BackgroundThe performance of peripheral blood transcriptional markers in evaluating the risk of type 2 diabetes (T2D) with normal weight is unknown. We developed a whole blood-based transcriptional risk score (wb-TRS) for nonobese T2D and assessed its contributions to disease risk and dynamic changes in glucose metabolism.Methods and findingsWe developed the wb-TRS in 1105 participants aged ≥40 years and in normal weight for up to 10 years from a well-defined community-based cohort with blood transcriptome data and validated it in an external dataset (253 overweight/obese participants from a dietary intervention trial with 3 repeated transcriptome data). Potential biology significance and causal inference were also explored. The wb-TRS included 144 transcripts. Compared to the lowest tertile, wb-TRS in tertile 3 associated with 8.68-folds (95% confidence interval [CI], 3.51-21.5), and each 1-unit increment associated with 2.57-folds (95% CI, 1.86-3.56) higher risk of nonobese T2D, after adjustments for traditional risk factors. Furthermore, baseline wb-TRS was significantly associated with dynamic changes in average, daytime, nighttime and 24h glucose and HbA1c, and area under the curve of glucose measured in the continuous glucose monitoring during 6-month of intervention. The wb-TRS improved the predicting performance for nonobese T2D in a model with fasting glucose, triglycerides and demographic and anthropometric parameters. Mitch analysis implicated oxidative phosphorylation, cholesterol metabolism and mTORC1 signaling involved in nonobese T2D pathogenesis. Transcriptome-wide Mendelian randomization supported causal effects of gene transcripts such as RAB1A and GCC1-PAX4 on nonobese T2D risk.ConclusionsA whole blood based nonobese T2D associated TRS was validated to predict dynamic changes in glucose metabolism. These findings also suggested several genes and biological pathways that might involve in the pathogenesis of nonobese T2D.