肥胖的遗传流行病学研究以及双变量关联定位的理论探讨和应用

发布时间：2018-04-08 23:32

本文选题：肥胖　切入点：受试者操作特征曲线　出处：《湖南师范大学》2010年博士论文

【摘要】： 最近的研究数据表明,白种人群的肥胖诊断标准可能不适用于亚洲人。虽然在白种人群中已经开展了许多关于肥胖相关表型的全基因组关联研究,但是在中国人群中至今没有成功运用全基因组关联研究定位肥胖基因的报道。在统计遗传学领域,已有大量关于单个性状的关联分析方法研究,但是对于两个相关性状的关联分析方法研究,还少有开展。本文首先使用全身脂肪含量(PBF)作为金标准,在1109个男性个体和879个女性个体中采用受试者操作特征曲线(ROC)分析估计了三个简易测量指标,即体重指数(BMI)、腰围(WC)、腰髋比(WHR)预测肥胖的能力。分析表明,BMI、WC、WHR都与PBF高度相关,相关系数为0.45-0.75。真阳性率在男性中是82.4%到94.1%,在女性中是68.8%到86.3%。真阴性率在男性中是64.1%到84.7%,在女性中是56.9%到79.0%。BMI和WC两个指标在男性和女性不同年龄组中曲线下的面积比较高(0.76-0.92),而WHR这个指标稍微低-些(0.74-0.88)。研究表明,BMI和WC是两个很好的预测肥胖的指标,WHR次之。该研究同时发现,用BMI在中国人群中预测肥胖的标准要明显低于白种人群。接下来我们在597个中国北方人群中运用Affymetrix 500k芯片开展了全基因组关联研究定位影响BMI变异的新基因,此研究进一步在2955个中国南方人中进行了重复验证。通过一系列的质量控制程序后,281,533个单核苷酸多态性(SNP)可以进行关联分析。经过FDR方法进行多重校正后,8个SNP仍然与BMI关联(FDRq=0.033-0.048)。最显著的SNP是rs4633,位于catechol-0-methyltransferase (COMT)基因的外显子上(p=5.45×10-7,FDRq=0.033)。另外,位于EIF2AK4(eukaryotic translation initiation factor 2 alpha kinase 4)基因内的两个临近的SNP rs4432245和rs711906也与BMI显著关联(?)值分别为4.38×10-6，6.39×10-6,FDRq=0.048)。在南方人群的验证研究中,这两个位点与BMI的关联获得了证实(p=0.03和0.01)。除此之外,我们通过大量的计算机模拟,运用广义估计方程2(GEE2)对两个相关性状的关联定位方法进行研究。通过与单个性状关联研究的功效比较,发现双变量模型比单变量模型统计功效更高。而且双变量关联分析的假阳性率与单变量分析相比没有差别,都在设定的显著性水平0.05左右。所以我们提出的双变量分析模型是合理的。进一步对肥胖和骨质疏松两个相关性状的实验数据进行的双变量关联分析验证了该方法的优越性。本研究的创新性在于：1)第一次在中国南方人群中对衡量全身肥胖的简易指标进行评估；2)第一次在中国人群中运用全基因组关联分析方法定位影响BMI变异的新基因；3)第一次尝试运用广义估计方程2拟合基于群体设计的两个数量性状的关联定位研究。
[Abstract]:Recent research data suggest that criteria for the diagnosis of obesity in white populations may not be applicable to Asians.Although many genome-wide association studies have been carried out on obesity related phenotypes in white populations, there have been no reports on the use of genome-wide association studies to locate obesity genes in Chinese population.In the field of statistical genetics, there have been a lot of studies on the correlation analysis of single traits, but few studies have been carried out on the correlation analysis of two related traits.In this paper, the body fat content (PBF) was first used as the gold standard, and three simple measurements were estimated by using the operating characteristic curve of subjects in 1109 male and 879 female individuals.Body mass index (BMI), waist circumference (WCC) and waist hip ratio (WHR) can predict obesity.The analysis showed that the WHR were highly correlated with PBF, and the correlation coefficient was 0.45-0.75.The true positive rate was 82.4% to 94.1% for males and 68.8% to 86.3% for females.The true negative rate was 64.1% to 84.7in males and 56.9% to 79.0%.BMI and WC in females. The area under the curve was higher in males and females in different age groups, but WHR was slightly lower (0.74-0.88g).Studies show that BMI and WC are two good predictors of obesity, followed by WHR.The study also found that BMI was significantly lower in predicting obesity in Chinese than in white populations.Then we used Affymetrix 500k chip in 597 northern Chinese population to carry out the whole genome association study to locate the new gene that affected the BMI mutation. This study was further verified in 2955 southerners.After a series of quality control procedures, 281533 single nucleotide polymorphisms (SNPs) can be analyzed.After multiple correction by FDR, 8 SNP are still associated with BMI 0.033-0.048.The most significant SNP is rs4633, which is located on the exon of catechol-0-methyltransferase mott gene.In addition, two adjacent SNP rs4432245 and rs711906 located in the EIF2AK4(eukaryotic translation initiation factor 2 alpha kinase 4 gene were also significantly associated with BMI.The values were 4.38 脳 10-6 and 6.39 脳 10-6 respectively.The correlation between the two loci and BMI was confirmed in the southern population.In addition, through a large number of computer simulations, we use the generalized estimation equation 2 / GEE2) to study the correlation mapping of two related traits.By comparing the efficacy with the single trait correlation study, it is found that the bivariate model is more effective than the univariate model.The false positive rate of bivariate correlation analysis was not different from that of univariate analysis.So the bivariate analysis model proposed by us is reasonable.The bivariate correlation analysis of two related traits, obesity and osteoporosis, proved the superiority of this method.The innovation of this study lies in: (1) for the first time in southern China, we evaluated a simple index of general obesity) for the first time, we used a genome-wide association analysis method to locate a new gene affecting BMI mutation in Chinese population.3) for the first time, the generalized estimation equation 2 is used to fit the correlation mapping of two quantitative traits based on population design.
【学位授予单位】：湖南师范大学
【学位级别】：博士
【学位授予年份】：2010
【分类号】：R589.2;R181.3

【同被引文献】