Serum Biomarkers Identification by Mass Spectrometry in High-Mortality Tumors

Cancer affects millions of people worldwide. Tumor mortality is substantially due to diagnosis at stages that are too late for therapies to be effective. Advances in screening methods have improved the early diagnosis, prognosis, and survival for some cancers. Several validated biomarkers are currently used to diagnose and monitor the progression of cancer, but none of them shows adequate specificity, sensitivity, and predictive value for population screening. So, there is an urgent need to isolate novel sensitive, specific biomarkers to detect the disease early and improve prognosis, especially in high-mortality tumors. Proteomic techniques are powerful tools to help in diagnosis and monitoring of treatment and progression of the disease. During the last decade, mass spectrometry has assumed a key role in most of the proteomic analyses that are focused on identifying cancer biomarkers in human serum, making it possible to identify and characterize at the molecular level many proteins or peptides differentially expressed. In this paper we summarize the results of mass spectrometry serum profiling and biomarker identification in high mortality tumors, such as ovarian, liver, lung, and pancreatic cancer.


Introduction
Cancer-related mortality is one of the leading causes of death worldwide. e most effective treatment to �ght cancer is still early diagnosis. On the other hand, it is known that the correct classi�cation of the tumor, coupled to a suitable therapy and to a stringent follow-up, helps to prevent and detect relapses. Cancer is a very heterogeneous disease, and, at the diagnostic level, is de�ned by many indexes such as histological grade, tumor stage, patient age, sex and, more importantly, genetic background and pro�les. Histological evaluation of tumor specimens obtained from tissue biopsy is the gold standard of diagnosis, but oen tumors with the same histopathological features respond differently to the same therapy. New generation diagnostic platforms, previously unavailable, have enabled to better characterize transcriptomic signatures that predict tumor behaviour, helping to de�ne diagnosis, prognosis, and the most appropriate therapies [1][2][3]. Tumor biomarker discovery in biological �uids, such as serum, plasma, and urine, is one of the most challenging aspects of proteomic research [4]. Many researchers have attempted to identify biomarkers in serum that re�ect a particular pathophysiological state. Since the expressed proteins, native, fragmented, or posttranslationally modi�ed, quickly change in response to environmental or pathological stimuli, the serum proteome is considered dynamic, oppositely to the stable nature of the genome. Proteins and their functions can determine the phenotypic diversity that arises from a set of common genes. e study of the serum proteome highlights differences in protein expression re�ecting a speci�c pathological state and provides useful information to diagnose a disease, to evaluate prognosis or therapy response [5].
cancer, are currently used. Most of these molecules have been isolated from animals immunized with tumor cells extracts or cell lines, with subsequent screening for monoclonal antibodies against cancer-associated antigens [6]. e abovementioned proteins increase the accuracy of diagnosis, even though there is an urgent need to isolate and use in clinical practice more speci�c biomarkers, or groups of biomarkers, to precisely characterize the disease at the diagnostic or prognostic level and to monitor its progression [7,8]. Biomarkers could also help to predict the response of the patient to anticancer therapy and thereby to guide physicians in choosing the best treatment. is research is more appealing due to the simplicity of obtaining blood samples, but, at the same time, shows limits due to the complexity of the serum protein mixtures. A plethora of molecules from almost every tissue of the body can be found in human serum/plasma. Many of the serum proteins are present at very low concentrations (less than pg/mL), while others are present in very large amounts (more than mg/mL). Serum and plasma are very complex mixtures of proteins and exhibit a broad dynamic range of relative abundance (up to 12 orders of magnitude) [9]. ey contain thousands of proteins, whose some are very abundant (e.g., albumin, immunoglobulins, apolipoproteins) and constitute approximately the 95% of the total protein content, but only the 0.1% of total protein species [10,11]. For these reasons, it is thought that many potentially important proteins and markers, if present at low concentrations, can escape the detection. Evidences show that the circulating fragments from unmodi�ed or post-translationally modi�ed proteins generated in the tumor tissue microenvironment can be used as diagnostic or prognostic markers. Proteolysis within the tissue or deregulated post-translational events (e.g. phosphorylation) generate protein fragments that diffuse into the circulation and could give information about the presence or the progression of the disease, then facilitating the management of the tumor. Among these fragments, the fraction with low molecular weight, the peptidome (<20 KDa, LMW peptides), is protected from renal clearance by interaction with abundant serum proteins and, in particular, seems to be an important source of biomarkers [12].
In order to simplify the biomarker discovery process, many prefractionation methods have been developed to remove highly abundant proteins. ese useful precursors of proteomic analysis include the use of immobilized dyes (cibacron blue) [13,14], immunoaffinity-based techniques [15,16], solid phase fractionation [17], liquid chromatography [18], or low-molecular weight fraction enrichment [19][20][21]. A promising method under investigation is based on the use of N-isopropylacrylamide (NIPA) nanoparticles which allow a fast one-step capture and concentration of analytes less than 20-25 KDa in molecular weight [22][23][24]. At the same time, the nanoparticles are able to protect the proteins from the degradation due to the "ex vivo" enzymatic activity of serum proteases, and, when conjugated with suitable chemical baits, show higher capability to sequester and retain many different proteins from whole serum, on the basis of their chemical and physical properties [25]. With this method, the captured analytes can be recovered by a simple electroelution and then analyzed by HPLC-MS/MS, western blotting or immunoassays for complete molecular characterization. To reduce the complexity of serum proteome, several studies are focused on targeting a speci�c subset of serum proteins [26]. Glycosylation is one of the most common posttranslational modi�cations in proteins. Approximately 50% of known eukaryotic proteins are glycosylated [27]. It is known that cancer cells express aberrant glycosylation patterns [28]. e analysis of glycosylated proteins (glycoproteome) has received great interest, because of the glycoproteic nature of the currently used cancer biomarkers. Two major methods have been developed to enrich glycoproteins or glycopeptides, based on chemical capture (reaction between aldehyde groups and hydrazide) [29,30] and lectin-affinity capture (speci�c recognition of protein glycan moieties by lectins) [31].

