®cing data quality. By incorporating such factors as parallel sample processing, high sample capacity, alternating column regeneration and smaller diameter columns to reduce cycle times, we can analyse the large number of samples produced and ultimately help expedite the drug-discovery process.

Konstantin Tsinman, pION, Inc., Woburn, MA, USA Co-authors: A. Avdeef, D. Voloboy, M. Stra¡ord and B. Kenney
The assessment of passive transport properties of over 20 drug and natural product molecules was made using the in vitro absorption model based on ® lter-immobilized arti® cial membranes (® lter-IAM), assembled from phosphatidylcholin e in dodecane, in buå er solutions at pH 7.4.
Several of the compounds were lactones extracted from the roots of the kava-kav a plant. Experiments were designed to test the eå ects of stirring during assays and the eå ects of varying the assay times.
Stirring during assay signi® cantly increased the observed permeabilities for highly mobile molecules.
In addition to permeability measurements, membrane retention of compounds was determined. Yangonin, desmethoxyyangonin , ketoconazole and phenazopyridine were > 60% retained by the arti® cial membranes containing phospholipids.
The in¯uence of hydrogen bonding was explored by determining permeabilities using ® lters coated with dodecane free of phospholipids. The membrane transport of phenazopyridine (strong hydrogen-bon d donor) is about twice as fast and retention is about twice reduced in the inert lipid membranes compared with phospholipidbased membranes.
In the ® lter-IAM method, concentrations were determined by microtitre plate UV (190± 500 nm) spectrophotometry and by LC/MS. Higher-throughpu t was achieved with direct UV by the use of 96-well microtitre plate formats and with LC/MS by the use of cassette dosing (5-in-1).

Kurt Roinestad, Purdue Pharma LP, Ardsley, NY, USA
This presentation will describe Purdue Pharma' s approach to an integrated solution between Laboratory Information Management Systems (LIMS), Chromatography Data Systems (CDS), Process Management and Compliance Systems (PMC), and Scienti® c Data Management Systems (SDMS). It will describe the vendorselection process and how Purdue Pharma created an environment to produce a best-of-breed information solution to creating the paperless laboratory. Working towards the creation of a paperless laboratory has the potential to result in signi® cant bene® ts including resource liberation and an electronic compliance platform for process management and data review. This paper will describe the process of working with four vendors to create a complementary data management solution to Purdue Pharma' s needs. The process of logging in samples, creating work-lists, executing analytical methods, capturing data electronically, storing electronic records and submitting reports will be reviewed. After data creation, electronic data review will be discussed enabling electronic`instant replay' of laboratory information.
The primary bene® ts of this complementary solution approach hold the possibilities to do the following.
. Increased capacity by liberating valuable resources from rework loops and the manual data review process. . Leveraging the approach to validation of the overall solution saving time associated with multiple computer validation plans and execution.
. Promise of reduced time to market for new drugs by using a paperless environment where possible.
This presentation will describe the challenges, successes and experiences surrounding the task of integrating complementary information systems in a pharmaceutical development environment.

High-throughpu t robotic workstation for performing gene-expressio n assays
Michael A. Kuziora, Gene Logic, Inc., Gaithersburg, MD, USA Metabolism and growth are dependent on a highly orchestrated interplay of a variety of proteins within cells. These proteins function as enzymes in metabolic pathways, signalling molecules for communication between and within cells, and as various cellular structural components. In a disease state, cells often modulate the amounts of speci® c proteins and may produce new proteins not normally found in a particular type of cell.
The alterations in the proteome observed in the diseased state most often result from changes in gene expression levels within a cell. Gene Logic believes that pharmaceutical companies can reduce the time, risk and cost associated with drug discovery if they know the expression levels of genes that play roles in the disease-associate d pathways. Such knowledge may help them discover drug targets, screen drug leads, and predict toxic and pharmacological responses to drug leads.
Using the Aå ymetrix, Inc. GeneChip 1 micro-array platform, Gene Logic has measured the expression levels of thousands of genes from a diverse range of normal and diseased tissues to create a reference gene expression database called the Gene Express 1 Suite. The bioinformatic analysis tools incorporate d in the Gene Express Suite facilitate the identi® cation of distinct sets of genes whose expression is consistently altered in a particular disease state. The expression patterns of these gene sets become a molecular ® ngerprint of the disease and thus not only re¯ect the disease status of metabolic pathways, but also could also serve to indicate the eå ect of potential treatments when aå ected cells are exposed to a potential drug.
A primary goal of a pharmaceutical company is to discover drugs that restore the normal functioning of the disease-aå ected pathways. In one scenario, a pharmaceutical company could use the information in the Gene Express database to identify a small set of genes that indicate a disease state. The expression of this gene set is then monitored when an appropriate cultured cell line that represents a disease state of interest is exposed to chemicals from a compound library. A potential drug can be identi® ed for further characterization if, for example, it is found to restore expression levels of the gene set to normal levels.
Unfortunately, the use of large micro-arrays to measure gene expression of a few genes in a high-throughpu t screen is not economically feasible at this time. We therefore investigated the use of alternative methodologies to measure gene expression. Bayer Corporation' s Quantigene TM assay, which uses branched DNA (b-DNA) for signal ampli® cation, and provides a sensitive and reproducible method for measuring mRNA levels of a small number of genes. The assay' s 96-well plate format and simple handling procedures makes it highly amenable to high-throughpu t robotics. This talk will describe a Zymark Staccato TM Workstation designed to perform the hybridization, ampli® cation and signaldetection steps of the Quantigene assay.

Lars A. Svensson, AstraZeneca, Molnda, Sweden Co-authors: Rimstedt Eva and Svensk-Ankarberg Anna
Correlation between drug dissolution results obtained by both liquid chromatograph y (LC) and ultraviolet-visua l (UV-VIS) spectrophotometry has been performed by multivariate calibration.
The dissolution of tablets or capsules, stored and exposed to harsh conditions (high humidity/elevated temperatures) at diå erent degrees, was carried out by an automated system based on the Zymark XP robot. The system had the capability of withdrawing samples to be subsequently prepared into LC vials at the same instance as an UV-VIS spectrum was recorded via a ® breoptic probe. The same probe was used in all vessels and could be attached directly adjacent to the ® lter tip for the sample withdrawal on one of the robotic hands.
The subsequent multivariate calibration was performed with Simca-P 8.0 (Umetrics AB, Sweden). Most of the calibration is explained in two components, with the ® rst mainly corresponding to the UV spectrum and the second taking into consideration the baseline oå set.

