Statistical Inference of Protein Identification Using Tandem Mass Spectrometry Data

• Activity Number: 19
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Consulting
• Author(s): Susmita Datta*+
• Companies: University of Louisville
• Keywords: Mass spectrometry ; MS ; MS/MS ; Bayesian

Abstract

In this research we develop a statistical methodology for more accurate protein identification using mass spectrometry, specifically tandem mass spectrometry (MS/MS) data. Most of the research in the area of protein identification from MS/MS data involves a two-step procedure of peptide identification first followed by the second step of protein identification process. In this setup the interdependence of present/absent peptides and proteins are neglected resulting relatively inaccurate protein identification. In this proposal, we develop an integrated, single-step, nested hierarchical Bayesian model to evaluate 'which proteins are present' and 'which peptides are identified correctly' simultaneously from MS/MS spectra of cancer tissue or fluid samples. We also incorporate the interdependence of proteins in a same pathway.