Genome-wide genotyping data have opened the prospect of precision medicine and disease prevention by developing tests specific for individual subjects' specific set of genetic variants. Conventional GWAS [1] [2] have been hugely successful over the past decade in detecting specific single nucleotide polymorphisms (SNPs)  in association with many diseases [3] [4] and traits of interest [5]. The challenge in the post-GWAS era is to develop new methods to implement precision medicine and prediction of disease risk.

We have used novel methods of prediction using whole genome-enabled Bayesian and other models. In contrast to univariate approaches (as implemented in GWAS),  genomic prediction models have multivariate characteristics enabling them to avoid colinearity between SNPs, which was known as a drawback of GWAS.  As is well known, prior probability density functions of biomarker effects [6] have been playing a key role in Bayesian methods. In this work we assigned a mixture of linear and Gaussian distribution on the marker effects. We illustrate the model setup and the performance of genome-enabled models in predicting human disease outcomes by using later-onset Alzheimer's disease dataset as an example. Proposed Bayesian models have shown superb predictive capability than ordinary linear models, and have identified novel mutations that could not have been found by traditional analysis methods.