Integrated clinical multi-omics identifies shared host transcriptomic and gut microbiota-metabolite signatures of diabetic microvascular complication multimorbidity

Background Diabetic microvascular complications frequently co-occur in patients with type 2 diabetes, yet the gut microbiota, circulating metabolic features, and shared host signals associated with diabetic microvascular complication multimorbidity remain unclear. This study aimed to characterize clinical multi-omics signatures associated with this multimorbidity phenotype. Methods This cross-sectional exploratory study included 80 patients with type 2 diabetes, comprising 41 patients with multiple diabetic microvascular complications and 39 patients without diabetic microvascular complications, together with 26 healthy controls for microbiome comparison. Faecal 16S rRNA sequencing, PICRUSt2-based functional prediction, untargeted metabolomics, and routine clinical indicators were analysed. Public GEO transcriptomic datasets for diabetic nephropathy, diabetic retinopathy, and diabetic peripheral neuropathy were integrated with GutMGene to prioritize putative shared host-microbiota-metabolite links. Multivariable models, Spearman correlation analysis, Elastic Net/LASSO regression, random forest, and XGBoost were applied. Results Gut microbial structure differed among groups, and 12 genera were significantly depleted in the multiple-complication group after covariate adjustment. PICRUSt2 identified 1,539 differential KEGG Orthology features, including 802 increased and 737 decreased features. Among 970 detected metabolites, 50 remained associated with multiple-complication status after differential screening and multivariable adjustment. Public transcriptomic analysis identified 33 shared candidate genes across diabetic nephropathy, diabetic retinopathy, and diabetic peripheral neuropathy, with CD163 consistently upregulated. GutMGene mapping and cohort microbiome results prioritized Blautia , Roseburia , and Collinsella for network analysis. These three genera showed 74 significant negative correlations with differential metabolites. Integrated diagnostic modelling suggested that combined microbiome, metabolomic, and routine clinical features showed good internal discriminative performance. A parsimonious four-feature signature comprising Blautia , Roseburia , D-(+)-Maltose, and Dihydrothymine was further derived, with supportive random forest and XGBoost analyses yielding AUCs of 0.986 and 0.965, respectively. Conclusions This study suggested that diabetic microvascular complication multimorbidity was associated with depletion of anaerobic fermentation-related gut taxa, circulating metabolic disturbance, and CD163-related shared inflammatory transcriptomic signals. These findings provide exploratory host-microbiota-metabolite signatures for future longitudinal and mechanistic validation.