Loss, Actions, and Decisions: Subjective Inference in High-Throughput Genomics

• Activity Number: 492
• Type: Invited
• Date/Time: Wednesday, August 3, 2011 : 10:30 AM to 12:20 PM
• Sponsor: IMS
• Author(s): Chris Holmes*+
• Companies: University of Oxford
• Keywords: genomics ; Bayesian inference ; decision theory ; nonparametrics

Abstract

In applications of high-dimensional data analyses arising in genetics and genomics we often have substantive prior knowledge on plausible effect sizes and on the credible dependence structures underlying the data. Subjective Bayesian analysis that exploits this knowledge and incorporates losses on action-outcome pairs is particularly well suited here. In this talk we will overview recent developments in Bayesian methods for high-dimensional cluster and regression analysis with an emphasis on nonparametric and decision theoretic approaches in genomics. We will discuss current research in decision theoretic Bayesian inference for partial and nonparametric likelihoods. The methods are motivated and illustrated using ongoing real world studies in cancer genomics, in particular surrounding statistical methods to model the occurrence of copy-number alterations in cancer genomes, whereby stretches of DNA are systematically duplicated or deleted during tumour formation, and their association with clinical phenotypes.