Mass Spectrometry: The Cancer Biomarker Discovery Tool
Many studies have been focused on the identi�cation of new serum biomarkers by mass spectrometry (MS). is powerful method enables to identify a protein without requiring the knowledge of its amino acid sequence. Further improvement of this technology has provided high accuracy to de�ne mass-to-charge ratio ( and to generate highresolution spectra. In addition, the development of tandem mass spectrometry (MS/MS), able to provide de novo protein sequence information, has enhanced the applications of this technology in proteomics [32,33]. Several MS methods have been used to characterize body �uids. Different combinations of ionization sources (e.g., MALDI, ESI), analysers (e.g., time of �ight TOF, quadrupole, Fourier transform and quadrupole ion traps), and fragmentation methods (e.g., CID collision induced dissociation, ETD electron transfer dissociation) can be used. MALDI-TOF (Matrix-Assisted Laser Desorption/Ionization-time of �ight) [34] is based on a so ionization method where a laser beam generates evaporation of a crystallized sample-matrix mixture. MALDI is used in biochemical areas for the analysis of proteins, peptides and oligonucleotides. Surface-Enhanced Laser Desorption/Ionization Time-of-Flight (SELDI-TOF), a modi�cation of MALDI-TOF, allows the identi�cation of proteins differentially expressed in serum by applying a small amount of sample directly on an array surface involving various chromatographic models based on classical chemistries (i.e., normal phase, hydrophobic, cationand anion-exchange), affinity-coated surfaces (IMAC, immobilized metal affinity capture), or biomolecular affinity probes [5,35,36], with minimal requirements for puri�cation and separation [37]. Selectively retained proteins are analyzed by laser desorption and subsequent ionization. e results are shown by a mass spectrum identifying ratios and peak intensities of peptides/proteins. Data preprocessing (i.e., calibration, baseline correction, normalization, peak detection, and alignment), bioinformatic and statistical analyses are performed in order to highlight and characterize any protein differentially expressed. SELDI-TOF allows the detection of many low molecular weight proteins [38].
Liquid chromatography/electrospray ionization tandem mass spectrometry (HPLC/ESI-MS/MS) technology is an alternative approach for serum biomarker identi�cation. e mixtures of analytes are subjected to HPLC then the solution is nebulized under atmospheric pressure and exposed to a high electrical �eld which generates a charge on the droplets� surface [39]. Due to the evaporation, droplets become much smaller and enter into the analyzer. HPLC-ESI-MS/MS couples protein fractionation with mass spectrometry, where peptide sequence tags can be produced from peptide fragments. Tandem mass spectrometry and data analysis by suitable bioinformatic tools, algorithms and databases, provides a powerful method to characterize peptides at the aminoacidic level, allowing a highly re�ned analysis [40].
e use of mass spectrometry for serum biomarker discovery is quite simple: spectral peaks (plots representing on the -axis the ratios of ions, and on the -axis the detected ion abundance), are identi�ed in a pathological group and compared with those obtained from normal control groups. Differences in spectral pro�les detect putative biomarkers. Essentially four strategies have been used in MS biomarker discovery: analysis of polypeptides separated by electrophoresis or chromatography, with or without prior fragmentation; analysis of enzymatic peptide fragments separated by HPLC and then analyzed by ESI or MALDI; analysis of proteins adsorbed on a solid surface; and analysis of speci�c serum fractions, such as the peptidome or the glycoproteic fraction [4].
Due to the lack of effective screening test, these methods have been applied to the serum biomarker discovery in many tumors, including those with high mortality, such as ovary, lung, liver, and pancreatic cancer. Here we report the results of studies focused on serum biomarker identi�cation in the above-mentioned types of cancer. Several protein pro�les and speci�c proteins (Tables 1, 2, 3, and 4) have been characterized and classi�ed as putative biomarkers.

Serum Biomarkers in Ovarian
Cancer. Despite advances in cancer therapy, mortality due to ovarian cancer is almost unmodi�ed during the last decades [50]. Ovarian cancer is usually diagnosed at a late clinical stage in more than 80% of patients [51]. In this group, the 5-year survival is approximately 35%. By contrast, the 5-year survival for patients with stage I ovarian cancer is more than 90% and surgery alone can be used as elective therapy [52]. Cancer antigen 125 (CA-125) is the most widely used biomarker for ovarian cancer. Elevated levels of CA-125 are detected in about 80% of patients with advanced-stage disease, but they are increased in only 50-60% of patients with early stage ovarian cancer [53]. e calculated positive predictive value for CA-125, considered as a single marker, is less than 10%, but this value was improved by ultrasound screening methods [54].

Proteomic Patterns for Ovarian Cancer Detection.
Analysis of sera by TOF-MS provided speci�c signature patterns which can be compared to distinguish ovarian cancer from benign disease or normal individuals. A training set of 50 sera from unaffected women and 50 from patients with ovarian cancer were analyzed by SELDI/TOF by using Platelet factor 4 (PF4) Decreased [49] a C16 hydrophobic interaction protein chip. An algorithm identi�ed a proteomic pattern able to distinguish cancer from non-cancer subjects. e pattern was then used to differentiate an independent set of 116 samples (50 ovarian cancer, 66 non-malignant disease or unaffected), yielding 100% sensitivity and 95% speci�city [67]. Some of the characteristic ion peaks in the previous signature have been sequenced and described in an independent cohort of ovarian cancer sera [36]. ovarian cancer, 15 benign, 5 normal). e four peak model showed a sensitivity of 90.8% and a speci�city of 93.5% in the training set, and a sensitivity of 87% and a speci�city of 95% in the blind test set in discriminating cancer from non cancer patients [69]. Hocker et al. [70] analyzed mass spectrometry peak differences in 35 ovarian cancer patients and 16 disease-free subjects. Proteomic pro�les distinguished early-stage from advanced cancer with a sensitivity of 80% and a speci�city of 93%.

Glycoproteomics in Ovarian
Cancer. An et al. [80] developed a glycomic approach to identify oligosaccharide markers for ovarian cancer by analyzing ovary cell lines supernatants and then con�rming their presence in sera from patients and healthy controls. Changes in glycosilation were monitored by MALDI-Fourier Transform Ion Cyclotron Resonance-MS. Approximately 15 unique serum glycan markers were detected in all patients and were absent in controls. Leiserowitz et al. [81] analyzed glycan markers and CA-125 levels in 48 sera from ovarian cancer women and 24 controls. Oligosaccharides were cleaved from serum glycoproteins and isolated using solid phase extraction. MALDI-Fourier Transformation-MS was used to identify peaks. Sixteen unique oligosaccharide signals were identi�ed in most of the cancer patients (44/48) and just in 1/24 controls (sensitivity 91.6%, speci�city 95.8%).

