An approach using random forest methodology for disease risk prediction using imbalanced case–control data in GWAS

Abstract As single nucleotide polymorphisms (SNPs) are known to be associated with the disease, prediction of disease risk of an individual based on SNP genotyping data using start-of-art prediction techniques is an important problem in the area of genome wide association studies (GWAS). In the present investigation, an approach based on random forest (RF) methodology has been proposed for the prediction of disease risk from imbalanced case-control data. The proposed approach was compared with the existing methods meant for imbalanced data, namely, balanced random forest (BRF) and weighted random forest (WRF) based on several performance metrics. The proposed approach was illustrated using a case–control data set of Ulcerative colitis and was found to perform better in terms of prediction accuracy over the existing methods.