Statistical Analyses of Genetic Epistasis Among Cancer Genes

• Activity Number: 124
• Type: Topic Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Audrey Fu* and Xiaoyue Wang and Megan E. McNerney and Kevin P. White
• Companies: and Peking Union Medical College and The University of Chicago and The University of Chicago
• Keywords: genomics ; cancer ; epistasis ; regression ; survival analysis ; interaction

Abstract

Quantitative genetic epistasis (or interaction) has been hypothesized to be an important factor in the development and progression of complex diseases, especially in cancers. However, as cancer mutations are uncovered at an unprecedented rate, determining which combinations of genetic alterations interact to produce cancer phenotypes remains a challenge. Here we present a seemingly unrelated regression model for the inference of data from combinatorial RNAi screening that knocked down a single gene or two genes simultaneously. We screened 1508 pairs of 67 genes, which are frequently co-altered in primary breast cancers based on the TCGA data, and selected three most informative but correlated molecular traits as the outcome of knockdown. Our regression model inferred dense and often previously undetermined interactions among cancer genes. We further carried out survival analysis for interacting gene pairs, and discovered that they are significantly associated with survival time when co-altered in patients, indicating that genetic interaction mapping may be leveraged to improve assessment of cancer risk in individual patients.