Abstract 1640: Unidentified genetic variants influence pancreatic cancer risk: An analysis of the PanScan genome-wide association study

Abstract

Abstract Genome-wide association (GWA) methods have been used to identify multiple pancreatic cancer susceptibility loci. However, it is likely that other unidentified common susceptibility variants exist. In this work we used data on 493,619 common autosomal single nucleotide polymorphisms (SNPs) from the PanScan GWA study (2,857 cases and 2,967 controls) to confirm the existence of unidentified susceptibility variants using two cross-validation methods: a polygenic risk score (PRS) approach and a power-replication approach. For the PRS approach, after excluding all SNPs from known pancreatic cancer susceptibility loci, we first constructed a polygenic risk models based on a training dataset and then tested the model in an independent testing dataset. The second approach (i.e., “power-replication) was based on calculating power to detect association for each SNP based on association estimates from a training set and then testing the power estimate for association with probability of replication in an independent testing data set. Using the polygenic risk score method, we demonstrate the existence of additional pancreatic cancer risk variants, the majority of which have P-values > 0.01 in the PanScan study. We also demonstrate enrichment for unidentified loci near gene regions. We reach similar conclusions using the power-replications approach, where SNPs with higher power to detect association were more likely to be replicated in an independent dataset at P<0.05, especially when analyses were restricted to SNPs in gene regions. While our results do not substantially improve pancreatic cancer prediction, they do provide further rationale for focusing association studies on genic regions and suggest that conducting larger GWA studies and innovative secondary analyses have the potential to reveal novel susceptibility loci. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 1640. doi:1538-7445.AM2012-1640