利用變異數型數量性狀基因座探討身體質量指數之基因與環境交互作用

Abstract

變異數數量性狀基因座 (Variance quantitative trait loci, vQTLs) 可識別與表型變異性相關的遺傳變異，並可能反映基因與環境 (G×E) 的交互作用。本研究利用台灣人體生物資料庫 (Taiwan Biobank, TWB)，探討了身體質量指數 (BMI) 變異性的遺傳架構。將分位數積分線性模型 (Quantile Integral Linear Model, QUAIL) 應用於合併數據集 (n = 118,908) 後，我們鑑定出 12 個達到全基因組顯著水準的 vQTL 基因座，其中主要訊號 (lead signals) 為位於 FTO 基因的 rs1421085 以及位於 MC4R 基因的 rs6567160。交互作用富集分析 (Interaction enrichment analysis) 顯示，相較於傳統的全基因組關聯分析 (GWAS) 基因座，排名前列的 vQTLs 在與規律運動習慣（定義為每週至少 3 次，每次持續超過 30 分鐘）的 G×E 交互作用中呈現較高的富集程度，尤其是在最嚴格的前 0.1–0.5% 閾值區間。 此外，我們從 TWB2 發現隊列 (discovery cohort, n = 93,708) 中推導出遺傳權重，構建了變異數多基因分數 (variance polygenic score, vPGS)，並在獨立的 TWB1 驗證隊列 (replication cohort, n = 25,200) 中評估其預測表現。分位數迴歸證實，vPGS 能顯著預測群體層級的 BMI 變異性 (β = 0.192, P = 2.78×10⁻¹¹)，且較高的 vPGS 五分位數與較大的 BMI 離散程度相關。vPGS 與運動的交互作用分析進一步揭示，vPGS 較高的個體對運動降低 BMI 的反應性較強(β = -0.070, P = 0.076)。這些發現支持 vQTL 與 vPGS 框架可作為描述台灣族群 BMI 遺傳可塑性 (genetic plasticity) 的有用工具。

Variance quantitative trait loci (vQTLs) identify genetic variants associated with phenotypic variability, and may reflect gene–environment (G×E) interactions. This study investigated the genetic architecture of body mass index (BMI) variability using the Taiwan Biobank (TWB). Applying the Quantile Integral Linear Model (QUAIL) to a merged dataset (n = 118,908), we identified 12 genome-wide significant vQTL loci, with the lead signals rs1421085 located in FTO and rs6567160 in MC4R. Interaction enrichment analysis demonstrated that top vQTLs showed superior enrichment for G×E interactions with regular exercise habits—defined as at least 3 times per week, with each session lasting more than 30 minutes—compared to conventional GWAS loci, particularly at the most stringent top 0.1–0.5% thresholds. Furthermore, we constructed a variance polygenic score (vPGS) by deriving genetic weights from the TWB2 discovery cohort (n = 93,708) and evaluating its predictive performance in the independent TWB1 replication cohort (n = 25,200). Quantile regression confirmed that the vPGS significantly predicted population-level BMI variability (β = 0.192, P = 2.78×10⁻¹¹), with higher vPGS quintiles associated with greater BMI dispersion. vPGS×exercise interaction analysis further revealed a trend toward greater BMI-lowering responsiveness to exercise among individuals with higher vPGS (β = −0.070, P = 0.076). These findings highlight the utility of vQTL and vPGS frameworks for characterizing the genetic plasticity of BMI in the Taiwanese population.