以臺灣人體生物資料庫探勘糖尿病未知基因之全基因體關聯性研究

Abstract

糖尿病在台灣盛行率高，且在不同族群中的致病基因並不完全一致，因此針對台灣族群進行GWAS研究，尋找台灣族群未知的糖尿病致病基因能提供治療及預防台灣族群糖尿病的新方向。 本文嘗試結合模型導向分群法 (Model-based clustering) 與序列核關聯檢定 (Sequence Kernel Association Test, SKAT) 分析全基因體資料。模擬資料分析結果顯示兩者對次要等位基因頻率 (minor allele frequency, MAF) 大於0.05的致病變異都有大於0.75的檢定力，所需計算時間差異不大。模型導向分群法對於罕見致病變異(MAF<0.05)的資料檢定力較低，而SKAT可以調整參數達到極高的檢定力。 從台灣人體生物資料庫經品質控管的資料挑選出4453位糖尿病患者與隨機挑選17812位未患病者為控制組，同時使用兩種檢定方法分析，嘗試尋找台灣族群糖尿病的未知基因。模型導向分群法共找到338個變異位點(FDR=0.05)；SKAT檢定出197個30kb的片段(p值<0.001)。兩方法重疊的顯著片段或變異位點皆為最顯著前幾名基因，而這些基因都是已知的糖尿病基因，與之前對於亞洲族群的GWAS研究結果相似。模型導向分群法與SKAT有一些顯著的變異位點或片段雖不重疊，但基因功能上卻與糖尿病相關，可能為致病基因。這些基因大致分成內質網壓力相關基因、胰島β細胞功能或胰島素相關基因、糖尿病併發症相關的基因。內質網壓力在糖尿病發展上扮演重要角色，能誘發未摺疊蛋白反應 (unfolded protein response, UPR) 與高基氏體逆向運輸等機制。與UPR及高基氏體逆向運輸相關的兩個基因內的變異可能是細胞無法順利解除內質網壓力的一部份原因。影響胰島β細胞功能或胰島素十個相關基因所涉及的生理機制眾多，顯示糖尿病的複雜性，而因為這樣的複雜所以糖尿病遺傳力在不同研究中變化很大。糖尿病併發症相關的兩個基因可能成為研究糖尿病性心肌病新的方向；UMOD及TGFBRAP1與糖尿病腎臟病變的關聯能幫助了解台灣族群的糖尿病腎臟病變機制。

In the past decades, studies have shown a marked increase in the prevalence of diabetes in Taiwan. The risk of diabetes exhibits population-specific causal relationships with environmental and genetic factors. Genome-wide association studies (GWAS) aims to test significant associations between a specific disease and genetic variants in human genome. So far, GWAS has successfully identified a number of associated variants responsible for type 2 diabetes. Such findings are important to provide guidelines for diabetes disease prevention and control in public health policy. In this study, we performed a GWAS for identification of SNPs that may be associated with type 2 diabetes via two methods, Model-based clustering and Sequence Kernel Association Test (SKAT). Results of simulation studies show that the powers of both statistical approaches are higher than 0.75 when minor allele frequencies (MAFs) of disease-associated single-nucleotide polymorphisms (SNPs) are higher than 0.05. There is no much difference of computation time between the two approaches. When MAFs of SNPs are lower than 0.05, the power of Model-based clustering method is lower and SKAT can remain high power performance by tuning parameters. A total of 4453 type 2 diabetes (T2D) patients were collected at the Taiwan Biobank dataset with a medical history and a total of 17812 controls were randomly selected from the non-diabetes subjects. Both model-based clustering and Sequence Kernel Association Test (SKAT) methods are applied to the dataset for identification of novel genes targeting the specific Taiwan population. As a result, the model-based clustering method identified 338 significant SNPs with controlling the FDR at 5%, and 197 regions of 30 kb were declared to be significant by SKAT at p-values < 0.001. Most of the overlap in significantly associated SNPs between two GWAS methods have been identified as T2D susceptibility genes and validated by several GWAS results for the Asian population. For those non-overlapping significant SNPs and loci, we attempted to verify their functionality via gene annotation. These genes consisted of endoplasmic reticulum (ER) stress related genes, β cell and insulin related genes, as well as diabetes complications related genes. ER stress which plays an important role in diabetes development induces the unfolded protein response (UPR) and retrograde transport from the Golgi to the ER. Variants in UPR and retrograde transport related genes could be part of the reasons that cells can’t release ER stress. Ten β cell and insulin related genes are simultaneously involved in many pathways, explaining the complexity of diabetes. Such complexity may lead to the heritability variation across diabetes studies. Our analysis revealed two potential genetic variants, UMOD and TGFBRAP1, showing significant association with diabetes complications. Our findings provide an insight into the diabetic nephropathy and these candidate SNPs might be valuable for future study in the development of type 2 diabetes complications in Taiwan population.