Enhancing the analysis of mass-spectrometry proteomic profiles using prior knowledge and past data repositories

Whole-sample mass-spectrometry proteomic profiling based on SELDI-TOF-MS technology has lead to many promising results in detection of various types of cancer and other diseases. However, the majority of SELDI-TOF-MS disease studies performed to this day do not attempt to identify protein species responsible for these promising results. The limitation of the protein identification is that it requires secondary lab-based analysis which increases the cost of the study, and that at the end, the identifications may not lead to any new biologically important result. To address the problem, our work focuses on computational approaches that provide early insights on the identity of putative biomarker signals found with a SELDI-TOF-MS instrument. We present two computational methods to achieve this goal: (1) labeling of mass-spectra peaks corresponding to high-abundance protein species, and (2) evaluation of disease-specific signals with the help of profiles from past case/control studies. The key benefit of our methods is that they can provide early characterization of discriminative signals while working directly with whole-sample profiles. As a result, they can be used to filter-out some of the MS signal candidates so that subsequent identification procedures are directed towards analysis of signal species that are likely to yield new information.