During the 2026 Scientific Sessions of the American Diabetes Association (ADA 2026), the team led by Professor Jianping Weng and Professor Xueying Zheng from the First Affiliated Hospital of the University of Science and Technology of China (Anhui Provincial Hospital) reported seven research advances. These studies cover heterogeneous subtyping of type 1 diabetes mellitus (T1DM), multi-omics profiling of complications in type 2 diabetes mellitus (T2DM), metabolic-inflammatory interaction mechanisms, and new approaches to digital health intervention. The main research summaries are presented below.
Hidden Markov Model-Based Analysis Reveals Social Support Behaviors in an Online T1DM Community
Abstract No.: 1630-P
This study used natural language processing and a hidden Markov model (HMM) to analyze 378,304 posts from 7,803 users in China's largest online T1DM community between 2019 and 2024. The analysis captured how users' behaviors evolved between seeking support and providing support.
The findings showed that 46.2% of users experienced at least one behavioral state transition. Movement from support seeking to support provision was more common than the reverse pattern, and was associated with longer disease duration, longer community engagement, and higher emotional intensity. The study identified three behavioral trajectories: emotion-driven, rational accumulation, and fluctuating. These results suggest that online communities may not only provide peer support, but also help patients gradually become active contributors to chronic disease management.
C-Reactive Protein-Triglyceride Glucose Index and Chronic Disease and Mortality Risk: Evidence from Multinational Cohorts
Abstract No.: 2229-P
The C-reactive protein-triglyceride glucose index (CTI) integrates inflammatory, lipid, and glucose metabolism signals, offering a practical way to assess metabolic inflammatory burden. This study evaluated CTI in multinational cohorts, including CHARLS, NHANES, the UK Biobank, and a hospital-based validation cohort.
Higher CTI was associated with increased risks of multiple chronic diseases and mortality. In CHARLS, higher cumulative CTI was linked to elevated risks of incident hypertension, diabetes, dyslipidemia, stroke, and asthma. In NHANES, higher CTI was associated with all-cause mortality and disease-specific mortality related to cancer, heart disease, lung disease, and diabetes. CTI also showed predictive value for dyslipidemia and mortality risk, with AUC values of 0.74 and 0.85, respectively. These findings support CTI as a simple, integrative marker for chronic disease risk identification.
Multi-omics Atlas of Multisystem Complications in Type 2 Diabetes
Abstract No.: 2283-P
This large-scale multi-omics study mapped molecular signatures associated with 19 major T2DM-related outcomes. By integrating proteomics, metabolomics, and longitudinal cohort analyses, the study examined how plasma proteins and metabolites relate to multisystem complications.
The results identified molecular signals shared across multiple T2DM outcomes, suggesting common biological pathways, while also revealing outcome-specific molecular patterns. Proteomics-based models outperformed conventional clinical models, and combined models delivered the strongest overall prediction performance. A simplified panel of more than 100 proteins retained robust predictive value, pointing to a clinically feasible route for risk prediction and precision intervention in T2DM complications.
Evidence for Age-at-Onset-Related Endotypes in Type 1 Diabetes
Abstract No.: 2371-P
Using DDRTree dimensionality reduction in 894 patients with T1DM, this study examined whether age at onset can capture clinically meaningful heterogeneity in T1DM. The analysis evaluated 13 clinical and immunometabolic phenotypes, including age at onset, metabolic indicators, and islet autoantibodies.
Age at onset emerged as a key phenotype, explaining 64.6% of the variation in the reduced-dimensionality results. The findings supported three age-at-onset-related endotypes: <7 years, 7-13 years, and >=13 years. These groups differed in autoantibody positivity, genetic risk, metabolic characteristics, and long-term cardiovascular risk. The study highlights age at onset as a practical and biologically informative marker for stratifying T1DM heterogeneity.
Deep Learning Identifies Multi-omics Features and Associated Mechanisms in T1DM Onset with Diabetic Ketoacidosis
Abstract No.: 2389-P
This study applied a deep learning-based multi-omics variational autoencoder (MOVE) to analyze newly diagnosed T1DM patients with or without diabetic ketoacidosis (DKA) at onset. The model integrated gut microbiota, serum metabolomics, lipidomics, and proteomics data to capture molecular differences associated with DKA onset.
Distinct multi-omics profiles were observed between the DKA and non-DKA groups. The DKA group showed elevated Bacteroides clarus, 2-arachidonoyl glyceryl ether, and PC (36:2), while the non-DKA group showed higher Burkholderiaceae, D-(+)-proline, LPC (16:0), and CCL28. Integrated and perturbation analyses suggested that gut microbiota may play a key role in DKA onset, with Lachnospira sp000437735 emerging as a potential protective signal. The study offers a new molecular perspective on T1DM onset heterogeneity.
Multivariate GWAS Reveals Shared Pleiotropic Genetics Between Obesity and MASLD
Abstract No.: 2615-P
Obesity and metabolic dysfunction-associated steatotic liver disease (MASLD) frequently coexist, but their shared genetic architecture remains incompletely understood. This study integrated GWAS summary data, LDSC, condFDR/conjFDR, MTAG, GTEx eQTL data, and SMR analysis to map genetic overlap and tissue-specific regulatory mechanisms.
The analysis confirmed significant genetic correlations between obesity-related traits and MASLD. It identified 31 independent MASLD-related loci, including 15 novel loci, as well as 184 obesity-related loci and 23 shared obesity-MASLD loci. Cross-method integration highlighted genes including ADCY3, FTO, TRIM66, COBLL1, L3MBTL3, VPS53, CDK6, ANAPC13, and TBX15 as potential pleiotropic signals. These findings deepen understanding of obesity-related MASLD and may inform future risk stratification and intervention target discovery.
