Genome-wide Association Study For Economically Important Traits In Chinese Holstein Based On One Optimized Algorithm

Genome-wide association study(GWAS)has been successfully used for detecting associated markers and identifying candidate genes in human,plant and animal during recent years.However,only genes of large effect are detected frequently in different studies,due to the disadvantages of the population structure and model of the analysis.These disadvantages may also cause false positive results,which increase the difficulty for the further study.Thus,it is a milestone to find an efficient associate approach to reduce the false positive and increase the power of testing.In this study,FarmCPU is implemented as an optimized algorithm of GWAS for growth traits and simulated threshold trait in Chinese Holstein.FarmCPU is used to detect SNPs associated to hip height and heart girth at 6,12,18,24 months for 3,325 Chinese Holstein.27 SNPs are significantly associated with the growth traits,and 66 genes are located by further candidate gene study.Comparing with the annotation information,8 of the genes,ATP1A1?DYRK1A?JUN?CEP135?CYP26B1?MYC?SOX6 and FGFRL1 are reported highly related to development and skeletal and muscular growth in human and mouse.These genes have never been reported in other studies in dairy cattle growth trait.This result illustrates that FarmCPU has advantage in GWAS for dairy cattle growth traits.A new approach to simulate the threshold trait is achieved based on logistic regression model.Different to the traditional approach based on threshold model,this new approach proposes a new idea called 'preferrence' and add a probability when converting the liability to categorical data.After comparing these two simulation approaches of threshold trait by using different combinations of simulation parameters,the new approach will weaken the association between phenotype and markers compare with traditional one.This result implies that the traditional simulation approach may cause the test of new method on GWAS of threshold trait over estimated.The two approaches are selected to simulate threshold trait,which is used to compare different GWAS method.The result is showing that it is not a decisive factor for choosing linear or non-linear model in GWAS of threshold trait.In other words,an evidence is provided for using linear model to analyze threshold trait.After comparing the performance of different GWAS methods,FarmCPU shows advantage than other methods.This study implements a GWAS on growth trait in Chinese Holstein using FarmCPU,and tests the performance of FarmCPU on analyzing simulated threshold traits.In a word,the result suggest that FarmCPU can be used in GWAS on important economic traits in Chinese Holstein.