基于靶向捕获测序的猪单倍型基因组选择效果研究

The Efficiency of Haplotype-Based Genomic Selection Using Genotyping by Target Sequencing in Pigs

【目的】探索靶向捕获测序技术的猪单倍型基因组选择研究效果,为我国猪分子育种提供借鉴。【方法】收集了1267头大白猪的生长性能测定记录和800头大白猪的繁殖记录,利用基于靶向捕获测序技术(genotyping by target sequencing,GBTS)开发的猪50K液相芯片(液相50K)以及靶向捕获重测序数据进行基因型分型,对靶向捕获重测序数据进行单倍型分析,比较固定SNP数目、固定物理间距和靶向block三种单倍型区块划分方法以及单个SNP数据的基因组选择准确性。其中靶向block方法是基于靶向捕获测序技术设计的一种单倍型区块划分方法,通过将靶位点与其上下游400 bp内SNP(mSNP)组合为一个block的方式构建单倍型。本研究对每个单倍型区块采用Beagle5.1进行单倍型推断后,将不同单倍型重新编码为单倍型等位基因,然后利用单倍型剂量模型构建单倍型基因型矩阵,采用一步法模型(ssGBLUP),估计达百公斤体重日龄(AGE)、百公斤活体背膘厚(BF)和总产仔数(TNB)三个性状的基因组育种值。对两个生长性状(AGE和BF)使用双性状动物模型进行估计,对总产仔数性状使用单性状重复力模型。使用年轻群体验证和5倍交叉验证两种验证方法,计算基因组估计育种值与育种值之间的相关系数和基因组估计育种值对估计育种值的回归系数,分别评价单倍型基因组预测准确性和无偏性。【结果】年轻群体作为验证群体的基因组预测结果表明,相较于液相50K SNP,基因型质量控制后,靶向捕获重测序标记数由液相50K的42302增加到了88105,但其对三个性状的基因组选择准确性反而有所下降。通过靶向block方法划分的单倍型区块平均包含SNP 2.08个、单倍型等位基因5.67个。基于三种单倍型区块划分方法进行的单倍型基因组选择准确性都得到了提高,其中靶向block提升幅度最大,AGE、BF和TNB三个性状准确性比液相50K分别提高了4.80%,1.98%和6.04%,靶向block多数情况下基因组预测无偏性最优。交叉验证结果与年轻群体验证相似,靶向block方法相较于单标记SNP和其他单倍型区块划分方法均有一定优势,固定间距400 bp划分单倍型的基因组选择效果与靶向block单倍型接近但计算时间增加;固定2个和5个SNP单倍型区块划分方法单倍型基因组预测准确性较单个SNP均有一定提升,但低于靶向block单倍型。【结论】由于液相芯片标记的特点,靶向重测序数据mSNP与液相50K SNP芯片标记多数处于高度连锁不平衡状态,利用靶向block划分的单倍型基因组选择准确性优于50K SNP以及其他两种单倍型区块划分方法,进一步利用了液相芯片的技术优势,拓宽了基于GBTS技术的液相芯片在基因组分析方面的应用。

【Objective】This study explored the efficiency of haplotype-based genomic selection using genotyping by target sequencing technology in order to provide the useful experience for molecular breeding of pigs in China.【Method】The growth traits records of 1267 Large White pigs and the reproduction trait records of 800 Large White pigs were collected and genotyped by using the porcine 50K liquid-phase SNP panel(liquid-phase 50K)based on the Genotyping by Target Sequencing(GBTS)as well as its target resequencing data.Three strategies for haplotype block partition,including fixed SNP number,fixed length,and target block,were proposed to construct haplotypes,and the corresponding haplotype-based genomic prediction were compared with single SNP genomic prediction.The target block partition strategy was mainly based on target resequencing data obtained by the GBTS,and one haplotype block contained one target site in liquid-phase 50K and its upstream and downstream SNPs(mSNP)within 400 bp from target resequencing data.In this study,after phasing haplotype using Beagle 5.1 within each haplotype block,the different haplotypes were recoded as haplotype alleles,and then haplotype allele matrix was constructed using haplotype dosage model.A single-step GBLUP model(ssGBLUP)was then used to estimate genomic breeding values for three traits of days at 100 kg body weight(AGE),backfat thickness at 100 kg(BF),and total number of born(TNB).A two-trait animal model was implemented for two growth traits of AGE and BF,and repeatability model for TNB.Younger validation population and five-fold cross validation were carried out to assess the efficiency of genomic prediction.The correlation coefficients between estimated breeding values and genomic estimated breeding values were measured as the accuracy of genomic prediction,and the regression coefficients of genomic estimated breeding values on estimated breeding values were measured unbiasedness,respectively.【Result】The results from the younger validation population showed that the target resequencing data yielded lower accuracies of genomic selection on three traits than liquid-phase 50K SNP,even its number of SNP after genotype quality control was increased from 42302 in liquid phase 50K to 88105.The haplotype blocks partitioned by target block strategy contained 2.08 SNPs and 5.67 haplotype alleles on average.The accuracies of haplotype-based genomic selection based on all three haplotype block partition strategies were improved,and the target block gained the largest improvement,yielding 4.80%,1.98%and 6.04%higher accuracies on AGE,BF and TNB than liquid-phase 50K,respectively.In addition,the target block strategy generated the lowest bias in most scenarios as well.The five-fold cross-validation obtained similar results as younger validation population did,target block gained advantages over both single SNP and other haplotype block partition methods.The fixed length of 400 bp block partition strategy performed comparable with target block,while it was time demanding.Although fixed 2 and 5 SNP haplotype block partition methods performed better than single SNP method,they were worse than target block.【Conclusion】Due to the short length,the linkage disequilibrium between most of the target sites in liquid-phase 50K SNPs and their mSNPs in same blocks are strong,resulting in target block strategy obtained higher accuracy in haplotype-based genomic selection than single SNP and other haplotype block partition strategies,which could make use of the technique advantage of GBTS,and further broaden the application of liquid-phase chip.