Abstract 3969: A novel statistical method for detecting association of copy number variations in genome-wide association studies of cancers

Abstract

Abstract Copy number variations (CNVs) are one of the major genetic resources in human populations. In large scale genome-wide association studies (GWAS), CNVs are usually detected from intensity data, quantified as Log R Ratio (LRR) and B-allele Frequency (BAF), generated by single nucleotide polymorphism (SNP) genotyping arrays. The standard approach for detecting CNV associations is based on two-step procedures: CNVs are first detected using intensity data and then association is tested based on the detected CNVs. Because the accuracy of CNV calling depends on the number of probes covered in the CNV region, short CNVs are generally detected with low sensitivity. Consequently, the statistical power of detecting short CNVs associated with diseases is low. We develop a novel method, variable thresholding exact statistic (VTES) implemented in a package SegCNV, for identifying genomic regions within which the frequency of CNVs is significantly higher in cases than in controls. VTES optimally summarizes the LRR and BAF of the probes to maximize the power. Extensive simulation studies demonstrated that VTES had a much better power for detecting rare, short CNVs associated with common diseases compared to standard two-step procedures, while maintaining correct type-I error rates. Particularly, the power of VTES is close to the ideal power when CNVs can be called perfectly. VTES is flexible to detect both recurrent CNVs and non-overlapping CNVs associated with diseases. VTES is computationally efficient and can analyze a GWAS of 2000 subjects within 20 hours using a single core. We applied VTES to a GWAS of testicular cancer and lung cancer, and identified regions that show strong evidences of CNV associations. We expect that VTES will prove useful for CNV association analysis in existing large scale GWAS of cancers. 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 3969. doi:1538-7445.AM2012-3969