QuasiNovo: An Information Theoretic Approach to De Novo Peptide Sequencing

University South Carolina Research Foundation, Columbia SC

Investigators

Abstract

The University of South Carolina is awarded a grant to develop tools for peptide identification. Tandem mass spectrometry has become increasingly important in high-throughput protein identification. The most popular approach to identifying peptides from mass spectra uses the experimental spectrum to query a database of known peptides. A major weakness of all database approaches is that they are unable to identify peptides that are not contained in the database. This is a particularly important limitation in the case of microbial peptides. It is well known that only 1%-10% of all microbes found in the environment can be cultured. Thus there are many bacteria that have not been previously identified. Indeed, even among culturable organisms many remain uncharacterized due to extreme diversity. Only de novo sequencing offers the possibility of identifying novel peptides. The goal of this project is to make significant improvements in the accuracy of de novo peptide sequencing. This will be accomplished through a systematic study of the use of amino acid usage models in fragmentation and cross-correlation peptide scoring functions. Preliminary results strongly support the hypothesis that a scoring function that considers amino acid usage patterns will be better able to distinguish between candidate peptides. This in turn will lead to much higher accuracy in peptide prediction. In addition, a Bayesian model for exploring the uncertainty of candidate peptides produced by de novo sequencing will be developed. A major difference between this model with that developed by other groups is that the proposed approach will incorporate the concept of proteome signature as a prior. Existing models for de novo sequencing do not expressly indicate amino acid usage, and thus implicitly assume flat priors for amino acid usage. The proposed research will be integrated into undergraduate and graduate courses taught in computer science and statistics at the University of South Carolina. The research will also be leveraged to support training and education through the development of training workshops to be presented at annual SC-INBRE Bioinformatics Core statewide meetings. The target audience of the annual workshops will be undergraduate and graduate students as well as faculty and staff in bioinformatics-related disciplines (biology, computer science, statistics, etc.) in South Carolina.