DEEP VEIN THROMBOSIS (DVT) may be identified more accurately through a combination of gut microbiome profiles and clinical data, with a machine learning model demonstrating significantly improved detection compared to clinical assessment alone.
Microbial Differences Linked to Deep Vein Thrombosis
Researchers analysed the stool samples from 58 patients with DVT, 56 patients with coronary artery disease (CAD), and 500 healthy individuals (controls), to investigate whether microbial signatures could help distinguish DVT cases from non-cases.
The study used RNA gene sequencing to characterise gut microbial communities at species level resolutions. Patients with DVT exhibited significantly lower microbial richness, compared with the healthy control individuals and patients with CAD.
At species level, Escherichia coli, Klebsiella pneumoniae, and Fusobacterium varium were among the most prolific organisms in patients with DVT, compared with healthy controls. In contrast, Bacteroides coprocola, Bifidobacterium pseudocatenulatum, and Collinsella aerofaciens were more abundant in healthy participants.
Following screening and feature selection, the researchers then reduced 95 candidate microbial features to 10 key variables for AI classification modelling.
Machine Learning Model Outperforms Clinical Assessment
Researchers found that a machine learning model, integrating microbial and clinical data, achieved stronger performance than a ‘clinical-only’ model.
The integrated model delivered a receiver operating characteristic (ROC) area under the curve (AUC) of 0.947 (95% CI: 0.870–0.991) compared with 0.874 (95% CI: 0.794–0.941) for the clinical model. Precision recall AUC was also higher at 0.793 (95% CI: 0.602–0.931), as opposed to 0.497 (95% CI: 0.274–0.724) for the clinical model.
The integrated approach also achieved better balanced accuracy (sensitivity and specificity), reaching 91.6% (95% CI: 84.7%–96.2%) compared with 79.3% (95% CI: 69.2%–87.5%). Microbial features represented eight of the ten most influential predictors identified by the model, highlighting their potential value as biological markers.
Metabolic Pathways Provide Biological Insight
The analysis suggested that microbial signatures associated with DVT were enriched for enzymatic and energy metabolism pathways, such as vitamin K₂ biosynthesis. Whereas, healthy controls were enriched with metabolic coenzyme pathways, associated with metabolic and cellular homeostasis.
While these findings support the potential of gut microbiome profiling to serve as a disease biomarker, further validation will be needed before this approach can be adopted in routine clinical practice.
Reference
Lu CR et al. Gut microbiome signatures discriminate deep vein thrombosis through machine learning and metabolic analysis. Sci Rep. 2026; DOI: 10.1038/s41598-026-55650-2.
Featured image: troyanphoto on Adobe Stock
- Author:
