以多基因法來偵測基因環境交互作用

Abstract

基因環境交互作用在遺傳性狀上扮演著至關重要的角色，於遺傳流行病學上益發受到重視。然迄今為止，從全基因組關聯研究中偵測出基因環境交互作用仍然十分困難。為了克服這項難題，我們提出一個多基因方法來偵測基因環境交互作用。我們將全部單核甘酸多型性與環境因子交互作用效應綜合在一個檢定量中。如此一來，可迴避單核甘酸多型性逐一檢定或基因逐一檢定中所面臨的多重檢定校正懲罰，繼而提高統計檢定力。此外，透過多方的統計模擬評估，我們的方法除保持適當的型一誤差外，在多數情況下都比文獻上常用的多基因風險評分法更具檢定力。我們進而將這項方法應用至臺灣人體生物資料庫，探索於舒張壓、收縮壓、高血壓、強制呼出時肺活量、第一秒用力呼氣量、醣化血色素值、三酸甘油酯等性狀上，是否存在基因與吸菸、飲酒間的交互作用。我們發現除了基因與吸菸的交互作用對收縮壓的影響不達顯著外，其餘的基因環境交互作用皆達統計上的顯著 (P值<0.05)。總結言之，對於偵測基因環境交互作用，我們的多基因法是一正確且更具檢定力的方法。透過這項新方法，我們可偵測出更多過去未曾被發現的基因環境交互作用，並有助於增進我們對疾病成因的認知。

Gene-environment interaction (G x E interaction) plays a vital role in hereditary traits and has gained much attention in the field of genetic epidemiology. However, detecting G x E interactions in genome-wide association studies (GWAS) remains challenging to date. To address this difficulty, we here propose a polygenic test for detecting G x E interactions. We combine the interaction effects from all single-nucleotide polymorphisms (SNPs) in only one test statistic. Hence, we improve the statistical power by avoiding the harsh multiple-testing penalty in the single-marker analysis or the gene-based analysis. We evaluate the performance of our method with comprehensive simulations. Our method is shown to preserve the type I error rate, and it is more powerful than the commonly-used polygenic risk score (or genetic risk score) approach in most situations. Furthermore, we apply our method to Taiwan Biobank data to explore G x smoking or G x drinking interactions on diastolic blood pressure, systolic blood pressure, hypertension, forced vital capacity, forced expiratory volume in one second, glycated hemoglobin, triglyceride, etc. All the G x E polygenic test results are statistically significant (P-value < 0.05), except G x smoking interactions on systolic blood pressure. To conclude, our polygenic test is a valid and powerful approach for detecting G x E interactions. With this novel approach, we can identify more important G x E interactions that have not been reported previously. In this way, our approach can help to enhance the understanding of disease etiology.