Multimodal Exploratory Analysis of Persistent Diabetic Macular Edema: Integrating Aqueous Humour Metabolomics, Clinical and OCT Features
5 affiliations
- 1Department of Ophthalmology, Hubei Key Laboratory of Otolaryngologic and Ophthalmic Diseases Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology Wuhan China
- 2Department of Ophthalmology West China Second University Hospital, Sichuan University Chengdu Sichuan China
- 3Department of Nutrition and Food Hygiene, School of Public Health Suzhou Medical College of Soochow University Suzhou Jiangsu China
- 4Department of Epidemiology and Biostatistics, School of Public Health Tongji Medical College, Huazhong University of Science and Technology Wuhan China
- 5Section of Statistical Multi‐Omics, Discipline of Clinical Medicine, School of Bioscience and Medicine University of Surrey Guildford UK
Abstract
Purpose: This exploratory study integrates aqueous humour metabolomic data with clinical and imaging features to examine candidate biomarkers and characterize multimodal patterns associated with persistent diabetic macular edema (pDME).
Methods: In this prospective cohort study, DME patients were treated with monthly anti‐VEGF injections over 3 months and categorized as pDME and non‐pDME groups based on central subfield thickness and best‐corrected visual acuity. We then carried out a nested case–control comparison of these groups, performing untargeted metabolomic analysis on aqueous humour samples. Comparison of clinical data, OCT parameters, and metabolites was performed. Logistic regression models were constructed to explore associations between multimodal features and pDME in the metabolomics subset.
Results: A total of 103 patients (146 eyes) were finally included in the analysis. The pDME patients had significantly higher BMI (25.20 vs. 23.27, p = 0.004) and triglyceride (1.80 vs. 1.50, p = 0.003) levels when compared with non‐pDME patients. A higher rate of disorganization of retinal inner layers (DRIL, p = 0.026) and vitreoretinal interface abnormality (p = 0.009), but a lower rate of SRF (p = 0.019), was found in the pDME group. Metabolomics analysis revealed 405 differentially expressed metabolites, and purine metabolism (p = 0.001, Q = 0.024) and DNA repair pathway (p < 0.001, Q = 0.013) were significantly enriched. Purine score and DNA repair score were both significantly higher in the pDME group compared with the non‐pDME group (p < 0.001). ROC analysis demonstrated metabolomic markers, particularly the purine score, showed stronger group‐wise separation for pDME than clinical and OCT features (LOOCV AUC 0.963 vs. 0.716 and 0.505). Additionally, vitreoretinal interface abnormality was associated with purine metabolism (OR = 5.820, 95% CI: 1.136–29.811) and DNA repair score (OR = 3.790, 95% CI: 1.096–13.109), while DRIL (OR = 0.214, 95% CI: 0.046–0.990) was associated with DNA repair score.
Conclusions: PDME is characterized by elevated BMI, serum triglyceride, low SRF incidence, but a higher rate of DRIL and vitreoretinal interface abnormality. Purine metabolism and DNA repair pathways were associated with treatment response and correlated with OCT structural changes. Metabolomic signatures, particularly the purine score, emerged as promising candidate markers associated with pDME in this exploratory cohort, warranting further validation in independent studies. A novel mechanism of pDME, the purine metabolism–DNA repair axis, was proposed and needs further verification.
