利用插補單核苷酸多態性探討台灣布魯蓋達氏症候群患者的全基因組關聯研究

Abstract

研究背景: 布魯蓋達氏症候群(Brugada syndrome, BrS)是一種心臟沒有結構上異常，卻突發性猝死的遺傳性心臟病，通常在半夜休息睡覺時發作，因此布魯蓋達氏症候群也被稱為突發不明性夜間猝死症候群。迄今已有多個與布魯蓋達氏症候群相關的基因被發現，但只能解釋30-35%的病患，其中SCN5A是最常被提及的致病基因，然在白人患者中，僅20-25%的患者具有此突變，而台灣患者具有此突變的比例更是不到10%，這說明不同種族的致病基因存在一定差異。本論文的研究目標是在台灣布魯蓋達氏症候群患者中，找到可以診斷疾病的致病基因，並比較白種人與台灣族群的致病基因差異。此外，布魯蓋達氏症候群的症狀，也可以分為較嚴重的猝死和較輕微的胸口不適，因此另一個目標為找到可以預測疾病嚴重程度的基因。 材料與方法: 資料來源為台大醫院的313位布魯蓋達氏症候群患者的單核苷酸多態性資料(Single nucleotide polymorphisms, SNP)，檢驗位點的晶片為Axiom Genome-Wide TWB 2.0 array與Axiom Genome-Wide TPM array，並以台灣人體生物資料庫(Taiwan Biobank)作為對照組，透過全基因組關聯研究(Genome-wide association study, GWAS) 探討兩組之間的基因變異。除了透過晶片上檢驗的數十萬個變異位點，我們使用IMPUTE2演算法來進行基因型插補，以將研究的變異位點擴增至全基因組，期望找到更多與布魯蓋達氏症候群關聯的基因。在實務應用上，我們嘗試建立多基因風險評分模型(Polygenic risk score, PRS)，幫助我們達到評估個體罹病風險的目的。 結果: 使用原始資料分析致病基因時，有50個變異位點達到邦佛洛尼校正法的顯著水準(P值<1.14e-07)；而在插補資料的分析中，有910個變異位點達到邦佛洛尼校正法的顯著水準(P值<1.36e-08)。將結果整合後，一共有185個彼此之間不存在連鎖不平衡的顯著變異位點，其中7個變異位點在已有的文獻中，證明了他們與心電圖、血液相關測量值的關聯。在多基因風險評分方面，設定P值閾值為4.04e-06，使用100個變異位點來計算罹病的風險分數，其解釋力可達到0.707。在原始資料的預後基因分析中，有2個變異位點達到顯著水準(P值<1e-05)；而使用插補資料分析時，有28個變異位點達到顯著水準(P值<1e-05)。整合分析結果後，有2個不存在連鎖不平衡的顯著變異位點，目前尚未有文獻說明他們與布魯蓋達氏症候群相關的性狀有關聯。在多基因風險評分方面，設定P值閾值為2e-04，使用50個變異位點來計算患者預後的風險分數，其解釋力僅0.022。 結論: 本論文所建立的多基因風險評分模型，在罹患疾病的預測中，具有70%以上的解釋力，因此透過基因檢測，可以幫助醫生更早地判斷個體風險，提供進一步的干預措施，例如:給予患者生活方式建議、提高家屬警覺，透過了解遺傳風險，做為決定治療方式的參考。

Background: Brugada syndrome (BrS) is a genetic disorder with no structural abnormalities of the heart, but BrS is highly associated with sudden cardiac death. It usually occurs in the middle of the night when resting and sleeping. BrS is also known as sudden unexplained nocturnal death syndrome. Till now, several BrS-associated genes have been reported; however, only 30-35% of BrS patients can be genetically diagnosed, even if we included the major gene, SCN5A. Notably, only 20-25% of Caucasian patients have SCN5A mutations, and such proportion of Taiwanese patients is less than 10%, indicating that there are huge differences in different races. To address these issues, this thesis aims to find the susceptible genes of BrS in Taiwanese patients and to compare their differences between Caucasian and Taiwanese patients. In addition to diagnostic genes, the symptoms of BrS can divided into 2 categories: more severe (sudden death) and less severe (chest discomfort). Therefore, the second aim is to develop a prognostic prediction model for BrS patients in Taiwan. Materials and Method: The BrS patients were enrolled from National Taiwan University Hospital and their SNP data were examined by using the Axiom Genome-Wide TWB 2.0 array and Axiom Genome-Wide TPM array. The total number of BrS patients analyzed in this study was 313, and the health control were selected from Taiwan Biobank. By using the case control design, a genome-wide association study (GWAS) was performed accordingly. In addition to the genotyped SNPs examined from the chip, we used the IMPUTE2 algorithm to obtain more SNPs in the human genome. Such approach can help us to to find more genes associated with BrS and extend the coverage of SNPs. To summarize the importance of identified SNPs, we developed two polygenic risk score (PRS) models to evaluate the incidence risk and the prognostic risk of BrS, respectively. Results: There are 50 SNPs reaching Bonferroni-adjusted significance (P-value < 1.14e-07) from the whole SNPs analyzed by using the array, and 910 significant SNPs were identified (P-value < 1.36e-08) from the imputed SNPs. After being integrated, a total of 185 lead SNPs were reported, and 7 of them were associated with ECG, and blood measurement. In the PRS model for the case control design, the explanatory power can be up to 0.707 under the settings as the P-value < 4.04e-06 and 100 SNPs in the model. Regarding the SNP analysis of the prognostic model in BrS, 2 SNPs reached significance level (P-value < 1e-05) from the whole SNPs analyzed by using the array, and 28 SNPs were identified (P-value < 1e-05) from the imputed SNPs. After being integrated, there are 2 lead SNPs. Currently, no literature shows the associations of these 2 SNPs with any trait known for BrS. In the PRS model for the prognosis, the explanatory power is 0.022 under the settings as the P-value < 2e-04, and 50 SNPs in the model. Conclusion: The PRS developed in this study had more than 70% explanatory power in predicting the incidence of BrS. The results suggest providing genetic testing for BrS may help medical doctors to evaluate the risk earlier. Further interventions can be provided such as giving lifestyle advice to patients at an early stage, and raising the awareness of family members. By understanding the genetic risk can further improve the timing of providing interventions and treatment decisions.