In Silico Screening of the Library of Pyrimidine Derivatives as Thymidylate Synthase Inhibitors for Anticancer Activity

: We here report the virtual screening of several series of pyrimidine derivatives for in silico Thymidylate Synthase (TS) inhibition to arrive at possible potential inhibitors of TS with acceptable pharmacokinetic or ADME (Absorption, Distribution, Metabolism and Excretion) properties. Library of the molecules was constructed based upon structural modifications of pyrimidines nucleus. Structural modifications in descending order were performed for the series of pyrimidines, viz from pyrimidines with five membered heterocyclic ring to pyrimidines with four membered heterocyclic ring to simple pyrimindine carboxylates in an order to arrive at pyrimidines with better inhibition scores ( G -Scores) as compared with Raltitrexed (RTX) and active metabolite of 5-Fluorouracil (5-FUMP). The molecules with better G -Scores were subjected to predict pharmacokinetic or ADME properties. The molecules with acceptable ADME properties and better G -Scores were prioritized for synthesis and anticancer evaluation. Three molecules from pyrimidine carboxylate series PIC 1 - 31 were found acceptable with G -Scores and pharmacokinetic properties. Thus a library of pyrimidine derivatives was constructed based upon the feasibility of synthesis and in silico screened to prioritize the molecules and to obtain potential lead molecules as TS inhibitors.


Introduction
In silico screening methods such as docking have a great advantage as compared to 2D similarity and 3D pharmacophore search methods as it utilizes the 3D receptor structure in a quantitative way 1 .Docking calculations alone or combined with the virtual screening has been carried out to develop the DHFR and tyrosine kinase inhibitors 2 , agonists and antagonists of H3 receptor 3 , acetylcholine esterase inhibitors 4 , glycogen phosphorylase inhibitors 5 , thymidylate synthase inhibitors 6 , glutathione and trypanothione reductase inhibitors 7 , COX-1 inibitors 8 , etc.

Thymidylate synthase
The TS (E.C. 2.1.1.45),is the most crucial enzyme in the folate pathway and plays significant role in the DNA synthesis and cell division [9][10] .Its inhibition has significant role in cancer 11 .5-Fluorouracil (5-FU) and its active metabolite 5-fluorouridinemonophosphate (5-FUMP) is the only available pyrimidine analogue used clinically for the treatment of various cancers and it acts by TS inhibition mechanism 12,13 .We here report the in silico prioritization of library of molecules to generate potential lead molecules as TS inhibitors with acceptable ADME properties.

Experimental
The computation was carried out in Schrödinger molecular modeling software.Molecular docking was performed using the GLIDE ® integrated Maestro ® 7.5 interface on the Linux operation system.The molecules were subjected to predict the pharmacokinetic or ADME properties using the QikProp ® 2.5 module.The ChemOffice software was used to draw molecular structures and for the conversion of the structure to mol files.

Selection of the protein file
For the docking purpose the PDB (Protein Database Bank) file 1I00 was selected after evaluating several files from the Protein Database bank.The file contains the 3D crystalline structure of thymidylate synthase from human origin.The enzyme file was subjected to structure validation procedures.The incorrect amino acid residues and their formal charges were adjusted at this step.

Structure validation of the enzyme
The errata report and the Ramachandran plot was obtained from the NIH MBI sever for evaluation of protein structures and are given in Figure 1& 2.

Molecular modeling Designing of the molecules and library design
A library of the molecules was generated by performing structural modifications in descending order on five series of pyrimidines each containing 124 molecules.Structural modifications were carried out on pyrimidines, viz from pyrimidines with five membered heterocyclic ring to pyrimidines with four membered heterocyclic ring and finally to simple pyrimindine carboxylates in an order to arrive at molecules with better in silico inhibition G-Scores as compared with Raltitrexed (RTX) and active metabolite of 5-Fluorouracil

Docking of the molecules
The ligands were prepared by LigPrep module of the software which produces a single lowenergy 3D structure with correct chiralities for each successfully processed input structure based upon the OPLS-2005 Molecular mechanics force field.RTX and 5-FUMP were included in the ligand sets to give inhibition scores (G-Scores) as compared with these drugs used clinically.Where Ar =Ar'' and ., Ar , Ar'' are different :

Energy minimization of the protein
The crystal structure of the human TS enzyme (1I00) was obtained from the protein data bank (RCSB PDB) and contains the chain A of the enzyme complexed with the known inhibitor Raltitrexed (RTX).This chain A was selected for the docking studies.The protein preparation wizard was used for the protein preparation where, the energy minimization was carried out at the default cut off (Root Mean Square Deviation) RMSD value of 0.30Å.

