基于贝叶斯算法的中国西门塔尔牛胴体性状基因组选择的初步研究

发布时间：2018-01-24 06:10

本文关键词： 全基因组选择全基因组关联分析贝叶斯筛选位点交叉验证　出处：《湖南农业大学》2015年硕士论文　论文类型：学位论文

【摘要】：自2001年Meuwissen提出全基因组选择(Genomic selection, GS)理论以来,动物育种技术从标记辅助选择(marker assistant selection, MAS)的分子水平提高到了GS的组学水平,实现了动物育种技术的又一次跨越。随着生物芯片技术、高通量测序技术以及计算机技术的快速发展,在当下生命科学领域的大数据时代,使基因组选择技术在动物育种领域的应用成为可能。本研究利用821头中国西门塔尔牛资源群体,用基于高密度基因分型芯片数据对其米龙、和尚头、黄瓜条、后腱子和臀肉共5个性状进行GS研究。首先,利用上述5个性状进行显著环境效应的筛选研究,并构建环境效应方程。其次,对上述5个性状进行ST-Bayes A、 ST-Bayes B和ST-Bayes CPi三种算法的单性状(single trait, ST)基因组估计育种值(Genomic estimate breeding value, GEBV),构建三种Bayesian算法的全基因组效应值图谱。其中ST-Bayes B算法的GEBV估计准确性最高,ST-Bayes A和ST-Bayes CPi的估计准确性相差不大,但ST-Bayes CPi算法的准确性稍好于ST-Bayes A。再次,对上述5个性状进行全基因组关联分析(Genome wide association study, GWAS),以位点显著性大于5作为位点的显著标准。结果表明,显著位点主要集中在6号染色体上。本研究还对821头中国西门塔尔牛资源群体进行了GS筛选位点。首先,利用GPOPSIM软件进行数据模拟,并对基于筛选位点的估计准确性评价方法(PA, Prediction Accuracy)进行研究。即比较基于回归系数的CV (Cross Validation)方法和传统CV方法,并证明前者具有一定的优势。对回归系数CV方法进行参数优化,研究证明基于ST-Bayes A算法的筛选位点,利用PA估计准确性过程当中进行2倍并重复8次的CV算法准确性最佳。筛选位点采用相等的遗传距离、ST-Bayes A和ST-BayesB算法估计的效应值以及GWAS的P值四种不同策略并利用ST-Bayes A算法进行GEBV的估计。研究证明：采用ST-Bayes A和ST-Bayes B算法估计的效应值进行筛选位点效果最好,在7-8 K左右即可达到较高的准确性。其次,关于相等遗传距离筛选位点的策略。研究表明：在15K左右达到平台期,而利用GWAS的P值效果在较低密度时上升明显,在40 K左右相对于均匀的位点筛选方法体现出优势。
[Abstract]:In 2001, Meuwissen proposed the Genomic selection (GSN) theory. The molecular level of marker-assisted marker marker assistant selection (MAS) was increased to the level of GS in animal breeding. With the rapid development of biochip technology, high-throughput sequencing technology and computer technology, in the current life sciences in the big data era. The application of genome selection in animal breeding was made possible. In this study, 821 Chinese Simmental cattle resource populations were used to make use of high-density genotyping microarray data to compare the data of Milosaurus and Sangtou. GS was used to study the five characters of cucumber strip, posterior tendon and gluteal meat. Firstly, the significant environmental effects were screened by using the above five characters, and the environmental effect equation was constructed. ST-Bayes A, ST-Bayes B and ST-Bayes CPi algorithms were applied to single trait for the above five traits. Genomic estimate breeding value (GEBV). The whole genome effect map of three Bayesian algorithms is constructed, and the GEBV estimation accuracy of ST-Bayes B algorithm is the highest. The estimation accuracy of ST-Bayes A and ST-Bayes CPi is not different, but the accuracy of ST-Bayes CPi is better than that of ST-Bayes A. Genome wide association studyy (GWAS) was used to analyze the five traits mentioned above. The results showed that the significance of the locus was greater than 5. The significant loci were mainly located on chromosome 6. GS screening sites were also used in 821 Chinese Simmental cattle resource populations. First, the data were simulated by GPOPSIM software. The estimation accuracy evaluation method based on screening sites was evaluated by PA. Prediction Accuracy. comparing the CV Cross validation method based on regression coefficient with the traditional CV method. The regression coefficient CV method is optimized and the screening sites based on ST-Bayes A algorithm are proved. In the process of estimating accuracy of PA, the CV algorithm with 2 times and repeated 8 times has the best accuracy. The screening site adopts the same genetic distance. The effect value estimated by ST-Bayes A and ST-BayesB algorithm and the P value of GWAS are four different strategies and ST-Bayes is used. A algorithm is used to estimate GEBV. The results show that ST-Bayes A and ST-Bayes B algorithms have the best effect on screening sites. Higher accuracy can be achieved at 7-8K. Secondly, the strategy of screening loci with equal genetic distance is discussed. The results show that the platform stage is reached at 15K or so. The P value of GWAS increased obviously at low density, and it showed an advantage over the uniform site screening method at about 40 K.
【学位授予单位】：湖南农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S823

【相似文献】