Integration of Genome Association and Prediction

One of the ultimate goals of genetic research is to provide an accurate prediction for individuals’ genetic potential so that a better health care can be managed and selection can be applied to improve agricultural production. A reasonable accurate prediction could be achieved based on genetic markers covering the entire genome. The Bayesian method that calculates the effect for each marker, gave the best accuracy among the available statistical methods of genomic prediction. Due to its heavy computation demand, it is not feasible for analyzing dense markers resulted from next generation genotyping technology, e.g. genotyping by sequencing. In this study, we modified the mixed linear model method that used to have lower accuracy of prediction, but have advantage on computing speed. The modification was based on the compression algorithm which was originally proposed for genome-wide association study. We applied the new genomic prediction method in datasets from four species (human, dog, rice and Arabidopsis ). The new method improve accuracy 5~15% than the Bayes B method and increased speed markedly.