Methodologies And Software Development In The Detection Of Main-Effect QTNs,and Environmentaland Epistatic Interactions In Association Mapping Populations

With the development of sequencing technologies and the decrease of sequencing cost,genome-wide association studies（GWAS）have become an important method in the genetic dissection of complex traits in order to promote genetic development in animal,plant and human.Since the establishment of mixed-model framework in GWAS,a series of new approaches have been proposed and their corresponding software packages have been developed to save running time and increase statistical power in quantitative trait nucleotide（QTN）detection,making them become the key methods and software packages for the genetic analyses of complex traits in human,animal and plant.However,these methods estimate only allelic substitution effect,additive effect is confounded with dominant effect,especially,no methods have been widely adopted to identify QTN-by-environment interaction（QEI）and QTN-by-QTN interaction（QQI）in association mapping population.To address these issues,in this study we proposed new methods to detect main-effect QTNs and QEIs of complex traits in partial NCⅡgenetic mating population and to identify QQI of complex traits in hybrid F₁ association mapping population.First,we established a unified three-variance-component mixed-model to detect main-effect QTNs,QEIs and QQIs,with high power and accuracies,quick calculation,and low false positive and negative rates.Then,a series of real datasets to be downloaded and simulated datasets were used to validate the 3Vmr MLM method.Finally,R software packages mr MLM for GWAS were developed.The main results are as follows.1)In main-effect QTNs detection in partial NCⅡgenetic mating design populations with 60 heterozygous parents and their 240 F₁ hybrids,there were 10000 markers for each individual,the phenotypic observation of quantitative trait for each individual was simulated based on six main-effect QTNs with the sizes（r²）of 0.03 to 0.15 and random error,and the replicates were 1000 times.As a result,the powers of the six simulated QTNs were 80,99,99,65,93 and 92（%）,respectively;their mean square errors（MSEs）for additive effect were 0.1135,0.4187,0.3590,0.2778,0.0798 and 0.1165,respectively;their MSEs for dominant effect were 0.2396,0.6677,0.6905,0.5002,0.1943 and 0.2453,respectively;false positive rate was 0.85‰,and false negative rate was 12.1%.The new method was used to associate the phenotypic observations of cotton fiber length in each environment and their BLUP values with each SNP marker in partial NCⅡgenetic mating design population with 60 heterogeneous parents and their 180 F₁ hybrids.As a result,27,33,32 and 38 significant QTNs for fiber length in upland cotton were identified under environments 1,2,3 and their BLUP values,respectively.Around these QTNs,5（18.5%）,6（18.2%）,7（21.9%）and 8（21.1%）known genes for cotton fiber length were mined,and six known genes were detected across various datasets;4（14.8%）,3（9.1%）,5（15.6%）and 4（10.5%）genes for cotton fiber length were found to be homologous to the known trichome-length-related genes in Arabidopsis,and two homologous genes were detected across various datasets.2)In QEI detection in partial NCⅡgenetic mating design populations with 54heterozygous parents and their 146 F₁ hybrids,there were 10000 markers for each individual,the phenotypic observation of quantitative trait for each individual under a single environment was simulated based on five main-effect QTNs with the sizes（r²）of0.03 to 0.10,environmental effect,five QEIs with the sizes（r²）of 0.03 to 0.15,and random error,and the replicates were 1000 times.As a result,the powers of five simulated main-effect QTNs were 100,96.7,68.0,100 and 98.1（%）,respectively;their MSEs for additive effect were 0.3086,0.0985,0.0651,0.0843 and 0.5117,respectively;their MSEs for dominant effect were 0.2410,0.1312,0.1203,0.2313 and 0.2446,respectively;the powers of five simulated QEIs were 70.1,89.4,99.6,99.7 and 98.8（%）,respectively;their MSEs for additive-by-environment interaction effect were 0.1382,0.0888,0.0754,0.2314 and 0.0728,respectively;their MSEs for additive-by-environment interaction effect were 0.2030,0.2167,0.1388,0.2448 and 0.2025,respectively;false positive rate was 0.48‰,and false negative rate was 8.0%.The new method was used to jointly associate all the phenotypic observations of upland cotton fiber length in three environments with each SNP marker in partial NCⅡgenetic mating design population with 60 heterogeneous parents and their 180 F₁ hybrids.As a result,39 main-effect QTNs and 37 QEIs for cotton fiber length under three environments were identified.Around these main-effect QTNs,7（18.0%）known genes for cotton fiber length were mined,and four known genes were also detected in a single environment;6（15.4%）genes for cotton fiber length were found to be homologous to the known trichome-length-related genes in Arabidopsis,and two homologous genes were also detected in a single environment.Around these QEIs,six（16.2%）known genes for cotton fiber length were confirmed by molecular biological experiments to interact with environments,and two known genes were also detected in a single environment;five（13.5%）genes for cotton fiber length were found to be homologous to the known trichome-length-related genes in Arabidopsis,four homologous genes were confirmed by molecular biological experiments to interact with environments,and one homologous gene was detected in a single environment.3)In QQI detection in 400 hybrid F₁ populations,there were 1000 markers for each hybrid F₁ individual,the phenotypic observation of quantitative trait for each individual was simulated based on eight QQIs with the sizes（r²）of 0.05 to 0.08 and random error,and the replicates were 100 times.As a result,the powers of eight simulated QQIs were54,82,74,80,55,80,62 and 58（%）,respectively;their MSEs for QQI effects were 0.150,0.056,0.092,0.184,0.027,0.148,0.113 and 0.096,respectively;false positive rates in main-effect QTNs and QQI detection were 0.60 and 0.10（‰）,respectively,and false negative rate in QQI detection was 31.9%.To further confirm the performances of the new method,it was used to associate the mean values of each yield or quality trait across two environments in 1439 rice F₁ hybrids with 15,943 bin markers.As a result,79main-effect QTNs and 622 QQIs were identified.Around these main-effect QTNs,43（54.4%）known genes for traits of interest were mined.Around these QQIs,19（3.05%）known gene-by-gene interactions for traits of interest were validated by molecular experiment evidence;two genes around the two loci in each of 23 QQIs were found to be associated with the traits of interest,and their predicted interaction scores were larger than0.4;23 pairs of gene-by-gene interactions with molecular experiment evidence were predicted to be associated with the traits of interest.4)To popularize our multi-locus GWAS methodologies,the code and interface versions of R software packages mr MLM and mr MLM.GUI（v4.0.2）were developed,including the mr MLM,FASTmr MLM,FASTmr EMMA,p LARm EB,p KWm EB and ISIS EM-BLASSO methods.The software packages were validated by Monte Carlo simulation studies under various numbers of markers,individuals and CPUs,along with international application studies,and real data analyses in rice,maize and Simmental beef cattle.This study provides a new method and its software package for detecting main-effect QTNs,QEI and QQI in association mapping populations,and will promote the application of GWAS in the genetic analysis of quantitative traits in animal,plant and human.