Ovarian Cancer �iomar�er Proteins �nvolve� in �n�am� matory
Processes. Several proteins involved in in�ammatory processes and acute-phase response have been identi�ed as putative biomarkers for ovarian cancer. In a study by International Journal of Proteomics 5 Zhang et al. [36], sera were fractionated by anion exchange chromatography. Aliquots were bound in triplicate with a randomized chip/spot allocation scheme to IMAC3-Cu, SAX2, H50, and WCX2 protein chip array. For biomarker identi�cation, proteins were puri�ed, separated by SDS-PAGE, and analyzed by MS/MS. ree proteins/protein fragments, described as acute-phase reactants (apolipoprotein A1, 28043, a truncated form of transthyretin, 12828, and a fragment of inter--trypsin inhibitor heavy chain H4, 3272), were identi�ed as putative biomarkers able to improve the detection of early stage ovarian cancer [36]. Ye et al. [42] identi�ed by SELDI an 11.7 kDa peak showing higher intensity in cancer sera. Aer protein puri�cation and LC-MS/MS analysis, the alpha chain of haptoglobin, an acute phase reactant, was identi�ed and further validated by western blot and ELISA as a potential biomarker for ovarian cancer, by using a speci�c polyclonal antibody against the peptide identi�ed by MS/MS analysis. e marker was 2-foldexpressed in cancer sera and had 64% sensitivity and 90% speci�city when used alone, or 91% and 95% sensitivity and speci�city, if combined with CA-125. Aer high abundance protein removal and two-dimensional gel electrophoresis (2-DE), Ahmed et al. [43] performed nanoelectrospray quadrupole-quadrupole time of �ight mass spectrometry (n-ESIQ-(q)TOF-MS) and MALDI/TOF analysis on six protein spots over-expressed in cancer sera. Protein isoforms of haptoglobin-I precursor (HAPI), a liver glycoprotein present in human serum, were identi�ed as putative novel biomarkers and con�rmed by 2-DE and western blotting on the serum from healthy controls and grade 1 and 3 ovarian cancer patients. Bergen III et al. [44] analyzed by nano-LC-ESI-TOF-MS or Fourier Tranform Ion Cyclotron Resonance (FT-ICR) the low molecular weight serum fraction obtained by ultra�ltration and identi�ed several candidate biomarkers. Among these, the �brinopeptide-A, already described as involved in acute phase reactions and elevated in many cancer including ovary, was detected. Two peaks (11.7 and 11.5 kDa) were identi�ed by SELDI-TOF in thermostable plasma fractions from 27 ovarian cancer and 34 control sera. A method involving cysteine modi�cations, 2-DE, and HPLC allowed to characterize the peaks corresponding to serum amyloid A1, an acute phase reactant, and its N-terminal arginine truncated form [45]. Kozak et al. [41] identi�ed by micro-LC-MS/MS four biomarkers for early stage ovarian cancer (transthyretin, apolipoprotein A1, transferrin, and betahemoglobin), corresponding to already described 13.9-TTR-, 12.9-TTR-, 15.9-Hb-, 28-ApoAI-, 79-Tf-kDa SELDI peaks [68]. Differential expression of these proteins in sera was also con�rmed by western blot and ELISA. Lopez et al. [47] set-up a work�ow using carrier protein-bound a�nity enrichment of serum samples directly coupled with MALDI/TOF. Subsequent tandem MS analysis de�ned the serum protein-bound peptides' sequence. e procedure was able to identify several speci�c biomarker panels to differentiate stage I ovarian cancer from unaffected and age-matched women. Among the peptides identi�ed, proteins involved in in�ammatory processes (complement component 3 precursor, complement component 4A preprotein, inter-alpha globulin inhibitor H4), as well as the glycosyltransferase-like 1B, a peroxisomal oxidation enzyme (D-amino-acid oxidase), two proteins involved in cancerogenesis (transgelin 2, casein kinase II alpha 1 subunit isoform a), �brinogen alpha chain isoform alpha preprotein, and keratin 2a were described as putative biomarkers. In a study on sera from patients with ovarian cancer, compared with sera from patients with benign masses or different type of cancer, it has been shown that the chemokines CC2 motif ligand 18 (CCL18) and CXC motif ligand 1 (CXCL1), identi�ed by MALDI-MS/MS analysis and further validated by ELISA, can be considered as novel circulating tumor markers for differential diagnosis between ovarian cancer and benign masses (sensitivity 92%, speci�city 97%) [48]. A nested case-control study performed on 295 sera from women pre-dating their ovarian cancer diagnosis and 585 matched control samples, showed that two peaks identi-�ed by MALDI, described as the connective tissue-activating peptide III (CTAPIII) and the platelet factor 4 (PF4), can be associated with CA-125 to improve early diagnosis [49].

Ovarian Cancer Biomarker Proteins Involved in Other
Functions. Woong-Shick et al. [46] studied by SELDI-TOF 35 sera from ovarian cancer patients in comparison to 10 from normal women. Aer protein puri�cation and N-terminal sequencing, hemoglobin-and chains were described as corresponding to the most distinctive peaks differentially expressed (15.1 and 15.8 kDa). ELISA validation test for intact hemoglobin indicated a sensitivity of 77% in sera from ovarian cancer patients. As above-mentioned, hemoglobin was also described as a putative ovarian cancer biomarker in a study by Kozak et al. [41].

