Abstract 5232: Using an integrated gene-based sequence kernel association test (intSKAT) to identify subtype specific single nucleotide variants in glioma

Abstract

Abstract One of the challenges for genome-wide association analyses is that the effect directions and allele frequencies (e.g., rare, common, or combination of them) of true causal variants are unknown. Built on a family of powerful approaches, sequence kernel association test (SKAT), we have devised a gene-based omnibus approach, integrated-SKAT (intSKAT) to perform association tests using next-generation sequence data. This includes a suite of 12 methods: Burden test, SKAT, SKAT-O, SKAT-C (Combined sum test of rare- and common-variant effects), SKAT-A (Adaptive sum test), SKAT-AR, three methods weighted by functional scores, and three rare-variant only methods. This expansive set allows the flexibility to detect the potential combination of different allele frequencies, effect directions, and/or (lack of) functional predictions. Minimum FDR was used to adjust for multiple comparison across methods. We applied intSKAT to investigate sub-type specific susceptibility loci between low-grade glioma (LGG) and glioblastoma (GBM) as proof of principle. We downloaded germline exome sequence data (N = 612) from TCGA, and aligned the sequence reads using the Burrows-Wheeler Aligner (BWA). Insertion/deletion realignment, quality score recalibration, and variant identification were performed with the Genome Analysis ToolKit (GATK). We used 80% SNV call rate for quality control. Following Principal Component Analysis, data for 544 Caucasian subjects were included in analysis. A total of 224K SNVs in 18,053 genes were studied. Ten genes were significantly associated with glioma subtypes (minFDR of 10%). Among these genes, a total of 9 significant SNVs with predicted possible damaging functions were identified in 6 genes, ATP5O, CKAP2L, SORBS1, STK19, VPS13B, and ZCCHC4. Five of the genes are significantly differentially expressed between LGG and GBM (all p<1.02×10-3), and consistently supported by both microarray and RNAseq platforms, with the only exception of ZCCHC4. Three of the significant genes, which would not have been identified using SNV-level univariate analyses, are CALML6, FGF22, and GRIN1. Weighted-SKAT was effective to identify genes with both deleterious and protective variants. Weighted-Burden test was powerful to detect genes with deleterious variants with predicted function. These findings demonstrate the power of our proposed gene-based method. Citation Format: Yian Ann Chen, Jamie K. Teer, Zachary J. Thompson, Rebekah L. Baskin, Yonghong O. Zhang, Kate J. Fisher, Zhihua Chen, Alvaro N. Monteiro, Kathleen M. Egan. Using an integrated gene-based sequence kernel association test (intSKAT) to identify subtype specific single nucleotide variants in glioma. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 5232.