High-throughpu t chemistry: an integration of chemistry, automation and informatics
Li Chen, Ho¡mann-La Roche, Inc., Nutley, NJ, USA One of many impacts that combinatorial chemistry has been made in the last decade is to stimulate intellectual creativity to invent new and more eae cient ways of making new compounds. Although much eå ort has been investigated into developing synthetic methods, automated synthesizers and analytical tools, the high productivity of compound library synthesis cannot be achieved without the integration of chemistry, automation and informatics into an eae cient process. I will emphasize what we have learned in how to establish an infrastructure that integrates functional modules containing a diverse set of tools for high-throughpu t organic synthesis applications. This modular approach provides the maximum eae ciency of combinatorial tool functions to support multiple project teams for drug lead generation, exploration and optimization.
The agenda includes the following.
. Data management and chemo-informatics. The Living Chip TM is a nanotitre plate consisting of a uniform and addressable through-hole array. The though-hole s are nominally 300 mm square and 500 mm deep, giving each well a 50-nl volume. Proprietary coatings make the surfaces of the plates hydrophobic and the interior of the wells hydrophilic. This allows the samples to be held in the wells by surface forces and prevents sample contamination from adjacent wells. Ten thousand-well chips are currently in use and 100 000-well chips are in production.
Massively parallel mixing of assay components takes place when chips containing diå erent components are stacked such that the through-holes align. The chips are imaged using a CCD array for readouts such as absorbance,¯uorescence and luminescence. Detection is performed in a transmission geometry, taking advantage of the bottomless wells. These key features and the sophisticated robotics built by BioTrove allow for the rapid acquisition of biochemical information, on the order of 10 7 measurements per day. Combining nanolitre reaction volumes and a simple interface to microtitre plates, the Living Chip TM format conserves compound libraries, increases analytical capabilities and decreases costs. The platform has applications in all aspects of drug discovery from materials' handling and storage, target ID and validation, to lead discovery and optimization.
Cell culture in the Living Chip TM using yeast (S. cerevisiae) and bacteria (E. coli) shows similar growth characteristics to cells grown in bulk. Additionally, E. coli cells expressing GFP inoculated into isolated channels show no crosstalk.
Diverse biological libraries can be introduced to the chip using a simple dip loading procedure. This technique was used by BioTrove collaborator Genofocus to dip load cells expressing a lipase enzyme library into a Living Chip TM . The cells were grown up overnight and then stacked with a chip containing a tagged substrate so that wells containing clones with increased lipase activity yielded a greater¯uorescence signal. The`hits' were harvested by a puå of air from a microsolenoid valve positioned above the well, into a microtitre plate waiting 125-300um >500um hydrophilic hydrophobic Some of the challenges and goals in many of today' s screening laboratories are to investigate, evaluate and validate new techniques that will enable the reduction in cost, provide an increase in eae ciency and strengthen the quality of screening data. Towards the end of reducing cost, we have been a Beta test site for a novel 96-well submicrolitre assay system called Arteas. The architecture of Arteas oå ers a solution to the problem of evaporation in small-volume assays.
We have successfully performed a¯uorescent-base d enzyme assay in Arteas with a reaction volume of 400 nl. We successfully reproduced published control IC 50 s and routinely generate Z' -factors in the range 0.6± 0.7. We will show these data and variability around some tests performed with actual library compounds. Translation of this device to production screening will be dependent upon the successful integration of nanoliquid handling onto our robotic platforms. Initially, we will integrate this new liquid-handling option onto our Fast Track robotic platform. Progress on this translation will be addressed.

Malcolm Willson, Systems Research, GlaxoSmithKline, Stevenage, UK Co-authors: David Brown and David Hayes
Programme-targete d SAR screening within Systems Research at Stevenage calls for a process able to progress compound activity determination (80 to > 320 compounds/week) in real time, enabling ongoing chemistry to continue based on known biological activity.
Eae cient compound dilution and generation of assay plates is required to meet weekly turn around times for IC 50 and cross-screening data generation.
Comparison of manual across-plate dilution, vertical dilution down a series of plates and Z-dilutions within a series of 384-well plates will be discussed.
Use of the twin head Biomek FX uses the¯exibility of both the 96-and 384-tip heads, allowing Z-dilutions, a very eae cient method for automated generation of IC 50 s. By having space on the stock plates for several extra compounds, a selected range of`standards' covering several assays can be included to enable assay performance and automation QC to be monitored for each set of compounds, along with suae cient control/blanks for statistical analysis of assay performance.
Cloning of assay templates reduces IT resource requirements for data analysis as templates can be cloned rather than written from scratch.
If compound numbers increase, the system can be converted to run vertical dilutions down a series of plates (diå erent dilution per plate). However, this only works if the assays are stable enough for the increased length of assay due to the higher plate numbers.
Conversion to 1536 plates doing Z-dilutions from 384well stock plates is also an option that would be possible when liquid-handling devices can cope with smaller volumes for assays.

Charles J. Manly, Discovery Technologies, Neurogen Corporation, Branford, CT, USA
Drug discovery today includes considerable focus of laboratory automation and other resources on highthroughput technologies, but lead generation and optimization to clinical candidates continues to be a lengthy and costly process. The real bene® t of today' s technologies is beyond the exploitation of each individually. Only recently have signi® cant eå orts focused on eå ectively integrating these complex discovery disciplines to realize their larger potential. Informatics, computationa l chemistry, virtual screening and data mining play a large role in this integration and in increasing the eae ciency of the drug-discover y process. Our experience in automating a genomics' laboratory for single nucleotide polymorphism (SNP) detection has shown us that automation projects are not often completed in the planned or expected manner. With automation playing an ever-increasing role in the research and development laboratory, we will discuss why some projects succeed while others fail. For instance, what are the real bottlenecks to address? Does experience make a diå erence? Is management expectation realistic? What are the advantages and disadvantage s of custom or commercial automation? What are the roles of the scientists, engineers, and laboratory managers in such projects?
The recent completion of our high-throughput screening project has enabled us to look back objectively and answer these questions for our genomics' application.
The high-throughpu t screening project at Aå ymetrix began in 1999. The initial goals of the project were to screen for SNPs across 40 unrelated individuals in an automated fashion. In the initial stages of this project, the process), which encompassed sample preparation to scanning, was done manually. This provided an output of about 2.3 Mb of sequence a week for every 4.5 people. With the collaborative eå ort of the executive management group, the high-throughpu t screening laboratory, the engineering group and the bioinformatics group, we set out to accomplish this task.
Unfortunately, as the project continued, con¯icts and diae culties arose. Each group had its own expectations for the project that were not necessarily communicated with the others eå ectively. This led to many problems including delays, miscommunications, unrealistic expectations, frustration and blame being placed on one another. In addition, another facet compounded the already diae cult process: reality. We found several instances where reality interfered in the development of the project including budget, time, resources and process bottlenecks. We also saw evidence of con¯icts such as when we needed to decide between 96-or 384-well plates, tubes or plates, custom or commercial or semi-automation, or even con¯icts between the scientists and the engineers.
Fortunately, we realized many of these issues and managed them eå ectively. We also proceeded to perfect the process in other ways using protocol enhancements and microscaling improvements. This increased our throughput to 1.4 Mb per day for every two people while reducing reagent costs.
At the conclusion of this project, approximately 25 000 genes were screened covering 8.2 Mb of the genome (including genes, regulatory regions and STSs). Since most of the projects were tested against 40 individuals, we screened 24 million bases of dsDNA and identi® ed 15 682 SNPs that have been deposited to the SNP public database. In addition, much of the software, automation and protocols developed during this project have been incorporated into other internal research laboratories.

Maria Styslo-Zalasik, R. W. Johnson Pharmaceutical Research Institute, Raritan, NJ, USA Co-author: Kathleen Cirillo-Penn
In the past 10 years, laboratory automation has become increasingly important for routine chemical analysis testing in the pharmaceutical industry. The use of automatization or semi-automated laboratory equipment can reduce the laboratory¯ow-through time of the samples in the QC area. Likewise, automated and semiautomated equipment can be used as a tool for research during method development and validation. This presentation will illustrate how semi-automate d sample preparation can aå ord rapid optimization of method parameters during early-stage method development and validation. The potential dissolution rate and solubility issues with high sample concentrations can be avoided through the evaluation of data from parameter optimization of the sample preparation using semi-automated instrumentation . Examples for the use of a stand alone extractor in an assay/purity method validation will be presented.