Serum Biomarkers in Liver
Cancer. Liver cancer is oen diagnosed at very late stage and is associated to poor prognosis, high recurrence and mortality. Hepatocellular carcinoma (HCC), the most frequent liver neoplasm, is the �h most common cancer, affecting approximately one million people every year, with an incidence almost corresponding to death rate [94] and a 5 year survival ranging from 17% to 50%. Some predisposing factors, such as viral infections [95], diabetes, metabolic syndromes, exposition to a�atoxin, or alcohol consumption, are frequently related to liver tumor initiation. is allows a management of the patients at risk, making it possible in some cases to diagnose the disease earlier. e most important liver cancer serum biomarker is alpha-fetoprotein (AFP), an oncofetal glycoprotein with elevated levels in patients affected by cirrhosis and HCC. e sensitivity and speci�city of AFP range between 60-80% and 70-90%, respectively [96]. For this reason, AFP test utility for screening procedures is questionable. Many studies have been focused on characterizing differentially expressed proteins in sera from patients affected by liver cancer or predisposing diseases and, similarly to ovarian cancer, proteomic pro�les and proteins have been identi�ed. International Journal of Proteomics markers panel (7486, 12843, 44293, 53598 Da) identi�ed HCC with a sensitivity of 100% and a speci�city of 85% in a two-way comparison of HCV-cirrhosis versus HCV-HCC training set. Sensitivity and speci�city for the correct identi�cation of HCC were 68% and 80% for random test samples. Fibrosis patients were distinguished from cirrhotic using a �ve marker panel (2873, 6646, 7775, 10525, 67867 Da, sensitivity and speci�city 100% and 85%, 80% and 67% in the training and random test samples, resp.). Aer puri�cation, MS/MS analysis and immunoassay validation, the 6646 Da protein was identi�ed as apolipoprotein C-I and described as a marker to differentiate liver �brosis from cirrhosis [97]. SELDI-TOF protein-chip technology was applied to analyze sera from patients affected by HCV-associated chronic liver diseases with (64 samples) or without (77 samples) HCC. Samples were randomly split into two analysis groups. Six selected protein peaks ( 3444, 3890, 4067, 4435, 4470, 7770) gave information to perform early diagnosis and to distinguish HCC from chronic liver disease in the absence of HCC (sensitivity and speci�city 83% and 76%). e model was also applied to the analysis of sera from 5 subjects HCCfree and from 7 HCC patients collected before the diagnosis by ultrasonography. e markers allowed to correctly predict the presence of HCC in 6/7 patients [98]. Wu et al. [99] identi-�ed serum proteins and peptide pro�les to differentiate HBVrelated HCC and HBV-related cirrhosis. Forty-�ve protein peaks distinguished HCC from LC (liver cirrhosis) samples. e most signi�cant SELDI 3892 peak showed sensitivity and speci�city of 69% and 83% and was identi�ed also in six AFP-negative patients. e 3892 peak was considered as a complementary diagnostic marker or a potential marker for positive or negative -fetoprotein HCC. SELDI-TOF analysis of sera from 120 patients affected by HCC and 120 affected by cirrhosis showed �ve proteomic peaks ( 3324, 3994, 4665, 4795, and 5152) able to achieve, especially for early stage HCC, a diagnostic value better than serum AFP (83% sensitivity and 92% speci�city in the test set) [100]. Cui et al. [101] formulated classi�cation trees, based on SELDI serum protein pro�les, able to distinguish patients affected by chronic hepatitis B, cirrhosis, and HCC from healthy individuals. Samples were divided into training and testing groups, each composed by HBV, liver cirrhosis, HCC patients matched with normal controls. Decision trees distinguished HCC with 90% sensitivity and 89% speci�city, cirrhosis with 100% sensitivity and 86% speci�city, and HBV patients with 85% sensitivity and 84% speci�city.

Glycoproteomics in Liver
Cancer. Goldman et al. [102] used a glycomic approach to evaluate the abundance of 83 N-glycans in a total of 202 sera from 73 HCC patients, 52 with chronic liver disease and 77 controls. Glycans were enzymatically obtained from serum and permethylated before MALDI-TOF analysis. e abundance of 57 N-glycans resulted signi�cantly altered in HCC samples. Six glycans were used to differentiate HCC cases from controls and showed sensitivity and speci�city of 73-90% and 36-91%, respectively. A combination of three N-glycans ( 2472.9, 3241.9, 4052.2) was able to classify HCC with 90% and 89% sensitivity and speci�city in an independent validation set of patients with chronic liver disease. Two-hundred and three serum samples collected from 73 HCC cases, 52 chronic liver disease, and 78 healthy subjects were treated for N-glycans releasing and then analyzed by MALDI. Seven glycan peaks achieved good performance in distinguishing HCC from chronic liver disease patients and normal individuals [103]. Aer depletion of high abundance proteins, Liu et al. [55] analyzed 27 sera from early HCC in comparison to 27 cirrhosis patients in order to identify glycoprotein biomarkers. A lectin array of 16 selected lectins was used to de�ne glycan structures showing changes between the two groups of samples. Samples were then analyzed by exactag labeling, lectin extraction and LC-MS/MS. Complement C3, ceruloplasmin, histidine rich glycoprotein (HRG), CD14 and hepatocyte growth factor (HGF), as validated by western blot, were considered putative biomarkers in differentiating early HCC from cirrhosis with a sensitivity of 72% and a speci�city of 79%.

