Finding Drug Targets in Small Sample GWAS

• Activity Number: 563
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Knut M. Wittkowski*+ and Benedetta Bigio
• Companies: Rockefeller University and Rockefeller University
• Keywords: autism ; GWAS ; small samples ; epistasis ; minor allele frequency

Abstract

Genome-wide (GW) association studies (GWAS) have failed to identify drug targets due to low sensitivity and specificity. Here, we address two problems: First, the expected distribution of p-values in GWAS is not uniform, because more significant results are more likely with higher minor allele frequency (MAF). Hence, the expected distribution should be convex, if it were not for the second problem. Increasing the sample size cannot guard against systematic, though unrelated differences between non-randomized samples from outbred populations. To obtain unbiased estimate, we propose to estimate the expected curve from chromosomes with a convex distribution function. To guard against true, but unrelated effects, we use a non-parametric wide-locus approach with higher power for epistatic, rather than isolated SNPs, which would have been selected against if they would reduce reproductive fitness. When applied to autism, samples of less than 1000 subjects sufficed to consistently identify the same mechanistically related clusters of genes in two studies, suggesting ion channel modulators given at the first symptoms (~1 yr) to prevent progression to the more severe end of the spectrum.