Eran Halperin, Ph.D.

Associate Professor of Computer Science, and Molecular Microbiology and Biotechnology, Tel Aviv University

“Finding hidden signals in whole-genome genetic and epigenetic data”

Abstract: Whole-genome genetic and epigenetic data sets the promise of detecting statistical correlations between phenotypes and genetic variants or epigenetic markers via genome-wide association studies (GWAS) and epigenome-wide association studies (EWAS). These correlations are useful for the generation of new hypotheses regarding the mechanisms involved, and they can be used for disease prediction and prediction of treatment outcomes. GWAS and EWAS studies, however, are complicated by the fact that correlations between the phenotype and confounders such as age, sex, batch effects, etc., may result in a large number of false positives. I will describe different approaches that deal with these confounders by directly predicting them from the data. Specifically, I will show how one can predict cell type composition and ancestry from either genotype or methylation data, using different variations of principal components analysis. These variations utilize the specific nature of each of the data types, resulting in a better performance than standard PCA. I will demonstrate how these approaches can be useful in specific studies of whole-genome genetic and epigenetic data.