AN ITERATIVE REGRESSION METHOD FOR GENOMIC PREDICTION

摘要

A method for improving the computational efficiency of estimation of Bayesian methods, such as Bayesian MCMC models, for performing genomic analysis of SNP data to estimate breeding values, Quantitative Trait Locus (QTLs), or genomic locations associated with a disease or trait. The method extends the BayesR method by using a blocked Gibbs sampling approach to estimate SNO effects comprising breaking the SNPS into non overlapping blocks of markers and sequentially processing each block according to a block ordering. Each regression coefficient in the block is then sequentially estimated, starting from a starting index. The method significantly reduces the computational time to estimate model parameters and can be applied to both single trait and multi-trait analyses.