New automated instrument for the characterization of biomolecula r interaction s
Mark L. Stolowitz, Prolinx, Inc., Bothell, WA, USA Surface plasmon-resonanc e technology is a powerful, label-free method for the analysis of biomolecular interactions. However, the use of this technology is limited by the cost, throughput and complexity of existing instrumentation and chemistries. Prolinx has developed a new instrument that will address these limitations and allow this powerful analytical platform to become ubiquitous in life science research and drug-discovery laboratories.
This technological breakthroug h is the result of combining the Texas Instruments Spreeta 1 2000 chip with Prolinx' s Versalinx TM Chemical Aae nity Tools. These technologies enabled the development of an instrument of moderate cost that incorporates eight parallel sensor surfaces that can be eae ciently modi® ed with molecular targets and exhibit low non-speci® c binding. The presentation will encompass technology, design and applications of this new instrument.

Mervyn Cadette, GlaxoSmithKline, Beckenham, UK
Relative to chemical assays in a pharmaceutical development environment, biological assays are relatively`low throughput' in comparison. However, the complexity of assays associated with biological products is often of a diå erent magnitude. Biological assays become attractive automation candidates based upon their complexity coupled with high demand for their use in biopharmaceutical analysis.
The Tissue Culture Infectious Dose 50% (TCID50) assay was used for the quantitation of viral titres used in vaccine products.
These assays required sterility, therefore requiring all equipment to be housed in Class II microbiological safety cabinets. The TCID50 assay was programmed on a Tecan RSP 200 using eight disposable-tip liquid-handling probes in conjunction with a robotic manipulator arm. Culturing of plates containing tissue culture, to support the TCID50 assay, was automated using a Zymark Twister coupled to a Multidrop dispenser.
Scheduling and liquid-handling software were combined to execute this automated assay. The use of the`joblist' function allowed the operator to select from a databas e of prede® ned dilutions within the Logic pipetting software.
A formal comparison of the automated TCID50 assay against the manual TCID50 was performed and proved comparability.

Integrated approach to high-throughpu t sample processing , characterizatio n and puri cation
Michael L. Moore, GlaxoSmithKline, King of Prussia, PA, USA The extension of high-throughpu t synthesis in support of lead optimization has imposed increasingly stringent requirements on compound characterization, purity and accurate concentration determination. We developed an integrated and highly eae cient process for compound analysis, puri® cation and sample processing with a capacity of 100 000 compounds/year at a 20-mg scale. Puri® cation is driven by ultrahigh-throughpu t analytical LC/MS, which minimizes the number of fractions collected and analysed. Custom software with a browser-based front end is employed to track samples and provide the required data to robotic workstations.

Michael R. Kozlowski, Axiom Biosciences, San Diego, CA, USA
The completion of the cloning of the human genome has provided the drug-discovery community with a wealth of potential drug targets. At the same time, it has produced challenges to the way in which drug discovery is done. Formerly, the mandate of drug discovery was to screen a relatively small number of well-validated targets against an immense number of compounds. Now, drug-discovery scientists are faced with processing a very large number of minimally validated targets. Most targets were formerly single proteins, which are amenable to screening in highly reductionist systems. Now it is clear that most biological processes, including those contributing to pathological states, must be the result of complex interactions between proteins. A way must be found to address this new level of complexity. In addition to these challenges, the availability of the entire sequence of the human genome raises expectations for the rapid introduction of more, and better, drugs.
These challenges demand new ways of thinking about how we carry out drug discovery, and about drugdiscovery automation, from assay development to AD-MET pro® ling. This talk will discuss some of these new approaches.

Michael West, P¢zer Global R&D, Groton, CT, USA Co-authors: Larry Cohen, Al¢n Vaz and Shawn Harriman
A¯uorescence-base d drug interaction assay using recombinant CYPs and a cocktail of CYP-speci® c probes in human liver microsomes was assessed as higher throughput methods for evaluating the potential for inhibition of CYP1A2-, CYP2C9-, CYP2C19-, CYP2D6-and CYP3A4-mediated metabolism. Comparisons of IC 50 obtained with the¯uorogenic and conventional drug probes in recombinant CYPs were similar for CYP1A2, 2C9, 2C19 and 2D6, but not for CYP3A4.
Additionally, using a single-point estimated IC 50 approach, compounds that were shown to be inhibitors using conventional drug probes with human liver microsomes were also classi® ed as inhibitors in recombinant CYPs using the¯uorescent probes. For the cocktail approach, it was shown that the CYP-speci® c reactions were not altered in the presence of multiple probes as indicated by no distinguishable eå ect on K m or V max . As expected from this result, the IC 50 s generated in the cocktail incubations were in good agreement to those obtained from individual incubations.

GigaMatrix TM ultrahigh-throughpu t screening platform
Mike La¡erty, Diversa, San Diego, CA, USA The myriad of microbes inhabiting this planet represent a tremendous repository of biomolecules for pharmaceutical, agricultural, industrial and chemical applications. Diversa Corporation has the unique capability of accessing this microbial diversity by taking a cultureindependent, recombinant approach to the discovery of novel proteins and small molecules. Diversa' s discovery programme uses genes and gene pathways captured from nucleic acids extracted directly from the environment, which are then constructed into complex, 109-member environmental libraries. These libraries often contain up to 5000 diå erent microbial genomes and, thus, require high-throughpu t screening methods to cover their diversity eå ectively.
Diversa has developed GigaMatrix TM , a new ultrahighthroughput screening platform. GigaMatrix plates have > 100 000 bottomless wells in the same footprint as a microtitre plate. The platform is automated and capable of screening 1 billion clones per day. Less equipment time and manpower are required and assay costs are dramatically reduced as compared with traditional microtitre plate-based screening. The power of the GigaMatrix platform to discover novel enzymes, small molecules, protein therapeutics and other bioactive molecules will be presented.

Building an electronic data-handling, compliant, foundation for the QA analytical laboratorie s at Westborough
Mike Stroz, AstraZeneca, Westboro, MA, USA A signi® cant portion of QA laboratory activities are dedicated to compliance and data handling in a paperbased system. AZ Westborough is implementing an electronic environment in the laboratories to ensure 21 CFR Part 11 compliance, improve eae ciency and reduce costs. The presentation will discuss the applications selected, and the validation and system architecture being installed to achieve these goals.

Min Chang, Abbott Labs, Abbott Park, IL, USA Co-authors: Huong Mai, Anita Shen, Brendan Swaine, Qin Ji and Tawakol El-Shourbagy
Time event-controlled column switching valves have been available to analytical scientists since the mid-1970s. Earlier uses of the column-switching valve included removing late elution peaks from the HPLC run, autosampler sharing, two-dimensional HPLC separation (heart cut), inline ® lter/guard column regeneration, online concentration, unattended column selection and fractions' collection. Although there is at least one column-switching valve in an HPLC system (the injector valve), the idea of the addition of another valve has not been widely accepted. Analytical scientists have turned away from the technology possibly due to the lack of ruggedness of the ® rst-generation air-actuated valve and other additional pieces of HPLC equipment including columns.
Recently, with the availability of commercial online solid-phase extraction system(s), the column-switching technique has become an acceptable option for bioanalytical sample preparation. At Abbott Laboratory, we have developed several online solid-phase extraction HPLC methods using automated valves, an internal reversed-phase guard cartridge and HPLC equipment by Shimadzu and Agilent. Internal standard forti® ed plasma were clari® ed by either centrifugation or ® ltered before the HPLC injection. These methods have been used successfully to analyse two Abbott compounds and their metabolites. This application of automated columnswitching valves has provided an alternative to o¥ine solid-phase extraction and liquid± liquid extraction and made it possible to select the best technology to increase assay throughput and sensitivity.