Liver Cancer �iomar�er Proteins �nvolve� in �n�am� matory Processes.
Paradis et al. [56] studied by SELDI-TOF eighty-two sera from patients with cirrhosis, either without (38 samples) or with (44 samples) HCC. irty protein peaks signi�cantly differentiated cirrhotic patients affected by HCC from those unaffected. An algorithm showing the six highest scoring peaks allowed the correct classi�cation of patients with or without HCC in 92% of individuals in the test set and in 90% in the validation set. Aer sera fractionation (IMAC-Zn spin column), analysis on NP20 chip array and protein recovery from tricine SDS-PAGE, tandem MS was performed and the highest discriminating peak (8.9 kDa) was described as the C-terminal part of the V10 fragment of vitronectin, a protein involved in cell adhesion, humoral defense mechanism as well as cell invasion. SELDI-TOF was used to identify differentially expressed proteins in hepatocarcinoma (55 samples in total, 31 HBV-related and 24 HCV-related) and chronic hepatitis patients (18 HBV and 30 HCV). Aer serum fractionation by anionic exchange chromatography, the proteins were characterized by 2-DE separation and LC-MS/MS analysis. e protein complement C3a (about 8.9 kDa), elevated both in chronic HCV and HCV-related HCC patients, was identi�ed as a candidate biomarker and further validated by PS20 chip immunoassay and western blot [57]. Yang et al. [58] used 2-DE combined to nano-HPLC-ESI-MS/MS to identify 14 proteins differentially expressed (12 up and 2 downregulated) in HCC patients with respect to normal controls. On the other hand, using whole serum trypsin-digested and then analyzed with nano-HPLC-ESI-MS/MS, twentynine proteins were identi�ed with high levels of con�dence. Six of them (Annexin VI isoform, Complement component 9, Ceruloplasmin-ferroxidase-, Serum amyloid A4, Serum amyloid A2, Serum amyloid A1 isoform 2), playing a role in immune and acute phase response or in membrane dynamics along endocytosis or exocytosis pathways (Annexin VI), were detected only in HCC patients. An 11 peak algorithm, generated by SELDI/TOF analysis, distinguished patients with HCC (41 samples) from those with hepatitis C cirrhosis (51 samples) better than the currently used biomarkers AFP, AFP L3 (Lens culinaris agglutinin-reactive AFP) and PIVKA-II (prothrombin induced by vitamin K absence-II). Within the 11-protein signature, the 13.4 kDa feature was puri�ed, identi�ed as cystatin C by MS/MS analysis and further validated by ELISA. e cystatin C, a cysteine protease inhibitor marker of in�ammation as well as renal function, resulted overexpressed in HCC samples and was described as a marker to distinguish HCC from HCV-related cirrhosis patients [59]. He et al. [60] performed by SELDI-TOF serum pro�ling on 81 patients with HBV-related HCC and 33 normal controls, randomly split into a training and a testing set. Six proteomic peaks ( 3157.33, 4177.02, 4284.79, 4300.80, 7789.87, 7984.14) were considered to construct the best classi�cation tree (sensitivity 95%, speci�city 100% in the testing set). Protein fraction corresponding to the 7489 peak was isolated and characterized by MS/MS analysis as the in�ammatory cytokine neutrophil activating peptide 2 (NAP-2). NAP-2 was validated by immunohistochemistry in HCC tissues and resulted speci�cally associated to hepatitis B-related HCC [60]. Sera from eighty-one patients with HBVrelated HCC and 80 healthy controls were divided in two sets and analyzed by SELDI-TOF. Candidate biomarkers were puri�ed and identi�ed by MS/MS and database searching. Two proteins, the thrombin light chain ( 4096) and the chemokine growth-related oncogene alpha (GRO-alpha) ( 7860) were selected as putative biomarkers. A clinical validation set composed by 48 HCC, 54 liver cirrhosis, 151 patients with other cancers and 42 healthy donors was analyzed to con�rm data by SELDI-immunoassay. e proteins, when associated to AFP, resulted in a sensitivity of 91.7% and a speci�city of 92.7% [61]. Sera from cirrhosis and HCC patients were analyzed by cleavable stable isotope labeling (cICAT) coupled to LC-ESI-MS/MS. Among 31 proteins differentially expressed, the alpha-1 acid glycoprotein (AGP), an acute phase reactant, was chosen for western blot assay and validation in a separate study. AGP was useful for discrimination of HCC from cirrhosis in patients with AFP less than 500 ng/mL [62]. A study based on 2D-gel electrophoresis and MALDI-TOF in patients with hepatocarcinoma or liver cirrhosis revealed �ve proteins differentially expressed (haptoglobin, Hp2, preprohaptoglobin, SP40 and SAA1). Western blot analysis showed haptoglobin, the most representative protein, as overexpressed in HCC patients. When used in association to AFP, the molecule improved the diagnostic accuracy. Serum haptoglobin also showed diagnostic potential in AFP-negative patients [63]. Several peptides in the serum low molecular weight fraction were identi�ed by MALDI and then characterized by LC-MS/MS. Differentially expressed peptides were described as truncations of N terminus of complement C3f, a �brinopeptide, complement C4alpha peptides, a zyxin peptide, a coagulation factor XIII peptide, and a biliverdin diglucuronide [64].

Liver Cancer Biomarker Proteins Involved in Other Functions.
Feng et al. [65] used a strategy based on sonication, albumin and immunoglobulin depletion, 2-DE and MALDI-TOF MS/MS to analyze 20 sera, respectively, from HCC, hepatitis B (HBV) patients and normal subjects. e same number of additional sera from corresponding groups was used for the validation test. Height proteins, involved in in�ammatory processes or classi�ed as acute phase reactants (alpha-1 antitrypsin, clusterin, ceruloplasmin, haptoglobin alpha2 chain, transferrin, and transthyretin) as well as alfa-fetoprotein and the heat-shock protein 27, a stressinducible protein acting in thermotolerance, cell proliferation, and apoptosis, were differentially expressed in the above-mentioned groups. Validation by western blot analysis revealed HSP27 expressed in 90% of HCC, in 10% of HBV and in none of normal sera. Wu et al. [66] compared by 2-DE and mass spectrometry sera from HCC patients and normal controls. Eight protein spots differentially expressed were analyzed and four proteins were identi�ed as putative biomarkers (MYH2 protein, mitochondrial ATP synthase, sulphated glycoprotein-2-clusterin SGP-2-, and glial �brillary acidic protein (GFAP). SGP-2, known to be involved in in�ammation and in the regulation of cellular proliferation, was also con�rmed by immunoblotting in an independent set of samples.

Serum Biomarkers in Lung Cancer.
Lung cancer is one of the leading causes of cancer-related mortality worldwide and is responsible for 1.3 million deaths worldwide annually [104,105]. e poor prognosis is evidenced by the 5-year survival rate which is less than 15% [106]. Lung cancers are grouped into small-cell lung cancer (SCLC) and nonsmall-cell lung cancer (NSCLC), consisting of adenocarcinomas, squamous cell carcinomas, and large-cell carcinomas [107,108]. NSCLCs comprise approximately 80% of all lung cancers [50], with adenocarcinomas and squamous cell lung cancers each accounting for approximately 30%. Many serologic biomarkers of lung cancer have emerged recently: these include carcinoembryonic antigen (CEA), the cytokeratin 19 fragment CYFRA21-1, cancer antigen CA-125 [109], plasma kallikrein [110], progastrin-releasing peptide (ProGRP), and neuron-speci�c enolase (NSE) [111].

Proteomic Patterns for Lung Cancer Detection.
In a study on 208 sera (158 lung cancer, 50 healthy controls), Yang et al. [112] identi�ed a 5 proteins peak pattern (11493, 6429, 8245, 5335, 2538 Da) which, in a blind test, achieved sensitivity of 86.9% (79% for stage I/II lung cancers), speci-�city of 80% and positive predictive value of 92.4%. In particular, the pattern sensitivity was 91.4% in the detection of NSCLC. In a group of sera (54 SCLC, 24 NSCLC, 32 pneumonia patients, and 40 healthy subjects), SELDI-TOF spectra data analyzed by support vector machine (SVM) gave three patterns able to distinguish SCLC from pneumonia, NSCLC patients and from healthy individuals better than neuron speci�c enolase (NSE). e sensitivity and speci�city ranged from 88% to 83% and from 91% to 75%, respectively [113]. A 17 MS protein signature was identi�ed in a study on 139 lung cancer patients (stage III-IV), 158 healthy individuals and then validated in two sets of 126 (63 lung cancer stage III-IV, 63 controls) and 50 (25 lung cancer stage I-II, 25 controls) individuals. e signature distinguished lung cancer patients from normal subjects and showed sensitivity and speci�city of 87.3% and 81.9% in the �rst validation set, and 90% and 67% in the second one [114]. Du et al. [115] 8 International Journal of Proteomics captured and concentrated serum peptides by using magnetic beads-based weak cation exchange on the ClinProt robotic platform. e peptides were analyzed by MALDI-TOF. A 5 protein �ngerprint distinguished SCLC patients (30 samples) from healthy individuals (44 samples) with a speci�city of 97% and a sensitivity of 90%. In particular, 89% of stage I/II SCLC were correctly diagnosed.

