Shotgun proteomics is one of the most commonly used approaches to MS-based biomarker discovery, due to its high throughput and sensitivity. The general strategy involves simultaneous protease digestion of all proteins in a mixture, liquid chromatography-based separation of peptides and analysis by tandem mass spectrometry (MS/MS) to produce fragmentation spectra of each peptide. Each experimental spectrum is searched against a protein database. Sequences that best match the experimental spectra are considered identified, while a set of reliably identified peptides from the same protein is necessary for a reliable protein identification. The main goal in the proposed work is to generate and interrogate MS/MS data from several proteomics platforms, including ESI/MS, MALDI/TOF/TOF, LC-IMS/TOF and MALDI-PID/TOF to develop customized computational tools that address several challenging problems in shotgun proteomics data analysis: peptide identification, protein identification and label-free protein quantification. Our proposed approach is data-driven. At its core is the application of machine learning methods to the prediction of peptide fragmentation spectra as well as the likelihood of peptide detection in a typical proteomics experiment. Improved peptide identification coupled with the predicted peptide delectability will then be used to develop new methods for improved protein identification and quantification. The methods proposed herein will be extensively evaluated and software will be made public both as web-based tools and open-source deliverables. These software tools will enable researchers using proteomics technologies to more effectively and efficiently study a variety of health related conditions. Such studies might entail disease diagnosis (biomarker discovery), disease progression (tissue profiling), or effects of treatment (drug-induced proteome changes). These studies will enhance understanding of diseases and hasten the development of effective treatments and cures. In addition, these tools will be useful in characterizing new analytical tools for proteome analysis. Here we propose to develop and extensively evaluate computational methodology that will be used to improve the interpretation of tandem mass spectrometry data. These software tools will enable researchers using proteomics technologies to more effectively and efficiently study a variety of health related conditions. Such studies that might entail disease diagnosis, disease progression, or effects of treatment, will enhance understanding of diseases and hasten the development of effective treatments and cures.