Muhammad Albarakeh, Barr Laboratories, Inc., Pomona, NY, USA Co-authors: Richard Ashley and Timothy Breuninger
Current laboratory demands related to the high throughput of samples, the need to keep costs and expenses minimal, yet keep the release function¯owing to prevent product backorder, have necessitated the use of laboratory automation and robotics systems. As a result of this automation, huge amounts of samples and related data are processed and generated.
Without a properly controlled and validated system, the very tools implemented to process high amounts of samples can bring your release function to a halt. Overviews on how to prevent cGMP logs jam and reach à zero backorder' release of product will be presented.

Methods' development for monoclonal antibody screening: how to think like a robot
Nanci E. Donacki, MedImmune, Inc., Gaithersburg, MD, USA A single fusion for the development of monoclonal antibodies will often generated 20 or more microplates that need to be screened. ELISA (enzyme-linked immunosorbent assay) is the most common method for screening newly developed monoclonal antibodies for antibody production and speci® city. The method, although speci® c for the antibodies, is highly repetitive for each step and can easily be automated. However, the way the assay is performed at the benchtop is not always the best way for an automated system to run the assay. Tips and techniques for transferring an assay from benchtop to automation will be presented.

Lean manufacturing and six sigma: an evaluation of the impact of these concepts on laboratory automation
Nigel North, Pharmaceutical Development, GlaxoSmithKline, Ware, UK The concepts of lean manufacturing and six sigma are now beginning to be applied to pharmaceutical manufacturing processes. Lean manufacturing has been successfully applied for many years in the automotive and aerospace industries with the key goal of reducing waste. Six sigma is a more recent concept involving reducing variation in manufacturing processes which has been implemented in the semiconductor industry resulting in signi® cant cost savings. The combination of lean manufacturing and six sigma provides a powerful combination of principles to deliver robust manufacturing processes with the elimination of non-value-adde d activities. The eå ect of these concepts on how we approach automation in the laboratory will be examined together with providing some perspectives on new technology that will be required to meet these challenges.

Applications of the Zymark Staccato for in vitro drug metabolism studies
Heather Sulkowski, Boehringer Ingelheim Pharmaceuticals, Inc.,

Ridge¢eld, CT, USA Co-authors: J. Richard Mount¢eld, Donald Tweedie and Drane è O'Brien
Advances in genomics, combinatorial chemistry and pharmacology screening have led to new challenges within drug metabolism due to large numbers of compounds requiring in vitro evaluation. To meet these demands within Boehringer Ingelheim Pharmaceuticals, Inc., the Zymark Staccato was evaluated as a robotic tool for conducting multiple in vitro assays. The criteria for an automated system included the following.
. A 96-well head allowing for maximum sample throughput. . Thermal block to maintain constant temperature. . Flexible deck layout. . Open-access capability As part of the evaluation process, assays were compared using both manual and automated methods. The results are described in this presentation Automation of several key in vitro metabolism assays has been achieved. The results indicate that the automated assays allow for a higher throughput while maintaining a high degree of precision. The work will be extended to additional in vitro assays, and the strategy of comparing manual versus automated processes will be continued as part of the validation procedures.

Patricia A. Fowler, Waters Corporation, Milford, MA, USA Co-author: Michael Swartz
Method validation is a tedious process performed to determine if an analytical method meets the requirements for its intended purpose. In the regulated laboratory, method validation may take many days to perform the necessary analytical tests. Data reduction and the statistical analysis of results performed can be a very time-consuming process. There is also a greater possibility of introducing error when calculations are performed manually. With the use of automated software to perform these calculations, method validation is much faster and easier, with less chance for error.
In this poster, we will show that an analytical method was validated using automated software. Chromatographic results were directly accessed from a relational database bypassing manual intervention. Statistical calculations were performed automaticall y and a report generated showing the results of the analyses from the Student, Cochran, Dixon and Fisher tests. Graphs were generated representing the results of the statistical analysis. In addition, we will show that the data reduction and statistical calculations necessary to validate the method, complete with the necessary documentation and report generation, are completed in signi® cantly less time. The fast pace of the pharmaceutical industry requires laboratories to reduce the analytical burden of their test procedures and increase productivity while still satisfying regulatory compliance. There are several waysto meet these challenges in the dissolution laboratory. By automating the dissolution process, pharmaceutical laboratories eliminate the slight variations that may occur in manual methods, insuring reproducible data, higher throughput and cost reduction. Validated single-source software control of the entire system, as well as dissolution data acquisition, calculations and reporting, can further streamline the work ow while maintaining FDA compliance with 21 CFR Part 11.
Automated online HPLC dissolution systems can pool dissolution samples signi® cantly to save time. Similarly, automated sampling at shorter intervals and analysis of a large number of samples by online HPLC may provide a more complete solution for the decision-making process in the early stages of drug development. In addition, these systems must be capable of handling increasingly complex formulations such as multiple actives and widely varying dosage levels, as well as diå erent media types such as buå ers and surfactants.
Our poster shows applications used to help reduce the analytical burden and increase the sample throughput and productivity in the laboratory.

Patrick B. Cahill, Genome Therapeutics Corporation, Waltham, MA, USA Co-authors: Dina Weymouth, Cori Gustafson, James Hurley, Michele Bakis, Veena Kamath, Doug Smith and Lynn Doucette-Stamm
Genome Therapeutics Corp has developed a high-throughpu t SNP assay based on the ability of some DNA polymerases to proofread the sequence as they extend. This Exo-Proofreadin g SNP assay can be accomplished, from sample addition to detection, in one tube. In addition, this SNP assay has the capability to be performed directly on genomic DNA. Results will be presented for this assay on multiple SNPs screened on large populations. The results will be compared with ASO and RFLP data with discrepancies resolved by sequencing. This assay provides a new, highly robust and fast method for large-scale SNP screening required for human disease and pharmacogenetic studies.

Development of a lead identi cation platform for kinase drug discovery
Paul Gallant, Millennium Pharmaceuticals, Inc., Cambridge, MA, USA The goal of any drug-discovery process is to eliminate quickly`poor' leads while rapidly advancing those with the highest potential. Millennium has used a combination of novel and mature technologies to establish a system to progress novel kinase targets rapidly through assay development, high-throughpu t screening, hit validation and hit characterization. By selecting focused technologies, automation, secondary screens and IT, an integrated process has been constructed which is capable of handling multiple targets quickly while identifying the compounds with the highest potential for lead optimization. This presentation will discuss the platforms chosen for assay development, high-throughpu t screening, hit con® rmation and potency/selectivity determinations. A process will be described to show how using a standard well-de® ned system can eå ectively move multiple kinase targets rapidly into the`Lead Optimization' stage.