Glycoproteomics in Lung
Cancer. Glycoproteomic approaches have been applied to identify biomarkers for NSCLC early diagnosis. Aer immunoaffinity depletion of highly abundant serum proteins, glycoproteins were captured and enriched by hydrazide chemistry, recovered and then analyzed by LC-MS/MS. irty-eight glycopeptides from 22 proteins distinguished cases from controls. ree of these proteins (alpha-1-antichymotrypsin ACT, insulinlike growth factor-binding protein 3 IGFBP3, lipocalin-type prostaglandin D synthase L-PGDS) were veri�ed by ELISA, showing correlation with MS results [116].

Lung Cancer Biomarker Proteins Involved in In�am� matory Processes.
Howard et al. [71] identi�ed by MALDI-TOF a peak ( 11702) differentially expressed in lung cancer patients (24 subjects) with respect to individuals with no evidence of cancer (17 subjects). Aer puri�cation and peptide mapping, the peak was described as the acute phase reactant serum amyloid protein A and further validated by ELISA. Bharti et al. [77] analyzed by MALDI-TOF differentially expressed albumin-depleted serum proteins from SCLC patients and controls, recovered from a silver-stained SDS-PAGE. e peptides were characterized by sequencing. Haptoglobin -subunit, validated by immunoblot, was considered as a biomarker with its level correlating with the disease stage. In addition, they analyzed by ELISA the levels of hepatocyte growth factor (HGF), a multifunctional protein which regulates both cell growth and motility, and described it as a potential SCLC biomarker. A study performed on 218 sera from 175 lung cancer patients and 43 controls by SELDI-TOF showed an 11.6 kDa protein peak signi�cantly elevated in cancer sera and increased in association to the clinical stage. Serum amyloid A protein was identi�ed by tricine SDS-PAGE and MALDI-MS/MS analysis as a biomarker to discriminate lung cancer patients from healthy individuals. e marker was validated by immunoprecipitation and ELISA in the same samples and showed sensitivity of 84% and speci�city of 80% [72]. In a study on 227 sera (146 lung cancer, 41 benign lung disease, 40 normal subjects) three peaks differentially expressed ( 13780, 13900, 14070) were identi�ed by SELDI. e peaks corresponded to native transthyretin (negative acute-phase reactant) and its two variant, as demonstrated by SDS-PAGE and ESI-MS/MS analysis and further validated by immunoprecipitation and immunoblotting. e transthyretin expression was signi�cantly lower in lung cancer sera compared with sera from normal individuals, but higher compared with those obtained from benign lung disease. Subsequent ELISA assay indicated that the levels of transthyretin were consistent with those obtained by SELDI, showing approximately 65-75% sensitivity and speci�city [78]. e diagnostic accuracy of MALDI in analyzing unfractionated serum was assessed in a study by Yildiz et al. [117] performed on 142 sera form lung cancer patients matched with 146 samples from normal controls. Samples were split into training and test set. A serum proteomic signature of seven features achieved an overall accuracy of 78% and 72.6% in the training and blinded test set, respectively. e peptides around 11500 Da were further analyzed by using SDS-PAGE separation and LC-MS/MS and described as a cluster of truncated forms of serum amyloid A protein. In a study on 154 sera from pretreated patients (55% early, 45% advanced-stage) an isoform of serum amyloid A, corresponding to an 11.6 kDa SELDI peak and characterized by SDS-PAGE and tandem MS, was found to be elevated in patients with poor prognosis. In this study, sera were prefractionated in six protein fractions on the basis of their isoelectric points [73]. Forty-nine proteins were found to be differentially expressed by LC-ESI-MS/MS in pools of sera from nonsmall cell lung cancer (adenocarcinoma and squamous cell carcinoma) with respect to healthy controls. Multiple reaction monitoring (MRM) assay was used to con�rm the abundance of four selected proteins (serum amyloid A-SSA-, alpha-1 acid glycoproteins 1 and 2-AAG 1 and 2-, clusterin-CLU-). SSA and AAG 1 and 2 showed higher spectral count in lung cancer serum pool [74]. Analysis by SELDI-TOF of 227 sera showed 5 peaks ( 11530, 11700, 13780, 13900, 14070) identifying native serum amyloid A protein and transthyretin, and some of their variants as lung cancer biomarkers [75]. Serum amyloid A1 and A2 proteins were identi�ed by LC-ESI-MS/MS in lung cancer pooled sera aer SDS-PAGE fractionation. e levels were higher in lung cancer patients with respect both to patients affected by other pulmonary diseases or different cancers and to healthy controls. e results were con�rmed by ELISA. Moreover, SSA expression in lung cancer samples was detected by tissue-microarray analysis [76].

Lung Cancer Biomarker Proteins Involved in Other
Functions. Ueda et al. [79] described a method based on enrichment of the peptidomic fraction and analysis by nano-LC-MS/MS. Aer further characterization by MRM-based relative quanti�cation, peptides from apolipoprotein A4 (APOA4), �brinogen alpha chain (FIBA), and limbin (LBN), a positive regulator of the hedgehog signaling pathway, have been identi�ed as useful biomarkers for early detection and staging of lung tumors.

Serum Biomarkers in Pancreatic
Cancer. Pancreatic cancer is the fourth leading cause of cancer death in both men and women. e high mortality associated with it can be essentially attributed to advanced stage of disease at patient presentation. Few patients with pancreatic cancer are cured without surgical resection. e overall 5-year survival is about 5%, and only 20% of patients are candidates for surgical resection and possible treatment. For this small percentage of patients undergoing resection, even when followed by multimodal therapy, 5-year survival rates are still less than 25% [118][119][120]. Current methods for diagnosing pancreatic cancer are relatively ineffective to identify small potentially curable lesions. e marker recommended in clinical practice is serum CA-19-9. However, this marker is of little utility in establishing early diagnosis [121]. To date, there are no efficient modalities to early detect pancreatic cancer and strategies to improve survival have focused just on chemotherapy in the neoadjuvant setting or aer resection.

