基于黄河鲤体质量性状的全基因组选择模型评估

Evaluation of genome-wide selection model based on body weight trait of Yellow River Carp(Cyprinus carpio)

为了对黄河鲤体质量性状进行全基因组关联分析及全基因组选择模型的预测准确性比较,采用鲤250K高密度SNP芯片对613尾黄河鲤(Cyprinus carpio)进行基因分型,并通过测定其体质量性状的表型信息进行全基因组关联分析,以及基于体质量性状、全基因组关联分析(genome-wide association study,GWAS)的不同变异数据集对GBLUP、贝叶斯、RKHS和机器学习模型等10种全基因组选择模型的预测准确性进行比较,以筛选出适用于黄河鲤体质量性状的全基因组选择模型。结果表明:通过GWAS定位到与体质量性状相关的5个SNP,位于1号和21号染色体上,进一步筛选关联SNP所在区域的基因,定位到WBP1L、GPM6B、TIMMDC1、RCAN1、EOGT基因;当选取与黄河鲤体质量性状表型相关的前100个SNP作为数据集,分析全基因组选择模型预测准确性时,机器学习模型XGBoost的预测准确性最高,为0.26,当SNP的数量分别为500、1000、3000、5000、20000时,GBLUP模型的准确性均最高,分别为0.3084、0.3444、0.4393、0.4526、0.4007,而XGBoost、LightGBM和GBLUP模型的变异系数则较低,说明模型预测的稳定性相对可靠。研究表明,本研究中共鉴定到5个与黄河鲤体质量性状相关的候选基因,分别为WBP1L、GPM6B、TIMMDC1、RCAN1、EOGT,10种全基因组选择模型中GBLUP模型的预测准确性最高,可用于黄河鲤体质量性状的基因组选育。

In order to compare the predictive accuracy of genome-wide association analysis and genome-wide selection models for body weight trait in Yellow River carp(Cyprinus carpio),613 samples of Yellow River carp were genotyped by a 250K high-density SNP chip and the phenotypic information of body weight trait was determined.The prediction accuracy of 10 genome-wide selection models,including GBLUP,Bayes,RKHS and machine learning models,were compared by genome-wide association analysis and different variation datasets based on body weight trait and genome-wide association study(GWAS)to screen out a genome-wide selection model suitable for body weight trait of Yellow River carp.It was found that the GWAS results for the weight trait of the Yellow River carp showed that significant and suggestive association sites were clustered,mostly located on chromosome 1,indicating that there were genes closely involved in the weight trait of the carp in this genomic region.A total of 5 significant SNPs were detected by Bonferroni correction,located on chromosomes 1 and 21.Further analysis of genes in the regions associated with these SNPs identified WBP1L,GPM6B,TIMMDC1,RCAN1,and EOGT genes.When selecting the top 100 SNPs related to the body weight phenotype of Yellow River carp as the dataset and analyzing the predictive accuracy of the genome selection model,the machine learning model XGBoost had the highest predictive accuracy(0.26).When the number of SNPs was 500,1000,3000,5000,and 20000,the GBLUP model had the highest accuracy,with values of 0.3084,0.3444,0.4393,0.4526,and 0.4007,with the lower coefficients of variation in models of XGBoost,LightGBM and GBLUP,indicating that the stability of model predictions was relatively reliable.The findings showed that five candidate genes related to body weight trait of Yellow River carp were identified,namely WBP1L,GPM6B,TIMMDC1,RCAN1 and EOGT.GBLUP model has the maximal accuracy and can be used for genome selection of body weight trait of Yellow River carp in the 10 genome-wide selection models.