Interpretive Risk Assessment on GWA Data with Sparse Linear Regression

Genome-wide association (GWA) studies provide large amounts of high-dimensional data. GWA studies aim to identify variables, i.e., single nucleotide polymorphisms (SNP) that increase the risk for a given phenotype and have been successful in identifying susceptibility loci for several complex diseases. A remaining challenge is however to predict the individual risk based on the genetic pattern. Counting the number of unfavorable alleles is a standard approach to estimate the risk of a disease. However this approach limits the risk prediction by only allowing for a subset of predefined SNPs. Recent studies that apply SVM-learning have been successful in improving the risk prediction for Type I and II diabetes. However, a drawback of the SVM is the poor interpretability of the classifier. The aim is thus to classify based on only a small number of SNPs in order to also allow for a genetic interpretability of the resulting classifier. In this work we propose an algorithm that can do exactly this. We use an approximation method for sparse linear regression problems that has been recently proposed and can be applied to large data sets in order to search for the best sparse risk predicting pattern among the complete set of SNPs.