Development Of Methods For Non-additive Genome-wide Association Studies And Its Application In Pigs,Rats And Mice

Factors contributing to the complexity of complex traits are usually classified into gene,environment and gene-environment interactions,and the genes could affect phenotypes by their additive,dominant and epistasis effect.However,in the field of agricultural animal,the model of general genome-wide association study(GWAS)often simply consider the additive genetic effect alone and the deficiency of non-additive effects could cause the omission of many quantitative trait loci(QTLs)and then cause difficulty in revealing the genetic architecture of complex trait systematically and completely.In this study,we try to optimize and upgrade the general GWAS model from two aspects,the gene-sex interaction and the dominant effect between two alleles.Then we built up two independent analysis pipelines which different from the additive GWAS and aim to detect and classify sex-specific QTLs and dominant/overdominant QTLs,respectively.Finally we applied these two pipelines into three different stocks,the pig resource stock(n=1005),the rat heterogeneous stock(n=1407)and the mice heterogeneous stock(n=2002)and explore the non-additive genetic architecture of various complex traits(hematology,immunology,carcass,growth,bone and several behavioral experiment traits related to anxiety or fear)and gene expression level in multiple tissues(liver,muscle,lung and hippocampus).On the aspect of the gene-sex interaction,we extracted 213 different traits in pig F2resource stock to analysis the sexually dimorphic genetic architecture.Firstly,we found that males and females represent a range of differences(from 0.74 to 1.58 fold)across all traits and 102 traits showed significant sex-specific differences(q value<0.05).Of those traits,the males had greater abdominal fat weight,greater intramuscular fat content,greater leaf fat weight and higher levels of saturated fatty acids in both their fat and muscle tissues.In contrast,females had higher levels of unsaturated fatty acids and a higher average daily gain from 21 to 46 days old.Next,we estimated the heritability of the 213 traits using male and female kinship matrices and found that the heritability of many traits are significant different between males(more heritable in meat quality,platelet distribution width and serum glucose levels)and females(more heritable in fatty acid composition,weight at 46 days old and skin thickness).Lastly,we performed a sex-stratified GWAS and assessed the sex-specific differences at each locus using z-statistics.A total of 43 significant sexually dimorphic loci(P<2.5×10^-5)that related to 22 traits(meat quality,hematological parameters,fatty acid composition and food intake)were found and the most significant signatures(P=4.8×10^-7)were detected for hue angle(H*)value on chromosome 6 which is adjacent to a potential causal gene(NUDT7).Moreover,2035 significant SNPs were detected only by the sex-stratified GWAS and this corresponds to 13.6%of the loci detected by the general GWAS.All these results have demonstrated an important role of gene-sex interaction for the genetic architecture of complex traits.On another aspect of the dominant effect,we investigated into all three stocks(pig,rat and mouse)for the analysis of dominant genetic architecture.First of all,based on a transformed linear mixed model,we corrected the population structure using both the additive and dominant genetic relationship matrix(GRM)and built up models to detect and classify the additive,partial-dominant,dominant and overdominant QTLs.Besides,we simulated several different phenotypes to test the specificity and sensibility of the detection models and it turned out that our models could present a robustly high performance with various phenotype that consisted of different causal loci(additive,dominant or mixed).Next,when we analysis the additive and dominant genetic architecture in three stocks,we found several new discoveries(1)The traits with high dominant variance component(VC)across all the three stocks are mainly enriched in hematology,immunology,serum biochemistry and growth trait.Besides,the traits with high dominant VC are always present a high additive VC.(2)Based on the full model including both additive and dominant effect,we detected352 QTLs,179 QTLs and 116 QTLs above the suggestive significant threshold in the pig,rat and mouse stock,respectively.We also classified these QTLs into three categories by the ratio of their additive and dominant effect,the proportions of different hereditary types among these QTLs across three stocks are quite consistent,which are 55.6%(additive):16.1%(dominant):28.3%(overdominant)on average.(3)By mining the dominant and overdominant QTLs across three stocks,we detected many interesting candidate genes.In pigs,we found two candidate genes named MLN and ADSL for the dominant QTL related to the body weight on age 240 days and another overdominant QTL related to the hemoglobin on age 46 days.In rats,we found candidate genes named FANCA and PLCL2 for two overdominant e QTLs related to the mean corpuscular hemoglobin and the absolute CD8+T cells.In mouse,we found candidate genes named BAT3 and APOA5 for two overdominant e QTLs related to the proportion of CD3+cells expressing CD8 and the glucose slope value.Lastly,we defined the cis-e QTLs as the e QTLs within 1Mb of the gene region and compared the relationship between hereditary types(additive or non-additive)and regulatory acting(Cis or Trans)of expression QTLs and we firstly discovered the trans-acting enrichment among non-additive e QTLs in mammals.