利用一种优化求解方法的中国荷斯坦牛重要经济性状全基因组关联分析

发布时间：2018-07-02 21:19

本文选题：全基因组关联分析 + 中国荷斯坦牛　；参考：《中国农业大学》2017年博士论文

【摘要】：全基因组关联分析(GWAS)近年来广泛用于人类疾病和动植物重要经济性状的候选基因挖掘工作中,多年来在各物种中都取得了一定的成功。但是,由于群体结构、分析模型等因素的限制,很多与性状相关的基因尚未被完全发掘。同时,大量的研究也报道了很多假阳性结果,降低了后续的工作效率。因此,找到合适的关联分析方法,降低假阳性比例的同时提高检验功效,对该领域的研究具有重要意义。本研究使用了一种优化求解的方法(FarmCPU)对中国荷斯坦牛的生长性状进行GWAS,并通过模拟对该方法在阈性状GWAS中的应用进行了评价。首先,本研究利用FarmCPU方法对3325头中国荷斯坦牛的6、12、18和24月龄体高和胸围进行了 GWAS分析,发现了 27个SNP位点存在与生长性状在全基因组水平的显著相关,并通过候选基因分析找到66个候选基因。在对候选基因信息的进一步挖掘过程中,找到ATP1A1、DYRK1A、JUN、CEP135、CYP26B1、MYC、SOX6和FGFRLI1等8个候选基因在人类和小鼠的骨骼和肌肉发育中发挥重要作用,但是这些候选基因在奶牛的生长性状研究中均为首次报道。这一结果证明FarmCPU方法可用于奶牛生长性状的GWAS研究。之后,本研究基于logistic回归模型(1ogistic regression model)理论提出了新的阈性状模拟方法并开发了相应的阈性状模拟程序。新方法有别于传统的基于阈模型理论的阈值模拟法,在模拟过程中引入"偏好性"概念,为阈性状的表型增加了随机性。通过对比不同参数组合下两种模拟方法模拟的表型在GWAS分析中的表现,发现现有的阈值模拟法进行模拟的阈性状表型进行GWAS检测时的检测效果要好于与新的模拟方法,这可能是导致对阈性状的GWAS方法类研究中模拟效果优于实际数据的原因。用两种模拟方法均可以证明,阈性状GWAS分析时,logistic回归模型在检测阈性状时的效果与一般线性模型没有显著差异。奶牛的繁殖性状中阈性状较多,且遗传力较低,GWAS分析的检测效果不理想。通过对比不同方法在模拟阈性状中的表现,本研究证明FarmCPU方法在阈性状GWAS分析中的表现优于其他方法。因此,本研究通过对FarmCPU方法在中国荷斯坦牛生长性状和使用该群体部分数据模拟的阈性状进行GWAS分析,发现FarmCPU在对中国荷斯坦牛重要经济性状的GWAS分析中具有优势。
[Abstract]:Genome-wide Association Analysis (Gwas) has been widely used in candidate gene mining for human diseases and important economic traits of animals and plants in recent years, and has been successful in various species for many years. However, due to the limitation of population structure and analysis model, many genes related to traits have not been fully explored. At the same time, a large number of studies also reported a lot of false positive results, reducing the efficiency of subsequent work. Therefore, it is of great significance for the research in this field to find a suitable association analysis method to reduce the false positive rate and improve the test efficacy. In this study, an optimal solution method (FarmCPU) was used to evaluate the growth traits of Chinese Holstein cattle by GWASS, and the application of the method in threshold traits was evaluated by simulation. Firstly, the body height and chest circumference of 3325 Chinese Holstein cattle at the age of 18 and 24 months were analyzed by using FarmCPU method. It was found that 27 SNP loci were significantly correlated with growth traits at the whole genome level. 66 candidate genes were found by candidate gene analysis. In the process of further mining candidate gene information, we found that the eight candidate genes, such as ATP-1A1DYRK1ANJUNP135, CYP26B1, MYCNSOX6 and FGFRLI1, play an important role in human and mouse skeletal and muscle development. However, these candidate genes are reported for the first time in the study of growth traits in dairy cattle. The results show that FarmCPU method can be used to study the growth traits of dairy cattle. Then, based on the logistic regression model (1ogistic regression model) theory), a new method of threshold trait simulation is proposed and a corresponding program is developed. The new method is different from the traditional threshold simulation method based on threshold model theory. The concept of "preference" is introduced in the simulation process, which increases randomness for the phenotypes of threshold traits. By comparing the phenotypic performance of the two simulation methods in Gwas analysis with different parameter combinations, it is found that the existing threshold simulation method is better than the new simulation method in detecting the phenotypes of threshold traits. This may be the reason that the simulation effect of Gwas method for threshold traits is better than the actual data. It was proved by two simulation methods that there was no significant difference in the effectiveness of the logistic regression model between the general linear model and the general linear model in the detection of threshold traits. There were more threshold traits in breeding traits of dairy cows, and the detection effect of Gwas analysis with lower heritability was not satisfactory. By comparing the performance of different methods in simulated threshold traits, this study proves that FarmCPU method is superior to other methods in threshold trait Gwas analysis. Therefore, this study analyzed the growth traits of Chinese Holstein cattle using FarmCPU method and the threshold traits simulated by part of the population data. It was found that FarmCPU had advantages in the analysis of important economic traits of Chinese Holstein cattle.
【学位授予单位】：中国农业大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：S823

【参考文献】