Statistical Analysis Strategies for Shotgun Proteomics Data

• Activity Number: 281
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Ming Li*+
• Companies: Vanderbilt University School of Medicine
• Keywords: shotgun ; Quasi-Poisson ; biomarkers ; high-dimensionality

Abstract

Recent advances in proteomic technologies have brought investigators a richer understanding of and capacity for detection of protein patterns and biomarkers. However, the complexity and high-dimensionality of LS-MS/MS (shotgun) data make the quantitative analysis quite challenging. The statistical tools for such type of data are still immature. We introduce Quasi-Poisson regression method. Different than most of the existing methods for such type of shotgun data (T-test, Fisher's Exact test, G-test, and Beta binomial test), we address the advantage of modeling shotgun data in the regression framework, which can be flexibly extended to model shotgun data from more complicated experimental designs. In addition, we compare our proposed method to a newly developed Bayesian model, the QSpec method. We conclude that Quasi-Poisson method performs equally well with less computational work.