Receptor grid preparation 15
The co-ordinates of the human TS chain A were obtained from the crystal structure of the protein file 1 I00 (processed file from the protein preparation wizard).The van der Waals radii were scaled up by the default value of 1.00 Å for the atoms with the partial charges of less than 0.25.The receptor grid was generated around the centroid of the ligand contained by enzyme file and the ligands with cut off size of 10 Å were allowed to dock.

Docking of the ligands
Docking was carried out using the Glide module of software which uses MCSA (Monte Carlo Based Simulated Algorithm) based minimization 16 .The ligands were docked flexibly to write up to 10000 poses per ligand in the standard precision mode.This produced result of docking of the ligands having the G-Scores as shown in Table 1.
Table 1.G-Scores of the two best scoring analogues from series 1-4 and best scoring three ligands from series 5 (PIC 1-31 ).
(4 th and 5 th Best scoring Ligands from Series 5 not mentioned.)

Validation of the docking protocol
Validation of the docking protocol was carried out by following methods: 1) Docking of the inhibitors with reported Ki values with TS enzyme and then finding a correlation coefficient with G-Score, as shown in Figure 6. 2) Pose regeneration of the ligand as seen in the crystallized PDB, overlapping the best dock pose on PDB File 1I00 to obtain same pose as seen in the PDB 1I00.

ADME or pharmacokinetic predictions of the best fit molecules
The ligands with the comparable scores with Raltitrexed were subjected to predict pharmacokinetic properties using the QikProp module of the software.QikProp 17 settings determine which molecules are flagged as being dissimilar to other 95% of the known drugs.Predicted significant ADME properties such as permeability through MDCK Cells (QPlog MDCK), Qik Prop predicted log IC 50 value for blockage of K + channels (QPlogHERG), Qik Prop predicted gut-blood barrier (QPPCaco) and violations of the Lipinski's rule of five (LROF) are reported here .The number of stars indicate the deviations from the 95% of the known drugs.These properties are mentioned in

Validation of the Docking 18
Structures with Ki values

Results and Discussion
Literature reveals that active binding site of TS enzyme comprises of Leu-221, Phe-225 and Ile-108.The PDB file 1I00 contained ternary complex of Tomudex with the enzyme TS.All the best scoring ligands shown in Table 1.Ligands from series 5 viz pyrimidinyl carboxylates were found to interact with at least one of the amino acid residue of the active binding site.Moreover all the molecules (PIC 1-3 ) were subjected for the ADME predictions and passed the ADME predictions with least number of stars and least number of violations, for being the suitable drug candidates for TS inhibition.The docking protocol was validated by correlation coefficient method and overlapping method.RTX was included in the docking run and the same pose as seen in PDB file 1I00 was regenerated confirmed by overlapping method.Moreover the correlation coefficient r 2 of 0.934 was obtained by comparing the reported experimental Ki values with G-score.It was found that heteroaryl substituents such as imidazolidinones, thiazolidinones, oxazolidinones and azetidinones built on the pyrimidinyl nucleus (Series 1-4) not only violate the ADME predictions but also results in low scoring compounds as compared with G-Scores of clinically used drugs such as RTX and 5-FUMP.

Conclusion
A library of pyrimidine derivatives was built and in silico screened to prioritize the molecules for synthesis and anticancer evaluation.Three molecules from the PIC series were in silico prioritized for the synthesis and actual in vitro anticancer evaluation from library of 2000 ligand molecules.

Figure 3 .
Figure 3. Virtual screening for arriving Good G-Scores for inhibition of TS as compared with 5-fluoro-uridine monophosphate.Four membered ring molecules not shown.

Table 2 .
ADME or Pharmacokinetic predictions of the molecule.