Exosomal miR-223-3p Targeting NLRP3-Related Neuroinflammation May Serve as a Biomarker for Cognitive Impairment in T2DM
Abstract No.: 1117-OR
This study explored the role of miR-223-3p in diabetes-associated cognitive impairment, focusing on NLRP3-mediated neuroinflammation. In diabetic mouse models, reduced hippocampal miR-223-3p expression was accompanied by cognitive decline, microglial activation, and increased NLRP3 and pro-inflammatory cytokines.
Mechanistic experiments showed that miR-223-3p directly regulates NLRP3: knockdown promoted neuroinflammation, while overexpression reduced it. In patients with T2DM, plasma exosomal miR-223-3p levels were significantly lower in those with mild cognitive impairment (MCI) than in cognitively normal patients, and were associated with cognitive and immediate memory performance. ROC analysis showed 71.4% specificity and 68.7% sensitivity for identifying MCI. These findings suggest circulating exosomal miR-223-3p as a potential biomarker for cognitive impairment in T2DM.
Commentary by Professor Jianping Weng
At ADA 2026, our team reported seven studies. Although these studies appear to span multiple areas--including T1DM heterogeneity, multi-omics analysis of T2DM complications, metabolic-inflammatory interactions, and digital health interventions--they are connected by a single overarching theme: diabetes care is accelerating from experience-based diagnosis and treatment of a "single disease" toward a new precision medicine paradigm built on multidimensional data integration.
What I find particularly important is that these studies collectively point to one core proposition: heterogeneity. Whether it is the age-at-onset endotypes revealed by DDRTree, the multi-omics features of DKA onset decoded by the MOVE model, or CTI-based risk stratification for chronic disease comorbidity, all are answering the same question: behind the same diagnostic label of "diabetes" lie markedly different molecular mechanisms, disease trajectories, and responses to intervention. In the past, we relied on blood glucose as a single indicator to manage patients in a one-size-fits-all manner. In the future, we must establish a multidimensional stratification system that incorporates the genome, proteome, metabolome, microbiome, and even digital behavioral phenotypes.
Equally thought-provoking is the integration and penetration of digital and intelligent technologies. The use of hidden Markov models to dynamically capture support behaviors in online communities is not only a study of digital phenotyping, but also a sign that chronic disease management is moving beyond hospital walls into a "community-home-digital space" continuum. When multi-omics atlases meet deep learning, and when clinical cohort data meet real-world behavioral data, we are in fact constructing an early prototype that approaches a "virtual artificial pancreas" and a "metabolic digital twin." This is precisely the direction of clinical translation that our team aims to advance next.
From a broader perspective, diabetes research is undergoing three major transitions: from a single-disease focus to a panvascular metabolic comorbidity network; from endpoint treatment to life-course risk warning; and from evidence based on Western populations to solutions tailored to Chinese population characteristics. The CTI multinational cohort validation, obesity-MASLD genetic overlap analysis, and exploration of early biomarkers for cognitive impairment reported at this ADA meeting all provide foundational evidence for these transitions.
I have always believed that diabetes research in China should not merely follow international guidelines. When our T1DM large models, artificial pancreas algorithms, and multi-omics prediction panels begin to generate "Chinese parameters," we are in fact participating in defining the global standards for the next generation of diabetes diagnosis and treatment. These seven ADA achievements are both a milestone summary of our team's previous work and a stepping stone toward the era of "computational precision therapy.”
Team Research Directions and Achievements
Professor Jianping Weng's team has long focused on precision diagnosis and treatment and clinical translation in diabetes, forming a systematic research framework around the following three dimensions:
First, exploring disease mechanisms using digital and intelligent technology tools. The team has established China's first intelligent management and big data center for T1DM, built the world's first T1DM-specific large model, and developed and validated a new artificial pancreas algorithm and a physics-informed personalized linear model control system (authorized patents include ZL202510565116.8). By integrating artificial intelligence technologies such as deep learning, hidden Markov models, and multi-omics variational autoencoders (MOVE), the team has built a digital research framework spanning molecular mechanism analysis to clinical risk warning, promoting the transformation of diabetes prevention and treatment toward intelligent and precision approaches.
Second, scientific research on diabetes complications and comorbidities. Based on large-scale prospective cohorts, the team has systematically conducted multi-omics atlas studies of multisystem complications in T2DM, integrating proteomics, metabolomics, and lipidomics data to construct molecular prediction panels for 19 related outcomes. Through DDRTree dimensionality reduction analysis, the team revealed age at onset as a key phenotype of T1DM heterogeneity and validated differences in cardiovascular outcome risk across endotypes. The team has also conducted multinational cohort research on the C-reactive protein-triglyceride glucose index (CTI) and risks of chronic diseases and mortality; explored shared pleiotropic genetic loci and tissue-specific regulatory mechanisms between obesity and metabolic dysfunction-associated steatotic liver disease (MASLD); and revealed the molecular mechanism by which exosomal miR-223-3p participates in diabetes-associated cognitive impairment through the NLRP3 pathway.
Third, serving clinical practice. The team's research findings have been cited by 35 international guidelines/consensus statements and authoritative textbooks in China and abroad, and the team has led the development of multiple clinical guidelines and group standards. Based on the "beta-cell rest" hypothesis, the team established a new paradigm for rational insulin use. A 24-year national multicenter cohort study demonstrated that early insulin therapy reduced the risk of stroke by 31% and the risk of hospitalization for heart failure by 28%. The team established a Panvascular Disease Management Center to promote integrated metabolic-vascular management; built an online health community behavior evolution model to provide evidence for digital health interventions in chronic diseases; and developed clinically feasible risk prediction and precision intervention frameworks to directly support clinical decision-making in high-risk population screening, risk stratification, and early intervention.