請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/49431
標題: | 建立台灣人心血管疾病預測模型並驗證不同種族的基因風險分數之應用 Predictive modeling of Cardiovascular Diseases for Taiwanese and Validating the Genetic Risk Scores(GRSs) Derived from Different Ethnics |
作者: | Chung-Yu Huang 黃涱煜 |
指導教授: | 盧子彬(Tzu-Pin Lu) |
關鍵字: | 心血管疾病,單核甘酸多型性,基因風險分數,傳統危險因子,平均混和模型,F1分數, single nucleotide polymorphisms (SNPs),cardiovascular disease (CVD),genetic risk scores (GRS),average blending model,F1-score, |
出版年 : | 2020 |
學位: | 碩士 |
摘要: | 全基因體關聯分析(genome-wide association studies; GWAS)已發現許多與心血管疾病相關的單核甘酸多型性 (single nucleotide polymorphism; SNP),然而大多數的研究集中在歐美,過去已有研究指出歐洲的基因風險分數 (genetic risk score; GRS) 與心血管疾病達顯著相關,因此現今預測模型的建立大多也會納入基因風險分數作為重要變數。然而由於種族的差異,目前尚無研究指出歐美的基因風險分數是否能有效地應用於台灣心血管疾病的病患上。因此本研究第一個目的是驗證歐美所發展的基因風險分數能否有效地應用於台灣資料,而第二個目的是建立台灣人專屬的心血管疾病預測模型。本研究之基因數據來自三個西方研究與一項中國研究,而該中國的研究均採用東亞全基因體關聯分析所發表的顯著位點作為分析標的,可視為亞洲研究之代表。在本研究中我們選取各研究分析的單核甘酸多型性標的並用相同的方法計算基因風險分數,再進行相關強度與區別能力的比較,區別能力則是以曲線下面積(Area Under Curve; AUC) 做判別。進行完驗證後,會根據台灣人體資料庫(Taiwan Biobank; TWB)挑選這些外部的單核甘酸多型性並且以風險等位基因頻率之倒數(inversed risk allele frequency; inversed RAF)為內部的權重做加權,而後以平均混和模型 (Average blending model) 搭配傳統危險因子 (traditional risk factors; TRFs) 來建立心血管疾病的預測模型,並評估其與外部基因風險分數的表現差異。在相關強度的比較結果中,唯有東亞的單核甘酸多型性組成之基因風險分數與台灣心血管疾病達統計上顯著,而其風險比(odds ratio; OR)為1.5 (95% 信賴區間,1.20 – 1.87),代表東亞基因風險分數可做為台灣心血管疾病的危險預測因子,但在曲線下面積的比較中發現東亞與歐洲的基因風險分數均無法顯著地改善模型區別能力。在模型表現比較中,東亞基因風險分數的F1分數(F1 score)改善程度優於其他歐洲的基因風險分數,說明東亞發表之顯著單核甘酸多型性相較於歐洲更適合於台灣人。若使用內部權重加權之基因風險分數,其改善模型的程度優於外部的東亞基因風險分數,且其改善程度並非隨機發生,這些結果說明了即使應用了外部發表的顯著之單核甘酸多型性,也須使用台灣人的內部權重來建立模型。然而,加入基因數據的改善程度有限,顯示出本模型尚無法應用於實際臨床端,仍有賴後續研究改善。 In spite the fact that multiple single nucleotide polymorphisms (SNPs) associated with cardiovascular disease (CVD) have been identified by genome-wide association studies (GWAS) in European populations, there were no evidences showing that the external genetic risk scores (GRSs) developed in European studies could be properly used in Taiwanese. This study aimed to validate the GRSs from European and East Asian, and build the predictive models with these GRSs. The data was obtained from Taiwan Biobank(TWB) and there were 924 cases and 13671 controls in this study. We selected three western studies and one Chinese study for GRSs validation. The effect sizes and AUCs were used for comparison. After validation, we used average blending algorithm to construct the predictive models with different external GRSs. In addition, we selected the SNPs in TWB with F1 scores and p-value to compute the GRSs with internal coefficients, which were the inversed risk allele frequencies (RAF). The results showed that only East Asian GRS was significantly associated with CVDs in Taiwanese and the quintile OR was 1.50 (95%CI, 1.20 – 1.87) after adjusted for traditional risk factors (TRFs). Besides, the performance of predictive model built with East Asian GRS was better than the other external GRSs, which suggested that East Asian GRS was the predictor of CVDs. Furthermore, the GRSs selected in TWB and weighted by internal coefficients were better in improvement, which suggested that external associated SNPs should be weighted by internal weights. However, the modest improvements in discrimination illustrated that the clinical application was limited at present. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/49431 |
DOI: | 10.6342/NTU202003097 |
全文授權: | 有償授權 |
顯示於系所單位: | 流行病學與預防醫學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-1208202015113100.pdf 目前未授權公開取用 | 1.52 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。