Christine Peterson is a postdoctoral scholar at Stanford working with Chiara Sabatti. Her current research is focused on the development of statistical methods to account for multiple hypothesis testing in the context of multivariate phenotypes. Before coming to Stanford, Christine earned an undergraduate degree in applied mathematics from Harvard and a PhD in statistics from Rice University. Her doctoral research was focused on the inference of biological networks using Bayesian graphical models, including applications to the inference of cellular metabolic networks and protein networks in cancer.
Talk: Multiple testing procedures for eQTL discovery
We are developing methods to both improve power and reduce errors in identifying genetic variants that are relevant to multivariate phenotypes such as imaging features, actigraphy measures, or gene expression. Since both the predictor and response variables are high-dimensional, this represents a massive multiple testing problem. For expression quantitative trait loci (eQTL) studies, which focus on gene expression as the outcome measure, the current standard approach to identify genetic effects is to compute pairwise tests of association for each SNP to its nearby (or cis) genes, then control the false discovery rate across all such tests. The search for distant (or trans) eQTLs is typically run separately and is often underpowered. In this talk, I will discuss improvements to this procedure which allow the integration of cis and trans eQTLs through appropriately chosen weights and which take into account the problem structure by focusing on families of hypotheses based on genomic location. The proposed methods will be illustrated both through simulations and through an application to a study of traits associated with bipolar disorder.
Wednesday June 4th, 2014
1:00 PM Lunch In Clark S360
1:15 PM Seminar
Location: Clark S360