<p>Diabetes has become a significant global public health challenge, and prediabetes (preD), as the earliest detectable stage of glucose metabolism disorder, constitutes a critical window for disease intervention. Currently, China has not established diagnostic criteria for HbA1c in preD, and there is limited understanding of the molecular mechanisms at the critical transition points during disease progression. This study employs a refined stratification strategy for HbA1c at intervals of 0.3%, integrating the Dynamic Network Biomarker (DNB) theory with high-throughput proteomics technology, combined with longitudinal cohort validation and machine learning methods, to systematically depict the disease progression characteristics, spanning from a healthy physiological state through the prediabetic stage and ultimately to the onset of type 2 diabetes (T2D). This cross-sectional cohort consisted of 110 subjects, divided into three groups: healthy controls (20), prediabetic (50), and type 2 diabetic (40). Of these, 28 went on to complete a 2-year longitudinal follow-up assessment. The results indicate that the HbA1c range of 6.0–6.2% (D stage) meets the three criteria of the DNB theory (high inter-molecular coherence, decoupling from external networks, and enhanced volatility), suggesting that this stage may represent a critical state for the transition from preD to T2D. A total of 37 core DNB proteins were identified through proteomics analysis, with the majority of functional annotations focused on biological processes involving extracellular matrix remodeling, inflammatory response, and insulin signaling pathways. In the longitudinal follow-up validation, the predictive efficacy of the ADAM10 protein for clinical outcomes was significant (AUC = 0.768), outperforming traditional indicators like HbA1c, demonstrating its potential value as a novel biomarker. There are several limitations in the present investigation. With the relatively small number of subjects, the short follow-up period, and the single center, single ethnic study design, the findings have limited generalizability. Additionally, the lack of functional validation of some of the important molecular targets is a methodological limitation that should be recognized. Therefore, the identified critical transition window and the clinical predictive value of DNB molecules need to be further validated through large-scale, multi-center, long-term cohort studies to obtain more reliable evidence-based medical evidence, providing a scientific basis for the precise diagnosis and early intervention of preD in the Chinese population.</p>