nPARS: A Comprehensive Search Algorithm for Constructing Bayesian Networks Using Large-Scale Genomic Data

• Activity Number: 472
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Yen-Yi Ho*+ and Leslie Cope and Giovanni Parmigiani
• Companies: The Johns Hopkins University and The Johns Hopkins University and Harvard University/Dana-Farber Cancer Institute
• Keywords: Chemotherapy resistance ; bayesian networks ; large-scale genomic data ; search algorithms ; eQTL study

Abstract

Background: Several studies have shown that the resistance to docetaxel and 5-FU in human cancer cell are significantly inheritable, but little is known about the underlying regulatory networks for the chemoresistance responses. Conventional gene association studies have reported on genomic location that could have effects on these differential treatment responses. However, much of the etiologic pathway mechanisms still remain largely unknown based on these study results. In this paper, we consider an extension of the traditional gene association study design by integrating information about genomic DNA variations and gene expression abundance and infer genetic networks that are associated with cellular responses to docetaxel and 5-FU. Results: We propose a comprehensive search approach named network partition reassembly search (nPARS) algorithm to construct Bayesian networks using large-scale genomic data. Based on our simulation results, we suggest that the proposed nPARS algorithm achieved better performance than commonly used search algorithms in terms of computational efficacy. We also demonstrate the implementation of nPARS algorithm in an experimental data set.