Genome-wide Regression Method For Powerfully Identification Of Large Effect Quantitative Trait Loci

Genome-wide multi-locus association analyses for the normally distributed traits were well developed to powerfully identify quantitative trait loci(QTLs),but they were difficultly extended to discrete traits and survival times.As important target traits in livestock breeding,discrete traits and survival time are often more significant than normal distribution traits.This study extends R/glmnet for linkage analysis to genome-wide association study(GWAS)by means of parallel computation,and proposes a genome-wide regression method,named the LASSO-LSR,which can powerfully identify large effect quantitative trait loci for normally distributed trait,discrete traits and survival time.In the LASSO-LSR,quantitative trait nucleotides(QTNs)candidates with nonzero genetic effects can be high efficiently chosen with LASSO by means of parallel,and then statistically inferred to identify QTNs with linear model for normally distributed traits,generalized linear model for discrete traits,or Cox proportional hazards model for survival times.By comparing the results of computer simulations and analyzing a series of publicly available data sets,the applicability and utility of LASSO-LSR are systematically demonstrated.This study confirms that the LASSO-LSR can more powerfully map and detect large effective QTLs and QTNs than genomewide mixed model for one SNP at a time(FaST-LMM)and the linear mixed models-least absolute shrinkage and selection operator methods(LMM-LASSO).