Peter D. Bryan, Forest Laboratories, Abbott Park, IL, USA Co-authors: Min S. Chang and Anita Shen
Now that LC-MS has become the technique of choice in the GLP bioanalytical laboratory, the work¯ow bottleneck has shifted from the separation± detection to the sample-processing portion of the bioanalytical method. To address the sample-processing bottleneck, eå orts have been concentrated on the conversion and validation of existing liquid± liquid extraction (LLE) to solid-phase extraction (SPE) methods. Profound increases in sample throughput have been achieved using LC-MS with automated SPE in the 96-well format over manual LLE extraction HPLC methods. Validation of automated 96-well format SPE methods have shown their equivalence to manual LLE extraction. Automated 96well format SPE is also much less tedious and allows the analyst to concentrate more on data analysis and compliance issues. For routine analysis, transfer from clinical tubes to the 96-well format has been accomplished using either the Biomek 2000 or the Hamilton MicroLab AT. Automated SPE is then performed on the Biomek 2000.
Additionally, semi-automated method development for 96-well format SPE has been automated using the Biomek 2000.

How small is enough?
Peter Grandsard, Amgen, Inc., Thousand Oaks, CA, USA Co-authors: Jim Petersen, Brian Rasnow, Mike Johnson, Chuck Li, Les Walling and Doug Overland A signi® cant technical and organizational challenge facing the biopharmaceutica l industry is the incorporation of miniaturization technologies into their business. The successful application of these technologies requires answers to the following questions. What is this new miniaturization technology? Which technologies should be applied to which work¯ows? Why? When should they be implemented? Should a technology be implemented early, through some sort of technology access programme (TAP), or should one wait until it becomes commercially available? Our views on these matters will be presented.
In collaboration with other Amgen R&D groups, the Research & Automation Technologies department has been actively cultivating critical collaborators with knowledge of submicrolitre liquid-handling, micromachining and detection technologies. For example, our TAP with Caliper, Inc., has increased our institutional understanding of miniaturization applications in the drug-discovery process¯ow, most particularly in small molecule screening and DNA analysis.
Miniaturization at Amgen has invoked a cascade of new projects or at least studies. The goals of these projects are to ® nd solutions for interfacing any miniaturized platform to the`macro-world' , such that the advantages of miniaturization are not undone by integration or certain interfaces. Operations that need to undergo changes to bene® t fully from miniaturization include the storage and retrieval of small molecules, detection and compound handling during screening, and information management. We will discuss some of our ® ndings and solutions.

Phil Small, Tripos Receptor Research, Bude, UK
A process-integratin g design, synthesis and analysis of combinatorial libraries has been implemented at Tripos Receptor Research. Crucial to the success of this process has been the development of a proprietary informatics system. This has so far been developed to manage reagent inventory, track samples, and record data from synthesis and analysis to provide a valuable database of compound information.
The combination of Tripos' s proprietary design software with automated synthesis and informatics leads to the production of drug-like libraries with well-de® ned purity and full synthetic history. This presentation will cover what were considered the important aspects involved in setting up and managing an automated chemistry facility.

Phil Waters, GlaxoSmithKline, Research Triangle Park, NC, USA Co-authors: James Ormand, David Igo and Pingyun Chen
Evaluation of the equilibrium solubility and chemical stability of compounds as a function of pH aids our understanding of the bioavailability of a drug candidate. Additionally, these data lead to rapid development of formulations for use in preclinical and clinical studies. Conducting these measurements often requires a signi®cant amount of human labour and time. Reducing the labour and time burden is expected to increase the frequency with which these measurements can be ap-plied, thus increasing sample throughput and improving the quality of decisions impacting compound selection and evaluation.
This presentation will describe sample-handling methodologies and automated instrumentation developed to conduct solubility measurements as a function of pH. The use of these methodologies and instrumentation has improved both throughput and eae ciency while maintaining the accuracy that can be achieved with manual methods. The accuracy and precision of the system will be illustrated using data on model compounds as well as drug candidates. Finally, we will describe the strategy and process used to prepare samples for stability determinations at user-de® ned pHs, and how this process is done in concert with the solubility determinations.

Quiyang Zhang, Cytoplex Biosciences, Plano, TX, USA Co-author: Alex Freeman
Cell-based assays are increasingly used for high-throughput screening in drug-discovery programmes because of their high information content. For the implementation of miniaturized cell assay, micro-arraying is an essential part of the technology. Initially, we had performed cell-adhesion studies and evaluated ® ve diå erent substratesÐ polystyrene, polycarbonate , silicon, glass and PDMSÐ and found that with polylysine coating, the glass, silicon and PDMS substrates served well for cell attachment. However, the conventional well-less spotting and assaying method was diae cult to implement in producing a predetermined array pattern of cells.
To overcome the above diae culty, we further developed both bottomed and bottomless well arrays to address the need for cell culturing and assays. PDMS and glass were used as the bottom surfaces for both bottomless well silicon substrates and bottomed wells. Novel liquid distribution arrays were fabricated in silicon to facilitate initial coating and liquid exchange from the entire array. The arrays were implemented for culture of adherent and non-adherent cells, immunochemical assays, and automatic liquid transfer all in high-density format. The current array platform will be also useful in cell library construction, combinatorial library synthesis, and continuous homogeneous and heterogeneous assays.

Rajesh K. Maheshwari, Schering-Plough Research Institute, Union, NJ, USA Co-authors: Annaniy Berenshteyn, Gary Kowalski and Joseph Norgard
This poster is about a robotic powder dispensing system we have developed. It can dispense on average 10 mg § 5% from any of the 288 dispensers on its carousel, into a sealed 24/96-well deep-well microtitre plate. The powders are kept in a sealed environment and are diå erentiated for dispensing purposes by their dispensing parameters, which are stored by in a Microsoft Access database . The databas e entry is pointed Abstracts of papers presented at the 2001 ISLAR to by a barcode label on the dispenserÐ hence, adaptive dispensing.
The user interface for this multiprocessor system consists of 250 000 lines of Multithreaded/Multitasking Visual C ‡ ‡ code. An Access database holds dispensing parameters' information for each powder. A Dynamic Query Screen has been designed to make it easy for a chemist to view the database selectively.
Such a system can ® nd other uses such as distributing compounds in the compound distribution centre, dispensing lyophalized microorganisms, dispensing resins for combinatorial chemistry, and any other application when powder is to be delivered precisely in a controlled environment.

Randall Engler, Kendro Lab Products, Newtown, CT, USA Co-authors: C. Elliot and R. Moody
Cell proliferation, cytotoxicity and other viability assays play an important role the drug-discovery process. Many screening programmes target anticancer compounds, requiring in-vitro characterizatio n of eae cacy. Additionally, all potential therapeutic compounds must be characterized for their cytotoxicity regardless of the target (ADME/Tox).
Cytokinetics has developed a system to automate the process of reagent addition, incubation and reading of cell viability assays using the MTS method of Promega, Inc. The integrated system includes hardware from a variety of sources and software developed by Cytokinetics.
Control of environmental conditions including CO 2 , temperature and humidity are important factors in accurate and reproducible cell-based assays. The system described here includes the Heraeus 1 Cytomat 1 6000 series incubator from Kendro for the incubation of cells.

Automation development for high-throughpu t phage library screening
Randy Yen, Genentech, Inc., San Francisco, CA, USA Co-authors: Sherry Yeh and Suki Hyare Phage display technology is a powerful tool in genomic and drug development. It allows scientists to screen quickly billions of peptides, antibodies and cellular proteins for binding to a target. It is a vital tool in studies aimed at identifying molecules that bind to a speci® c target and at improving particular features of pre-existing molecules. At Genentech, we automated the binding assays for the speci® c target to screening the phage library. By running fully automated 384-well assay on the robotic system, we screened thousands of clones in a multiple target in 1 day and ® nished each library in a short time. We knew as a laboratory that we could not function with only one type of dissolution system. Our research led us to two automated options. The ® rst, a peristaltic pump sampling system using a single-¯ow cell approach, oå ers the detection of the sample. The second, a syringe pump sampling system using a multi¯ow cell approach, provided us the options to detect, detect and collect, and collect and dilute the sample. Each of these dissolution set-ups is paired with an HP8453 UV-VIS while remaining 21 CFR Part 11 compliant.

