On the Development of a New Framework for the Joint Analysis of Genomic and Pharmacological Data

• Activity Number: 343
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: International Chinese Statistical Association
• Author(s): Haisu Ma*+ and Ray Liu
• Companies: Graduate Program in Computational Biology and Bioinformatics, Yale University and Millennium : The Takeda Oncology Company
• Keywords: Bayesian statistics ; Drug target discovery ; Tensor Factorization

Abstract

Drug target discovery involves a variety of complex data sources. Genome wide expression data, metabolics data, and drug sensitivity profiles are examples of commonly used data. Traditional analysis methods consider each type of data one at a time, but pooling the data could reveal new information. Here we propose a new statistical method using tensor factorization and Bayes' theorem for the joint modeling of various data sources. This model enables incorporation of prior knowledge on the associations between genes, drugs and pathways through the Bayesian sparse model set-up. The model has multiple usages, including the prediction of novel drug targets. Performance of this model is evaluated via various simulations.