Proteomic Patterns for Pancreatic Cancer Detection.
In a study on 15 healthy controls, 24 cancer and 11 chronic pancreatitis patients prospectively collected, the low molecular serum fraction was enriched and analyzed by MALDI-TOF. An eight peaks serum signature ( 4470, 4792, 8668, 8704, 8838, 9194, 9713, 15958) differentiated cancer patients from normal individuals (sensitivity and speci�city of 88% and 93%), cancer from pancreatitis patients (sensitivity and speci�city of 88% and 30%), and cancer from healthy plus pancreatitis-affected individuals (sensitivity and speci�city of 88% and 66%). e most signi�cant peak, 9713, was described by MS/MS analysis as the apolipoprotein CIII [82]. Liu et al. [122] used SELDI-TOF technology to differentiate cancer from different pancreatic conditions, by studying 118 serum samples, split in training and test set. Two MS patterns, differentiating pancreatic adenocarcinoma from healthy controls and chronic pancreatitis, yielded in the test set sensitivity and speci�city of 91.6% (cancer versus controls) and sensitivity of 90.9%, speci�city of 80% (cancer versus chronic pancreatitis).

Pancreatic Cancer Biomarker Proteins Involved in In�ammator�
Processes. An orthotopic nude mouse model of human pancreatic cancer was used to detect serum biomarkers [83]. Mice were injected with a human pancreatic cancer cell line and then were divided in groups treated with anti-cancer drugs for some weeks. Sera were recovered and analyzed by SELDI Proteinchip technology. Plasma from 135 pancreatic cancer patients and 113 healthy individuals were at the same time examined. An 11.7 kDa protein peak, correlating with tumor weight, was detected in mice sera. Aer puri�cation and separation by SDS-PA�E, the corresponding protein was identi�ed as serum amyloid protein A and con�rmed by western blotting. e level of SAA detected in plasma of pancreatic cancer patients correlated with the clinical stage. Ninety-six sera from pancreatic cancer patients undergoing surgery were fractionated by chromatography and analyzed by SELDI in comparison with as many sera from healthy controls. Twenty-four differentially expressed peaks were identi�ed. Twenty-one of them resulted in downregulated pancreatic cancer samples. Aer puri�cation, several proteins were identi�ed by peptide mapping and postsource decay-matrix-assisted laser desorption ionization-TOF-MS. Down-regulated apolipoprotein A-II, transthyretin, and apolipoprotein A-I were described as potential markers in pancreatic cancer [84]. Hanas et al. [85] studied sera from pancreatic cancer patients by gel electrophoresis in order to highlight protein bands differing quantitatively. e proteins were analyzed and characterized by ESI ion-trap tandem MS. ree high mass proteins ( -2 macroglobulin, ceruloplasmin, complement 3C) were elevated in cancer sera with respect to controls. e ESI-MS analysis revealed great heterogeneity especially in the low mass region. By statistical analysis, twenty low-mass serum peaks correlating to controls and 20 different peaks correlating to cancer sera were found. A study performed by Fiedler et al. [86] on forty sera from patients matched with forty samples from healthy controls was focused on MALDI-TOF peptidome pro�le analysis aer using magnetic beads for protein fractionation. Data were validated by using an additional 20 plus 20 sera set. Two signi�cant peaks ( 3884 and 5959) showed 86.3% sensitivity and 97.6% speci-�city for discriminating patients from controls. e 3884 peak was described and further validated by immunoassay as platelet factor 4 (PF4). PF4, used in combination with CA-19-9, signi�cantly improved sensitivity and speci�city for the identi�cation of pancreatic cancer [86]. Rong et al. [87] used immunoaffinity depletion of highly abundant proteins and 2-DE to identify 16 protein spots differentially expressed (8 iper-and 8 ipoexpressed in cancer sera). e proteins were analyzed and sequenced. Mannose-binding lectin 2 and myosin light chain kinase 2, a serine/threonine kinase, were identi�ed as potential biomarkers for the pancreatic cancer diagnosis and further validated by western blot in an independent set of sera from pancreatic cancer patients and normal controls. Low molecular weight (<60 kDa) serum proteome from a training set composed by 24 patients with pancreatic cancer and 21 controls was analyzed by HPLC-ESI-MS/MS. Among many peaks identi�ed, a peptide from CXC chemokine ligand 7 (CXCL7) was signi�cantly reduced in cancer sera. Data were con�rmed by high-density protein microarray in a large cohort of 140 patients affected by pancreatic cancer, 10 patients with chronic pancreatitis and 87 healthy controls. Combination of CXCL7 and CA-19-9, improved the discriminatory power for pancreatic cancer [88]. In order to limit the complexity of the plasma proteome, Pan et al. [89] employed multidimensional fractionation followed by HPLC-MS/MS. Many proteins/peptides were iden-ti�ed with this method. A group of differentially expressed proteins was selected and evaluated on a separate cohort of samples from pancreatic cancer, chronic pancreatitis patients, and nonpancreatic disease control. A composite marker of the tissue inhibitor of metalloproteinases TIMP1 and the adhesion molecule ICAM1, as characterized by ELISA, showed signi�cant better performance than CA-19-9 in distinguishing pancreatic cancer, pancreatitis, nonpancreatic diseases and healthy controls. Forty-�ve samples from patients affected by pancreatic cancer and 20 from healthy controls were analyzed by 2-DE and LC-MS/MS. Seven protein spots were differentially expressed. Serum isoforms of alpha-1-antitrypsin (AAT), also con�rmed by western blot, were described as upregulated and potential serum biomarkers for pancreatic cancer [90].

