Proteomic approaches to the identification of novel biomarkers for cancer diagnosis and staging have traditionally relied on the identification of differentially expressed proteins between tumor cells and their normal counterparts based on the patterns of protein expression observed by two-dimensional gel electrophoresis (2D-PAGE). Recent advances in mass spectrometry and in the informatics and statistical tools necessary to interpret mass spectrometric data have revolutionized the approach to defining new tumor markers. The combinations of SELDI mass spectrometry, retentate affinity chromatography, and statistical algorithms for pattern recognition have engendered a great deal of interest in 'proteomic profiling' as a diagnostic tool. However, the ability of new mass spectrometers to provide unambiguous identification of low abundance proteins from mixtures as complex as human serum also provides a mechanism for the discovery and mechanistic validation of small sets of specific proteins that are amenable to more traditional formats for clinical assays.