Additional file 1 of Variability of polygenic prediction for body mass index in Africa

Additional file 1: Table S1. Development of the multi-ancestry (MAMA) and UKBB PRS in AWI-Gen and the Estonian Biobank. Figure S1. A. Grid of p-value thresholds (5E-08 to 1) at which PRS were computed to determine the best predictive one. B. Selected bar plots of p-value thresholds at which PRS were computed, indicating the best predictive one. Table S2. Development of the South African PRS using the AWI-Gen dataset. Table S3. Development of the West African PRS using the AWI-Gen dataset. Table S4. Clumping and LD parameters in the full AWIGen target dataset before splitting into training and validation. Table S5: Clumping and LD parameters in the AWIGen South target dataset before splitting into training and validation. Table S6. Clumping and LD parameters in the full AWIGen West target dataset before splitting into training and validation. Table S7. PRS and sex interaction models. Table S8. PRS and sex interaction models using inverse rank normalized BMI adjusted for age and principal components. Table S9. PRS and socioeconomic status interaction full model. Table S10. PRS and alcohol interaction full model. Table S11. PRS and smoking status full model. Table S12. PRS and physical activity interaction full model.

补充材料1：表S1。基于AWI-Gen与爱沙尼亚生物库（Estonian Biobank）构建多祖先（multi-ancestry, MAMA）与英国生物库（UK Biobank, UKBB）多基因风险评分（Polygenic Risk Score, PRS）的流程。
图S1。A. 用于计算PRS的p值阈值网格（范围为5×10^-8至1），用于筛选最优预测效果的PRS模型；B. 经筛选的p值阈值柱状图，标注了对应最优预测效果PRS的阈值。
表S2。基于AWI-Gen数据集构建南非人群PRS的流程。
表S3。基于AWI-Gen数据集构建西非人群PRS的流程。
表S4。划分训练集与验证集前，完整AWIGen目标数据集的聚类（clumping）与连锁不平衡（Linkage Disequilibrium, LD）参数。
表S5。划分训练集与验证集前，AWIGen南部人群目标数据集的聚类与LD参数。
表S6。划分训练集与验证集前，完整AWIGen西部人群目标数据集的聚类与LD参数。
表S7。PRS与性别的交互作用模型。
表S8。经年龄与主成分校正后的逆秩标准化BMI与PRS、性别的交互作用模型。
表S9。PRS与社会经济地位的全量交互作用模型。
表S10。PRS与饮酒情况的全量交互作用模型。
表S11。PRS与吸烟状态的全量交互作用模型。
表S12。PRS与身体活动水平的全量交互作用模型。