Identification of Protein Phosphorylation Sites Using Zero-Inflated Poisson Regression

• Activity Number: 669
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Shu Yang* and Eric Kolaczyk+ and Simon Kasif and Martin Steffen
• Companies: Boston University and Boston University and Boston University and Boston University
• Keywords: cancer ; Phophorylation ; protein ; Identification ; zero-inflated Poisson Regression

Abstract

Differential expression analysis is a basic task in bioinformatics. Most work on this task has been done for microarray data. Recently, methods for detecting differentially phosphorylated residues are being actively developed. The data from protein phosphorylation sites can provide a direct link to the cellular processes involved in tumorigenesis, making it particularly appropriate for the creation of statistics to identify cancer from normal tissue and enable future treatment. However, these new data pose certain unique statistical challenges: they contain low-level counts and are frequently inflated with artificial zeros. Standard permutation tests can be notably conservative on these data. For better identification, we propose a framework based on zero-inflated Poisson regression. Results from both simulation and tyrosine phosphorylation data of lung cancers are given as validation.