Cross-functiona l team approach to implementin g and managing laboratory automation in the quality control laboratory
Timothy Reilly, Novartis Pharmaceutical Corporation, Su¡ern, NY, USA Poorly de® ned project goals and an inadequate infrastructure have hindered previous attempts to initiate automation projects in the Novartis Quality Control laboratory. This haphazard approach has been replaced with well-de® ned project expectations and agreement among the functional groups that make up the project infrastructure to support aggressively the automation project goals.

Rolf Rolli, Sotax Ltd, Allschwil, Switzerland
New FDA regulations allow biowaivers for Class I drugs based on solubility, permeability and dissolution testing. BCS-based biowaivers can be requested for signi® cant post-approva l changes (e.g. Level 3 changes in components and compositions) to a rapidly dissolving immediate release product containing a highly soluble, highly permeable drug substance, provided that dissolution remains rapid for the post-change product and both pre-/post-change products exhibit similar dissolution pro® les. Dissolution tests have to be performed at pHs 1, 4.5 and 6.8. With this regulation, automation of dissolution testing gains further importance with respect to pH changes.
A fully automated test system with the necessary software is described. This solution oå ers one the capability to run up to 15 USP 2 tests in series. With such a system, all steps are fully automated, from medium preparation, to tablet input and up to the printout of the reports. The system includes a very eae cient cleaning device that prevents any carry-over between tests. Tests requiring baskets are handled with the Basket-Station . This system allows up to 10 USP 1 tests to be analysed. With this automated dissolution concept, biowaiver studies are executed rapidly allowing for accelerated drug development and SUPAC changes.

Integration of reagent management and computationally biased combinatoria l synthesis
Scott M. Harris, DuPont Pharmaceuticals Research Laboratories, San Diego, CA, USA One of the key aspects to successful high-throughpu t drug discovery is the ability to integrate several functions. These include medicinal chemistry, computationa l design, high-throughpu t synthesis and puri® cation. A potential bottleneck in this process is a lack of established chemistries linked with readily available commercial and proprietary building blocks. Early in the process, it is critical that a protocol exists that allows for planning and synthesizing libraries, developing initial structure± activity relationships and validating screening leads.
It is also important to have access to novel building blocks and reagents for preparing targeted or focused libraries. DuPont Pharmaceuticals has developed a simple custom Web-based monomer request system that uses a JAVA GUI that allows for tracking of orders and report generation. The data are housed in a Reagent Inventory Tracking System using Oracle and the physical reagents are stored in a controlled ventilated cabinet system. The typical throughput and average turn around time will be discussed. The overall bene® ts of this programme are lower costs, better inventory control and an increase in eae ciency in library production and chemistry. Conversion of our core DNA sequencing laboratory to the 3700 DNA analyser (AB) and sequencer for BioLIMS sequence analysis and databas e software (AB, genecodes) has greatly reduced the amount of manual intervention required for generating, analysing and maintaining DNA sequence data.

Automation of chemistry in
The set-up of the DNA sequencing reaction is the next process that has been selected for improvement. Currently, requests for sequencing from the Therapeutic Area laboratories are (1) submitted online, (2) the sample information is transferred to the 3700 plate record and (3) the addition of DNA and primers to the thermal cycle plate is performed manually.
Manual chemistry requires up to 4 h of hands-on time per day depending on the number of samples. While a liquid handler working oå a ® le can perform the hit-picking required for core terminator chemistry, the set-up of the source racks is time-consuming and a possible source of errors. The poster will present the use of Web submission, individually bar-coded sample tubes and a liquid handler for automating chemistry set-up in a core DNA sequencing facility.

Shane Weber, Millennium Pharmaceuticals, Cambridge, MA, USA
Millennium' s nylon cDNA micro-array platform is successfully ® lling our drug-and marker-discovery platforms. This platform is eå ective because of its sensitivity, gene content, throughput eae ciency and cost. The platform' s eå ectiveness results from the industrialized process¯ow of liquid handling, printing and information tracking. Eå ectiveness is constantly assessed by measurements of quality and eae ciency. The process ow and quality assessment will be reviewed.

William R. Hall, GlaxoSmithKline, Research Triangle Park, NC, USA Co-authors: Melissa Lindsay and Dean Phelps
The Library Puri® cation Group at Research Triangle Park is responsible for puri® cation of both large and small libraries from discovery and targeted research. We use mass-directed preparative HPLC to purify library samples with masses of 5± 100 mg per well. Our system incorporates at-column dilution, separate waste collection for each sample and UV veri® cation of peak collection. Eae cient evaporation of solvents, robotic liquid handling and reformatting, and fast QC on puri® ed samples allows for high-throughpu t of puri® ed libraries. As part of the evolution and industrialization of the BMS Lead Discovery organization, a core group of specialized scientists focused solely on automated screening was assembled. The mission of the Automated Screening Core is to adapt and prosecute high-throughpu t screen campaigns on fully automated robotic screening systems. This talk will cover the implementation of the BMS automated screening infrastructure, the rationale for the establishment of the Automated Screening Core and the organizationa l impact of such a group.

Determining feasibility and parameter values for compound orders based on transfer amounts, liquid-handler capabilitie s and container-volum e capacities
Steven Ho¡man, Bristol-Myers Squibb PRI, Princeton, NJ, USA Co-author: Mark F. Russo An algorithm is described for determining the feasibility and calculating parameter values for one or more compound processing orders. Feasibility is determined primarily by applying a series of inequality constraints that result in a valid target concentration range within which orders can be processed. These inequality constraints are obtained from initial compound amounts, source and destination container-volum e capacities, the liquid-handler transfer volume range and the amount of compound to be processed. If feasible, calculated parameter values include the dilution, transfer and top-oå volumes necessary to ful® l the order. Sample code implementing the algorithm is given in Microsoft Visual Basic 1 . We carried out a¯uorescence polarization-base d (FP) high-throughpu t screen (HTS) to detect inhibitors of the 3 0 -5 0 -exonuclease activity of the human cytomegalovirus DNA polymerase (the HCMV exonuclease). We used a 23-mer ssDNA substrate labelled with TAMRA at the 3 0 end. The emitted light from TAMRA in an intact substrate molecule was highly polarized (about 240 mP), but, upon substrate digestion with the HCMV exonuclease, the emission from TAMRA free in solution exhibited a lower polarization (about 40 mP).

High-throughput screen to detect inhibitors of the
The assay was designed to run in a 384-well format with a ® nal volume of 50 ml. The K m of the HCMV exonuclease for the substrate was determined to be 20 nm. In the assay, 50 nm substrate was digested by 0:01 £ stock HCMV exonuclease in 50 mm HEPES, pH 7.6, which produced a 50% reduction in mP in 8.3 min. We screened 117 843 compounds at 10 mm (2.5% DMSO). The screen identi® ed 647 compounds that inhibited the HCMV exonuclease 3 30%.

