Evolving SNP Panels for Genomic Prediction

The use of genetic variation (DNA markers) has become widespread for prediction of genetic merit in animal and plant breeding and it is gaining momentum as a prognostic tool for propensity to disease in human medicine. Although conceptually straightforward, genomic prediction is a very challenging problem. Genotyping organisms and recording phenotypic traits are time consuming and expensive. Resultant datasets often have many more features (markers) than samples (organisms). Therefore, models attempting to estimate the effects of markers often suffer from overfitting due to the curse of dimensionality. Feature selection is desirable in this setting to remove markers that do not appreciably affect the trait being predicted and amount to statistical noise.We present a differential evolution system for feature selection in genomic prediction problems and demonstrate its performance on simulated data. Code is available at: https://github.com/ianwhale/tblup.