以預測轉錄體來推論環境因子會減弱或加劇候選基因之不利影響

Abstract

基因-環境的交互作用(G×E)已經被發現對許多人體表型 (phenotypes) 與疾病具有影響。於全基因組關聯分析(Genome-wide association study，GWAS)中，由於多重檢定校正之懲罰(p值 < 5 x 〖10〗^(-8))，G×E 常難以被檢測出來。本研究發展新方法，避免多重檢定校正之嚴苛懲罰，推論環境因素是否會減弱或加劇候選基因的不利影響。我們先以 PrediXcan 的方法得到「基因校正表現」(genetically regulated expression，GReX) 與「轉錄風險分數」(transcriptome risk score，TRS)，再總和所有相關表型的TRS × E的標準分數(Z-score)平方值，並使用「重採樣」(resampling) 檢測TRS與E的交互作用是否顯著，以及推論環境因素會減弱或加劇候選基因的影響。 我們應用此方法於臺灣人體生物資料庫中25,460位個案，分析於肥胖有顯著影響的FTO基因(Fat mass and obesity-associated protein，Chromosome 16: 53,737,875-54,155,853)。從事規律運動與否則當作環境因子E (Environmental factor)。我們結合所有與肥胖有關的表型：「腰圍」(waist circumference)、「臀圍」(hip circumference)、「體脂肪率」(Body Fat Percentage) 以及「身體質量指數」(Body Mass Index)。以本論文之方法，於數個人體組織中確實能夠觀測到規律運動會減弱FTO基因的不利影響。 我們觀察到在不同組織下的FTO基因與規律運動的交互作用，能被顯著察覺者包括甲狀腺 (Thyroid) (p值 = 0.00074)、乳腺 (Breast Mammary) (p值 = 0.00072)、脛神經 (Nerve Tibial) (p值 = 0.0014)。甲狀腺分泌多寡與肥胖有關，有趣的是我們於此組織中觀察到FTO基因與規律運動的交互作用。本研究的方法能推論環境因素會減弱亦或是加劇候選基因的不利影響。

Gene-environment interactions (G×E) have been found to influence many human phenotypes and diseases. In genome-wide association studies (GWAS), G×E are usually difficult to detect due to the penalty of multiple testing correction. To address this issue, we here develop a new method to infer whether an environmental factor can attenuate or exacerbate the adverse effects of candidate genes while bypassing the harsh penalty of multiple testing. We first used PrediXcan to obtain the “genetically regulated expression” (GReX) and “transcriptome risk score” (TRS). Then, we calculated the sum of squared Z scores of TRS×E from all relevant phenotypes, and used resampling to detect whether the interaction between TRS and E is significant, and inferred whether environmental factors can attenuate or exacerbate the influence of a candidate gene. We applied our method to 25,460 subjects in the Taiwan Biobank. We analyzed the FTO gene (Fat mass and obesity-associated protein, Chromosome 16: 53,737,875-54,155,853 in GRCh37) that is related to obesity. Performing regular exercise or not was served as an environmental factor (E). We combined several obesity-related phenotypes, including waist circumference, hip circumference, body fat percentage and body mass index. According to our proposed method, several tissues suggest that performing regular exercise can attenuate the adverse impact of the FTO gene on obesity. We observed significant interactions between the FTO gene and regular exercise in different tissues, including thyroid (p-value = 0.00074), breast mammary (p-value = 0.00072), and tibial nerve (p-value = 0.0014). The amount of thyroid secretion is associated with obesity, and it is interesting that we detected FTO-by-exercise in this tissue. The method described in this study can infer whether an environmental factor attenuates or exacerbates the adverse influences of a candidate gene.