High-throughpu t protein function screening for immune system target validation
Stewart D. Chipman, Immunex, Corporation, Seattle, WA, USA Validation of the biochemical function of a novel gene/ protein in the pathophysiolog y of a disease is the ratelimiting step in novel target validation for drug discovery. Many groups have applied the technology and methodology developed for small molecule high-throughput screening and clinical testing to assays that test for novel protein function. This presentation will address how we have integrated gene discovery and expression analysis, protein production and inventory capability, cell-based protein function assays, laboratory automation tools and data management systems into an integrated target discovery and validation platform for immune system targets. Topics to be discussed are automated and multiplexed cytokine detection, cognate ligand detection, calcium mobilization, high-throughpu t optical imaging, large-scale cell preparation, large-scale protein expression and puri® cation, workstation style laboratory automation and automated data handling.

Laboratory automation: a cost-e´ective solution to increasing laboratory workloads
Stephen Green, Forest Laboratories Ireland Ltd, Dublin, Ireland Forest Laboratories, Inc., markets Celexa, an SSRI antidepressant, in the USA. The bulk tablets are produced at the manufacturing site in Dublin, Ireland. Market projections in 1999 indicated a 250% increase in volumes to be produced at the Dublin site. Several laboratory strategies to cope with this increased production were assessed and the use of automated robotic systems was shown to be the most cost-eå ective solution.
The introduction of the automation into the laboratory had several phases. Initially, there were physical changes required to the laboratory layout to accommodate the robots. Second, there was the quali® cation of the robotic systems and the training of the initial users. Finally, there was the development and validation of several analytical methods, which was performed in conjunction with the Zymark MTOV group.

Characterizatio n and e´ective use of the molecular properties of reaction vessel surfaces in high-throughpu t assay development
Sven Erik Rasmussen, Nalge Nunc International, Naperville, IL, USA Co-author: Nalge Nunc An understanding of the molecular surface of a reaction vessel is critical to achieving the sensitivity and speci® city for high-throughpu t or high-content molecular and cellbased assays. We developed and characterized several molecular surfaces for reaction vessels that can be eå ectively used in genomic and proteomic assays. These surfaces, in conjunction with the physical properties of the substrate material (optical or thermal properties, for example) along with format (MicroWell, slide, or chip), are compared with regard to assay performance.
Through XPS analysis (also known as ESCA), the molecular and atomic compositions of various MicroWell surfaces have been de® ned. These data will be used to demonstrate the speci® c utilization of the surface in molecular and cell-based assays. For example, a hydrophilic, highly charged surface, MultiSorp, preferentially binds complex phospholipid molecules while excluding other molecules, including glycoproteins such as IgG. Another`high-binding' hydrophilic surface, MaxiSorp, binds a signi® cant amount of glycoproteins such as immunoglobins. These surfaces and others such as the Nunclon Delta cell culture surface and non-treated sterile polystyrene surfaces are molecularly de® ned by ESCA techniques accompanied by speci® c procedural examples.
Well geometry, round, square, columnar or`shallow' , in¯uences assay performance, particularly when an active surface is used. Additional parameters come into eå ect when well volumes are miniaturized as in a 1536well plate. The eå ect of a molecular surface changes again as the assay is further miniaturized to a twodimensional format as in a slide or chip and the in¯uence of the substrate material may increase.
Also important to the development of a speci® c, sensitive and eae cient assay are the physical characteristics of various substrate materials. With regard to optical properties, the¯uorescent and re¯ective properties of solid black and solid white as well as black or white optical bottom plates (OBP) were compared by measuring signal-to-noise ratios, light cross talk and sensitivity. Data and practical applications will be discussed.

New polymer substrate for DNA micro-array s
Svend Erik Rasmussen, Nalge Nunc International, Naperville, IL, USA Co-authors: T. Kristensen, K. HolmstrÖm and L. Pedersen The quality and surface properties of the micro-array substrates are fundamental factors for a successful DNA micro-array system. At present, coated glass-slides (amine-, alehyde-or epoxy-coated) are the dominating substrates on the market. However, an often-encountered problem using these substrates is a non-homogenous surface, which can lead to problems with spot uniformity and morphology. Nunc A/S has developed a pre-activated, but uncoated, NucleoLink TM polymer MicroArray Slide that can be used, without any further modi® cation, for creating DNA micro-arrays. The Nu-cleoLink TM polymer MicroArray Slide can be used for both covalent attachment of DNA as well as immobilization of DNA by electrostatic forces and/or ionic bonds.
Studies of auto¯uorescence properties, DNA-binding capacity and DNA-hybridization eae ciency have been performed and compared with conventional glass sub-strates. Data from oligonucleotide arrays as well as cDNA arrays will be presented.

Flexibility paradigm in laboratory automation management
T. C. Ramaraj, Roche Discovery Technologies, Ho¡mann-La Roche, Nutley, NJ, USA Considerable developments and rapid changes have taken place in high-throughpu t screening and compound management in the last 5 years. This is a result of a concerted eå ort between the pharmaceutical industry, laboratory automation companies, instrument companies and manufacturer s of consumable laboratory ware. The evolution of new automation technologies has fuelled the growth and prospects for various modes of automation in virtually all phases of drug-discovery research. Here, we will try to highlight our experiences in managing a complex HTS and compound management infrastructure at Hoå mann-La Roche.
The infrastructure consists of automation platforms that have evolved over several years. All require proper care and maintenance for sustained performance. Some recently installed con® gurations require continued validation to achieve expected levels of performance and reliability to match ever-increasing throughput requirements. Promising new automated screening technologies require careful evaluation. Validation of these new technologies also require allocation of signi® cant resources for meeting demanding requirements of miniaturization, speed,¯exibility, cost eae ciency and increased performance levels. It is no easy task to steer winning methodologies to their fullest potential and at the same time take a risk in exploring unproven avenues and technologies.
To maintain a balanced perspective among available choices requires¯exibility in approach, quick implementation strategies, a diverse pool of expertise and talent that can work as a team, contribute as individuals, share the successes as well as failures, learn from the mistakes, revise the expectations and still meet corporate goals and objectives every year. The presentation will also include examples from some of our most recent integration experiences, the upgrade of existing systems, the implementation of new detection modes in existing instruments and our experience with latest 384-well pipetting systems.