Related
- Associations between serum lactate and renal dysfunction in diabetic kidney disease: An exploratory analysis integrating animal experiments, clinical cohorts, and bioinformatic approaches
- Exploratory Integration of Metabolomics and Machine Learning Identifies Candidate Signatures of Spontaneous Preterm Birth
- Comprehensive Multimodal Diagnostics and Rehabilitation in Persistent Post-Concussion Syndrome Following Motor Vehicle Collision
- Metabolic Profiling, Antioxidant Activity, and Mechanistic Insights of Chicory from Northern and Southern China: Integrating Metabolomics, Machine Learning, Molecular Docking, Molecular Dynamics Simulations, and In Vitro Assays
- The new frontier of newborn screening: integrating genomics, metabolomics, and data science for inborn errors of metabolism
BibTeX
@article{tao2026multimodal,
title = {{Multimodal Exploratory Analysis of Persistent Diabetic Macular Edema: Integrating Aqueous Humour Metabolomics, Clinical and OCT Features}},
author = {Yunhan Tao and Huimin Yu and Yunxi Ma and Mingzhu Yuan and Guochong Chen and Jun Liu and Xi Chen and Xufang Sun},
journal = {Diabetes, Obesity \& Metabolism},
year = {2026},
month = sep,
doi = {10.1111/dom.71173},
url = {https://doi.org/10.1111/dom.71173},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Multimodal Exploratory Analysis of Persistent Diabetic Macular Edema: Integrating Aqueous Humour Metabolomics, Clinical and OCT Features
AU - Yunhan Tao
AU - Huimin Yu
AU - Yunxi Ma
AU - Mingzhu Yuan
AU - Guochong Chen
AU - Jun Liu
AU - Xi Chen
AU - Xufang Sun
PY - 2026
DA - 2026/09/20
JO - Diabetes, Obesity & Metabolism
DO - 10.1111/dom.71173
UR - https://doi.org/10.1111/dom.71173
AB - Purpose: This exploratory study integrates aqueous humour metabolomic data with clinical and imaging features to examine candidate biomarkers and characterize multimodal patterns associated with persistent diabetic macular edema (pDME). Methods: In this prospective cohort study, DME patients were treated with monthly anti‐VEGF injections over 3 months and categorized as pDME and non‐pDME groups based on central subfield thickness and best‐corrected visual acuity. We then carried out a nested case–control comparison of these groups, performing untargeted metabolomic analysis on aqueous humour samples. Comparison of clinical data, OCT parameters, and metabolites was performed. Logistic regression models were constructed to explore associations between multimodal features and pDME in the metabolomics subset. Results: A total of 103 patients (146 eyes) were finally included in the analysis. The pDME patients had significantly higher BMI (25.20 vs. 23.27, p = 0.004) and triglyceride (1.80 vs. 1.50, p = 0.003) levels when compared with non‐pDME patients. A higher rate of disorganization of retinal inner layers (DRIL, p = 0.026) and vitreoretinal interface abnormality (p = 0.009), but a lower rate of SRF (p = 0.019), was found in the pDME group. Metabolomics analysis revealed 405 differentially expressed metabolites, and purine metabolism (p = 0.001, Q = 0.024) and DNA repair pathway (p < 0.001, Q = 0.013) were significantly enriched. Purine score and DNA repair score were both significantly higher in the pDME group compared with the non‐pDME group (p < 0.001). ROC analysis demonstrated metabolomic markers, particularly the purine score, showed stronger group‐wise separation for pDME than clinical and OCT features (LOOCV AUC 0.963 vs. 0.716 and 0.505). Additionally, vitreoretinal interface abnormality was associated with purine metabolism (OR = 5.820, 95% CI: 1.136–29.811) and DNA repair score (OR = 3.790, 95% CI: 1.096–13.109), while DRIL (OR = 0.214, 95% CI: 0.046–0.990) was associated with DNA repair score. Conclusions: PDME is characterized by elevated BMI, serum triglyceride, low SRF incidence, but a higher rate of DRIL and vitreoretinal interface abnormality. Purine metabolism and DNA repair pathways were associated with treatment response and correlated with OCT structural changes. Metabolomic signatures, particularly the purine score, emerged as promising candidate markers associated with pDME in this exploratory cohort, warranting further validation in independent studies. A novel mechanism of pDME, the purine metabolism–DNA repair axis, was proposed and needs further verification.
SN - 1462-8902
SN - 1463-1326
ER -