Supplementary file for "Whole blood-based transcriptional risk score for nonobese type 2 diabetes predicts dynamic changes in glucose metabolism"

<strong>Supplemental Material</strong> <strong>Title:</strong> Whole blood-based transcriptional risk score for nonobese type 2 diabetes predicts dynamic changes in glucose metabolism <strong>Authors: </strong>Yanan Hou, Huajie Dai, Na Chen, Zhiyun Zhao, Qi Wang, Tianzhichao Hou, Jie Zheng, Tiange Wang, Mian Li, Hong Lin, Shuangyuan Wang, Ruizhi Zheng, Jieli Lu, Yu Xu, Yuhong Chen, Ruixin Liu, Guang Ning, Weiqing Wang, Yufang Bi, Jiqiu Wang, Min Xu <strong>Supporting information</strong> <strong>Table S1. </strong>Clinical characteristics of study participants. <strong>Table S2. </strong>Weights of 144 gene transcripts included in wb-TRS, ordered from most negative to most positive associated with nonobese type 2 diabetes. <strong>Table S3.</strong> Association of z-score normalized wb-TRS with cardiometabolic traits in initial cohort. <strong>Table S4. </strong>Receiver operator characteristic curves for prediction of nonobese type 2 diabetes. <strong>Table S5. </strong>Top30 Reactome pathways associated with genes in wb-TRS. <strong>Table S6.</strong> Used IVs for wb-TRS in initial cohort. <strong>Table S7. </strong>Transcriptional MR for causal relationships between gene transcripts included in wb-TRS and T2D in initial cohort. <strong>Table S8. </strong>Transcriptional MR for causal relationships between gene transcripts included in wb-TRS and T2D in AGEN. <strong>Fig S1. </strong>Flowchart of study participants selection. <strong>Fig S2. </strong>Cross-validation results for least absolute shrinkage and selection operator (LASSO) logistics-model for nonobese type 2 diabetes in training dataset. <strong>Fig S3.</strong> Violin plot of wb-TRS and gene transcripts most negatively (FUOM) and positively (RPF1) associated with nonobese type 2 diabetes. <strong>Fig S4. </strong>Stratified analysis. Data are odds ratio (OR) and 95% confidence interval (CI) calculated using multivariable logistic regression models. The results were adjusted for age, sex, body mass index, family history of diabetes, waist circumference, smoking and drinking status, education level, physical activity, systolic and diastolic blood pressure, total cholesterol, triglycerides, high-density and low-density lipoprotein cholesterol.