Convolutional model for predicting SNP interactions

Abstract

Single-nucleotide polymorphisms (SNPs) are genetic markers that empower researchers to examine for genes associated with complex diseases. Several efforts have been contributed by researchers to study the interaction effects between multi-locus SNPs for discerning the status of complex diseases. However, the current conventional machine learning techniques are still left with several caveats. Deep learning is a new breed of machine learning technique that elucidates the hidden structure of the raw data by transforming it into multiple high levels of abstractions, using the power of parallel and distributed computing. It promises empirical success in the number of applications including bioinformatics to drive insights of biological complexities. The deep learning approach in the multi-locus interaction studies is yet to meet its potential achievements. In this paper, a convolutional neural network is trained to identify true causative two-locus SNP interactions. The performance of the method is evaluated on hypertension data. Highly ranked two-locus SNP interactions are identified for the manifestation of hypertension.