Genetic Connection to Drug-Induced Liver Injury (DILI) Through Statistical Learning Methods

Abstract

AbstractDrug-Induced Liver Injury (DILI) is one of the major causes of drug development failure or drug withdrawal from the market after development. Therefore, investigating factors associated with DILI is of paramount importance. Environmental factors that contribute to DILI have been investigated and are, by and large, known. However, recent genomic studies have indicated that genetic diversity can lead to inter-individual differences in drug response. Consequently, it has become necessary to also investigate how genetic factors contribute to the development of DILI in the presence of environmental factors. Thus, our aim is to find appropriate statistical methods to investigate gene-gene and/or gene-environment interactions that are associated with DILI. This is an initial study that only explores statistical learning methods to find gen-gene interactions (epistasis). We introduce Multifactor Dimensionality Reduction (MDR), Random Forest (plus logistic regression), and Multivariate Adaptive Regression Splines (MARS), as the few potential methodological approaches that we found. Next, we attempt to improve the MARS method by combining it with a variable selection method.