Pancreatic Cancer Protein Biomarkers Involved in
Other Functions. Bloomston et al. [91] analyzed by highresolution 2-DE thirty preoperative sera from pancreatic cancer and thirty-two from healthy individuals. Differentially expressed spots were recovered and analyzed by MALDI-TOF and LC-MS/MS. Approximately 150 proteins resulted commonly overexpressed in all cancer patients. Four proteins discriminated 100% of pancreatic cancer and 94% of normal samples. Among them, �brinogen-was identi�ed as putative biomarker and further validated by enzymatic analysis in sera and immunohistochemistry in tumor tissues. One hundred and twenty-six sera form pancreatic cancer patients (84 with diabetes) were examined by SELDI-TOF in comparison to 61 sera from chronic pancreatitis (32 with diabetes), 24 from type 2 diabetes mellitus patients, and 12 from healthy controls. Classi�cation algorithms obtained by MS analysis resulted to improve the diagnostic accuracy of CA-19-9 in pancreatic cancer diagnosis and to facilitate the differential diagnosis between pancreatic cancer and type 2 diabetes mellitus. Among the large number of peptides, that described with the 3519 was identi�ed as a member of the EGF-like family [123]. Fiy-eight sera from patients with pancreatic cancer were compared with 18 samples from patients affected by benign disease and 51 healthy controls. Sera were analyzed using a strong anionic exchange chromatography proteinchip and SELDI-TOF. Sixty-one protein peaks were detected to construct multiple classi�cation trees to distinguish the disease groups, reaching 83% sensitivity and almost 100% speci�city in discriminating cancer from controls and benign disease. Putative protein biomarkers were identi�ed� one ( 4016) showed a downregulated trend in preoperative versus post-operative sera, three ( 4155, 4791, 28068) were detected in the differential diagnosis of the 3 test groups. C14orf166, a protein involved in modulation of mRNA transcription by Polymerase II, was identi�ed as corresponding to the 28068 peak by ProteinChip immunoassay. e molecule showed levels signi�cantly higher in pancreatic cancer patients, as con�rmed by immunoenzymatic methods. C14orf166 was also iper-expressed in tumor cells [92]. A SELDI-TOF protein panel derived from the study of a training set composed by 38 pancreatic cancer sera, 54 disease controls, and 68 healthy volunteers was further validated on a �rst validation set (40 pancreatic cancer, 21 disease controls, 19 healthy volunteers) and then, by ELISA, on a second one (33 pancreatic cancer, 28 disease controls, 18 healthy volunteers). Some proteins corresponding to peaks of interest were puri�ed and identi�ed. A simpli�ed diagnostic panel comprising CA-19-9, apolipoprotein C-I, apolipoprotein A-II, additionally validated by ELISA on the second validation set, resulted to improve the diagnostic ability of CA-19-9 [124]. Potential prognostic markers were initially identi�ed by nano-LC-MS/MS in 4 groups of sera, each from 10 patients, selected on the basis of survival (long or short) and therapy (gemcitabine plus bevacizumab, or gemcitabine plus placebo). Alpha1-antichymotrypsin (AACT) was negatively correlated with overall survival and considered as a prognostic marker for pancreatic cancer [93].

Conclusions
Cancer is one of the leading causes of death worldwide. Advances in screening methods signi�cantly improved early diagnosis with consequent enhancement of prognosis, survival and treatment efficacy. Unfortunately, some tumors are difficult to diagnose before the disease is in advanced or metastasizing state. erefore, there is an urgent need to discover novel biomarkers which provide sensitive and speci�c disease detection. Over the past decade, serum biomarkers have been identi�ed in sera from cancer patients by using powerful high-throughput technologies. Mass spectrometry allowed the identi�cation of hundreds of proteins within complex biological samples such as tissues, serum, plasma, and urine. MS analytical attributes in biomarker discovery are its high mass accuracy, resolution and ability to characterize the peptides at the level of their aminoacidic sequence. Several work�ows including methods for serum samples preparation (e.g., high abundance protein removal, serum fractionation), SDS-PAGE and 2D-GE, LC, different MS platforms, and protein chip arrays have been used for biomarker discovery. Many differential MS peak pro�les were identi�ed and several proteins were characterized and described as potential biomarkers for high mortality tumors (Tables 1-4), achieving different levels of sensitivity and speci�city to diagnose the disease. Many of these proteins are involved in fundamental processes such as in�ammation, cellular differentiation and proliferation, and apoptosis. Among them, positive (i.e., serum amyloid A, ceruloplasmin, complement factors, haptoglobin) and negative acute-phase reactants (i.e., transthyretin, transferrin) were differentially expressed in sera from ovary, lung, liver, and pancreatic cancer patients. Some putative ovarian cancer biomarkers described in this paper, such as the keratin 2a, the glycosyltransferase-like 1B, involved in glycosylation processes, and the casein kinase alpha 1, a serine/threonine kinase involved in cellular differentiation, proliferation and apoptosis, have been associated with processes related to cancer [125][126][127][128]. Similarly, the mannose binding lectin 2 (MBL2), a mediator of in�ammation which results iperexpressed in pancreatic cancer sera, is involved in cancer processes. Genetic alterations of MBL2 can increase colon cancer susceptibility in African Americans and a MBL genetic polymorphism, associated to a reduction of vaginal MBL concentration, may be a risk for development of ovarian cancer [129,130]. e chemokine CCL18, here described as candidate ovarian cancer serum biomarker, was also considered as a urine biomarker for bladder cancer detection [131]. Likewise, the HCC biomarker cystatin C, an inhibitor of cystein proteinases, showed signi�cantly higher levels also in sera from lung cancer patients [132], and the HCC and lung cancer putative biomarker alpha-1 acid glycoprotein 1, an acute phase protein, was found as well elevated in sera and tumor tissues from patients affected by gastric carcinoma [133]. e here described liver cancer biomarker heat shock protein 27, a protein with cytoprotective and anti-apoptotic activity, measured by immunoenzymatic assay, was con�rmed to be elevated in an independent cohort of sera from HCC patients [134].
Despite the great advances in the application of MS in serum biomarker discovery, several challenges remain. e identi�cation of differential serum protein pro�les and speci�c molecules able to discriminate normal from diseased subjects requires a technology able to highlight small differences and to process large series of serum samples. Although MS is the most powerful approach for biomarker identi�cation, there are some boundaries in the analysis of serum. ese can be attributable to the complex nature of serum and its tremendous dynamic range, to diurnal variation in protein expression, instability of proteins due to in vivo or ex vivo protease activity, pre-analytical methods reproducibility as well as to the intrinsic MS sensitivity (> g/mL) [135] in detecting analytes which usually range between 50 pg/mL and 10 ng/mL [136]. Accurate selection of cases and controls, standardization of sample collection and storage conditions, utilization of adequate and effective methods focused on reducing the complexity of serum/plasma prior MS analysis, use of different protein array with complementary binding conditions, re�ned bioinformatic and statistical analysis to process data, and suitable validation work�ows by immunoassay on larger sets of independent samples are necessary elements to circumvent criticisms and improve the biomarker discovery process.