Variety of automation tools for drug metabolism
Thomas Lloyd, DuPont Pharmaceutical Co., Newark, DE, USA Automation continues to be developed for new diversi® ed applications within drug metabolism. The choice of an automated tool can vary considerably depending on the speci® c requirements of the application. An overview of the automated tools that have been introduced at DuPont over the last 5 years will be presented along with speci® c application examples. An evaluation is oå ered for how the diå erent tools have proliferated throughout this function.
Automation application areas include bioanalytical sample preparation techniques including solid-phase extraction, liquid± liquid extraction, protein precipitation Abstracts of papers presented at the 2001 ISLAR and online approaches for in vivo samples, in vitro sample preparation processes, sample handling, data handling and training. Application areas include many functions within drug metabolism ranging from discovery through clinical trials.
Characterizatio n and e´ective use of the molecular properties of reaction vessel surfaces to achieve optimal assay performanc e Tom Cummins, Nalge Nunc International, Rochester, NY, USA Co-authors: Janne N. Knudsen, Lena B. Larsen, Svend Erik Rasmussen, Tom Cummins and Joseph Granchelli An understanding of the molecular surface of a reaction vessel is critical to achieving the sensitivity and speci® city for high-throughpu t or high-content molecular-and cellbased assays. We developed and characterized several molecular surfaces for reaction vessels that can be used eå ectively in genomic and proteomic assays. These surfaces, in conjunction with the physical properties of the substrate material (optical or thermal properties, for example) along with the format, are compared with regard to assay performance.
Through ESCA analysis, the molecular and atomic compositions of various MicroWell 1 surfaces have been de® ned. These data will be used to demonstrate the speci® c use of the surface in molecular and cell-based assays. For example, a hydrophilic, highly charged surface preferentially binds complex phospholipid molecules while excluding other molecules, including glycoproteins such as IgG. Another`high-binding' hydrophilic surface, MaxiSorp TM , binds a signi® cant amount of glycoproteins such as immunoglobins.
Well geometry, round, square, columnar or`shallow' , in¯uences assay performance, particularly when an active surface is used. Additional parameters come into eå ect when well volumes are miniaturized as in a 1536well plate.
Also important to the development of a speci® c, sensitive and eae cient assay are the physical characteristics of various substrate materials. With regard to optical properties, the¯uorescent and re¯ective properties of solid black and solid white as well as black or white optical bottom plates (OBP) were compared by measuring signal-to-nois e ratios, light cross-talk and sensitivity. Data and practical applications will be discussed.

Intelligent automated dissolution system! A challenge worth striving for
Umesh V. Banakar, Pharm-Assist International, Carmel, IN, USA Co-author: Manoj Bagool Dissolution testing requirements have been extended to all oral dosage forms over the past decade. Additionally, the dissolution regulations for pharmaceutical formulations have been tightened world-wide. Furthermore, the past decade has also witnessed an increase in popularity of modi® ed release pharmaceutical s that inherently require extended dissolution protocols. The increased emphasis on correlating dissolution and bioavailability data points to the desirability of more frequent sampling points to validate a dissolution pro® le. As a result, these developments have signi® cantly increased the requirements of equipment, personnel, data processing and validation, which translates into increased production time for oral dosage forms. Thus, automation in dissolution testing has become a necessity in order to handle these demands.
Automation in dissolution testing is not new to the pharmaceutical industry. Simultaneous to the development in pharmaceutical technology over the past decade, there have been advances in automated dissolution testing, although at a slower pace. The virtual explosion in the diverse and yet speci® c requirements associated with a plethora of dosage form types and their respective functions have curtailed the outright all-encompassing development in automated dissolution testing. While the advantages of automated dissolution testing are unequivocal, the limitations of automated dissolution testing systems cannot be refuted. There is a constant quest to improve an existing automated system that meets speci® c requirements, either a unit function of the test or to accomplish a speci® c expectation of a dosage form, thus limiting the universality of the automated system. Robert Kennedy once said,`Some people watch things happen and ask the question, why; and there are those that dream of things and ask the question, why not!' It is beyond doubt that an automated dissolution testing system that addresses the current limitations of the existing ones and yet is¯exible and universal enough is the need of the day. Going beyond the minimum expectations of such an automated dissolution test into its utility during early drug development and above all in simultaneous prediction of bioavailability, can be a dream worth striving for. A vision for such an automated dissolution test system will be presented.

Development of a radiofrequency identi cation (RFID) data carrier interface with an analytical balance
Je¡rey Veitch, GlaxoSmithKline, Ware, UK Co-authors: Tony Allcock and James Ormand The use of process instrumentation such as analytical balances within the laboratory is commonplace and the need for capturing related process measurement data are often a necessity. Quite often these data need to be exported to databases and invariably to laboratory information management systems (LIMS). Laboratories regularly use process equipment such as balances, pH meters, etc. that are not attached to a dedicated PC. At GlaxoSmithKline, we have developed a way of capturing process data (e.g. weight values) directly to radiofrequency identi® cation (RFID) smart labels, where these labels are used as data carriers.
This evaluation has proven to be a success and is potentially the stepping stone towards a more speci® c use of RFID within GlaxoSmithKline.

System integration for genomics using biorobot workstations
Achim Wehren, QIAGEN Hilden, Germany, and Valencia, CA, USA Co-authors: Fred Siegmanand Carola Schade Automated robotic workstations are frequently used as stand-alone systems for sample puri® cation, reaction setup or sample re-array in the ® eld of genomics. However, a variety of external instruments often must be integrated to provide full automation of several sequential tasks or to increase the range of applications that can be performed.
QIAGEN 1 Instruments provides complete solutions to automate molecular biology and liquid-handling applications by integrating BioRobot workstations with a wide range of accessory instruments, including thermal cyclers, spectrophotometer s and the BioRobot RapidPlate 1 , a 96-channel pipetting system with capabilities for both 96-and 384-well pipetting. These integrated systems provide rapid and fully automated processing for applications such as DNA template normalization, PCR, sequencing reaction set-up and sample transfer tasks.
BioRobot 1 3000 extended-arm systems are designed to accommodate complementary instruments on both the left and right sides. Additional integration of the BioRobot Twister TM increases system storage capacity for microplates, blocks and disposable tips, and allows the integration of accessory instruments with both extended-arm and standard BioRobot workstation con-® gurations.
BioRobot 1 8000 robotic workstations can be integrated with accessory instruments using the BioRobot Twister II external arm for walkaway front-end processing, e.g. from plasmid puri® cation to ® nal reaction set-up. The large size of the BioRobot 8000 worktable and the high storage capacity of the Twister II robotic arm allow unattended operation over extended periods.
QIAsoft TM , the BioRobot Operating System, communicates with external instrument software, controls processes performed on the BioRobot and coordinates the actions of the BioRobot and external instruments.
Dispensing precision for the SciClone TM automated liquid-handlin g workstation for 96-channel pipetting

Rudy Willebrords, Janssen Research Foundation, Beerse, Belgium
One of the ® rst steps in drug discovery involves identi® cation of novel compounds that interfere with therapeutically relevant biological processes. Identi® cation of`lead' compounds in all therapeutic areas included in a drugdiscovery programme requires labour-intensive evaluation of numerous samples in a battery of therapy-targete d biological assays. To accelerate the identi® cation of lead compounds, JRF has developed an automated highthroughput screening (HTS) based on the unattended operation of a custom Zymark-tracked robot system. Automation of enzymatic and cellular assays was realized with this system adapted to the handling of microtitre plates. The microtitre-plate technology is the basis of our screening. All compounds within our chemical library are stored and distributed in micronic tube racks or microtitre plates for screening. An eae cient in-house-develope d mainframe-base d laboratory information management system supported all screening activities. Our experience at JRF has shown that the preparation of test compounds and making serial dilutions have been rate-limiting steps in the overall screening process. To increase compound throughput, it was necessary both to optimize the robotized assays and automate the compound supply processes. In HTS applications, one of the primary requirements is highly accurate and precise pipetting of microlitre volumes of samples into microplates. The SciClone TM is an automated liquid-handling workstation capable of both 96-and 384-channel high-precision pipetting. For high-throughpu t applications, the SciClone TM instrumentation can pipette a variety of liquid solutions with a high degree of accuracy and precision between microplates (interplate variability) and tip-to-tip (intraplate variability) within a single plate. The focus of this presentation is to review the liquid-handling performance of the SciClone TM system as a multipurpose instrument for pipetting aqueous or organic solutions, and virus suspensions into 96-and 384well microplates. The capabilities of the system and the resulting bene® ts for our screening activities will be described.
Abstracts of papers presented at the 2001 ISLAR