Background: Obesity exhibits heterogeneity in physiological states and metabolic phenotypes cross individuals. A recent study identified a proteomics signature of obesity which better captures such heterogeneity than classical Body Mass Index (BMI). Aims: We aimed to create a proteomics-inferred BMI signature and investigate whether the proteomic-inferred BMI was associated with incident Type 2 Diabetes (T2D). Methods: This study included 34,317 adults in the UK Biobank who were free of T2D and had complete data on proteomics metrics assessed by antibody-based Olink assay. Sixty-seven pre-identified proteins were available and used to create the proteomic BMI signature. Cox proportional models were used to estimate the association between proteomic BMI signature and incident T2D. Results: The proteomic BMI signature was strongly correlated with actual BMI (Pearson correlation coefficients=0.75, P <0.001). During a mean follow-up of 13.2 years, a total of 1,651 T2D events were documented. We observed that per standard deviation increase in proteomic BMI signature was significantly associated with a 250% higher risk of T2D (HR, 2.50, 95%CI, 2.37, 2.64). Compared with the lowest Quartile, the HR (95%CI) for the Quartile 2 to the Quartile 4 of proteomic BMI signature were 2.06 (1.61, 2.64), 3.56 (2.83, 4.48), and 10.0 (8.08, 12.4), respectively ( P -trend<0.001). Also, we observed that the associations between proteomic BMI signature and incident T2D were stronger in women than men ( P for interaction < 0.001). Compared with the lowest Quartile, the HR (95%CI) for highest Quartile of proteomic BMI signature were 16.63 (9.75, 28.36) for women and 8.06 (6.37, 10.19) for men. Conclusions: Higher proteomics-inferred BMI was significantly associated with higher risk of incident T2D, and such association was stronger in women than men. Our results suggest that proteomic BMI signature may serve as a surrogate metabolic indicator to better understand the mechanism underlying the association of BMI and T2D.