Abstract 1593: Genetic interactions in lung cancer using machine-learning approaches in genome-wide association studies

Abstract

Abstract Genome-wide association studies (GWAS) consisting of hundreds of thousands or millions of single nucleotide polymorphisms (SNPs) have presented the complex inheritance patterns of disease/trait. Although GWAS have identified thousands of SNPS associated with diseases, few studies have explored interactions among these SNPs that may influence disease risks. Machine learning applications can identify SNPs that influence disease risk through interactions and define how these SNPs jointly influence disease risks. Tree-based machine-learning applications; classification and regression trees (CART) and random forest (RF) methods, have become increasingly popular and convenient tools for understanding interactions influencing disease development. Here we apply these methods to understand the genetic architecture of lung cancer. To elucidate the SNP-SNP interactions that influence lung cancer risk, we applied supervised tree-based approaches using 18,444 cases and 14,027 controls from lung cancer OncoArray GWAS data that is the largest lung cancer GWAS so far. To reduce the space of SNPs to consider in modeling, we first selected the SNPs very significantly (p<0.00001) associated with lung cancer risk. Random Forests, which consists of systematically fitting classification trees, was run 1,000 times to identify the most influential SNPs that jointly influence lung cancer risk. Subsequently we applied a classification tree approach to summarize interactions that predict risk. The final parsimonious tree included effects from genetic variants in rs55781567(CHRNA5): rs452384(CLPTM1L): rs9258608(LOC105375010): rs6154144(DNAJC5): rs116506680(HLA-G): rs9271365(near HLADQA1): rs421629(CLPTM1L). The nodes of this tree had Odds ratios for lung cancer ranging from 0.87 to 1.63. Machine-learning approaches in genomics can provide an important genetic mechanism in lung cancer development. Citation Format: Jinyoung Byun, Younghun Han, Christopher I. Amos. Genetic interactions in lung cancer using machine-learning approaches in genome-wide association studies [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1593.