Protein Quantitation through Targeted Mass Spectrometry: The Way Out of Biomarker Purgatory?

The enormous potential of biomarkers to revolutionize clinical practice and improve patient care has been well documented (1, 2 ). Molecular-based diagnostic and prognostic tests, particularly those aimed at protein analytes, could be used to detect disease earlier, enabling treatment to start sooner and possibly cure rather than to merely delay further injury or death. These tools could also be used to stage disease more accurately and to predict response to therapy, thereby helping to select the correct treatment. Biomarkers can also be used to stratify patients for the assessment of new drug therapies and to serve as surrogate endpoints in early-phase drug trials, thereby lowering the overall cost of drug development and producing more effective treatments. Given their high potential therapeutic and financial impacts, that so few new protein biomarkers have been introduced into widespread clinical use recently is, on the surface, surprising. In fact, only 5 new protein markers have been approved by the US Food and Drug Administration in the last 5 years for measurement in plasma or serum [the information on protein markers in Anderson and Anderson, 2002, has been updated with information from the Center for Devices and Radiological Health, US Food and Drug Administration] (3, 4 ). The reasons for the dearth of new protein biomarkers are gradually becoming clearer. They are related to the high false-discovery rate of “omics” methods (regardless of the technology used), combined with a lack of robust methods for biomarker verification in large clinical sample sets (5– 8 ). It is now common for differential analyses of tissue or plasma samples by multidimensional liquid chromatography–tandem mass spectrometry (LC-MS/MS) (the workhorse tool for unbiased discovery) to confidently identify thousands of proteins, hundreds of which can vary in concentration by 5-fold or more between case and control samples in small discovery studies. To access proteins at lower abundances (e.g., 500 g/L in plasma, concentrations at which occur many of the known protein biomarkers, such as carcinoembryonic antigen, prostate-specific antigen, neuron-specific enolase, and the troponins), these studies always employ multidimensional fractionation at the protein and/or peptide level, thereby exploding a single patient sample into up to 100 subfractions, each requiring lengthy LC-MS/MS analysis. It is not uncommon for the analysis of a single case/control sample pair to take up to 2 weeks of oninstrument time, which limits the numbers of samples that can be practically analyzed to typically 10 (or fewer) case– control comparisons. These numbers are very small relative to the high dimensionality of the proteome (hundreds of thousands or more possible components, when posttranslational modifications and other variants are taken into account) and the scale of typical variation in the human population. Thus, a very large fraction, possibly exceeding 95%, of the protein biomarkers “discovered” in these experiments are false positives that arise from biological or technical variability. Clearly, discovery “omics” experiments do not lead to biomarkers of immediate clinical utility, but rather produce “candidates” that must be “qualified” and “verified” (7, 8 ). Until recently, verification technologies capable of testing large numbers of protein biomarker candidates emerging from discovery “omics” experiments in large sample sets ( 1000 –2000) have not been available. In principle, antibody (Ab)-based measurements could be used; however, the required immunoassay-grade Ab pairs exist for only a small number of the potential candidate biomarker proteins. Developing a new, clinically deployable immunoassay is both very expensive (US $100 000 to $250 000 per biomarker candidate for a research assay, or $2– 4 million for a Food and Drug Administration–approvable assay) and time-consuming (1–1.5 years). This fact restricts the use of immunoassays to the short list of already highly credentialed candidates. For the large majority of new, unproven candidate biomarkers, what is required is an intermediate verification technology with shorter assay-development time lines, lower assay costs, effective multiplexing of 10 –50 candidates, low sample consumption, and a high-throughput capability for analyzing hundreds to thousands of serum or plasma samples with good precision. The goal of such a verification approach would be to identify from the initial list of hundreds of candidate protein biomarkers the few that are worth advancing to traditional candidate-validation 1 Nonstandard abbreviations: LC-MS/MS, liquid chromatography–tandem mass spectrometry; Ab, antibody; SID-MRM-MS, stable-isotope–dilution multiple reaction–monitoring mass spectrometry; LOQ, limit of quantification; SISCAPA, stable isotope standards with capture by antipeptide antibodies. Clinical Chemistry 54:11 1749–1752 (2008) Editorial