Genomic Prediction for Quantitative Traits Is Improved by Mapping Variants to Gene Ontology Categories inDrosophila melanogaster

• DOI: 10.1534/genetics.116.187161
• Publication Date: 2016-05-28T02:49:56Z
• Authors: Stefan M Edwards; Izel F Sørensen; Pernille Sarup; Trudy F C Mackay; Peter Sørensen
• Keywords: Genetic Markers; 0301 basic medicine; Genomic selection; Genotype; Drosophila Genetic Reference Population; Genome, Insect; Quantitative Trait Loci; Startle response; Genomics; Shared data resource; Polymorphism, Single Nucleotide; 3. Good health; 03 medical and health sciences; GenPred; Drosophila melanogaster; Gene Ontology; Phenotype; Best linear unbiased prediction; Genomic feature models; Animals; Chill coma recovery time; Starvation resistance

Predicting individual quantitative trait phenotypes from high-resolution genomic polymorphism data is important for personalized medicine in humans, plant and animal breeding, adaptive evolution. However, this difficult populations of unrelated individuals when the number causal variants low relative to total polymorphisms individually have small effects on traits. We hypothesized that mapping molecular features such as genes their gene ontology categories could increase accuracy prediction models. developed a feature best linear unbiased (GFBLUP) model implements strategy applied it three traits (startle response, starvation resistance, chill coma recovery) unrelated, sequenced inbred lines Drosophila melanogaster Genetic Reference Panel. Our results indicate subsetting markers based increases predictive ability standard (GBLUP) model. Both models use all markers, but GFBLUP allows differential weighting genetic marker relationships, whereas GBLUP weighs relationships equally. Simulation studies show possible further complex using model, provided are enriched variants. prior information can predictions provides formal statistical framework leveraging evaluating across multiple experimental provide novel insights into architecture