Fuzzy Clustering and Bayesian Model Selection for Peptide/Protein Identification

• Activity Number: 342
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Soyoung Ryu*+ and Vladimir Minin and Dave Goodlett
• Companies: University of Washington and University of Washington and University of Washington
• Keywords: Mass spectrometry ; Clustering ; Bayesian model selection ; False discovery rate ; Filtering

Abstract

Identifying proteins in biological samples is a critical step in all proteomics analyses. We propose a novel clustering-based method to accomplish this task. Tandem mass spectrometry generates from thousands to millions of spectra that can be used to identify proteins present in complex biological samples. After filtering noise spectra, we cluster similar spectra. We assess clustering uncertainty using a Fuzzy clustering approach. Then, the probabilistic model for clusters is built and all the spectra in a cluster are scored against candidate peptides using Bayesian model selection. Lastly, we assign significance scores to top-ranked peptides (and corresponding proteins) based on a decoy database assisted FDR procedure. The performance of our method is illustrated by applying our procedure to the standard protein mixture and to proteins from yeast grown under glucose conditions.