Machine Learning Approach to Identify Gut Microbiota Biomarkers in Patients With ST-Elevation Myocardial Infarction Presenting With Primary Ventricular Tachyarrhythmias

BackgroundRisk stratification for primary ventricular tachycardia/ventricular fibrillation (VT/VF) in patients with ST-elevation myocardial infarction (STEMI) remains limited. We aimed to identify gut microbiome biomarkers and microbial metabolic pathways associated with primary VT/VF in STEMI using machine learning (ML). MethodsStool samples from 33 STEMI patients (7 with primary VT/VF and 26 without VT/VF) underwent full-length 16S rRNA sequencing during the acute ([&le;]7 days) and recovery (30-60 days) phases. An ML pipeline integrating taxonomic features identified reproducible features associated with primary VT/VF. Core biomarkers were defined as taxa demonstrating both univariate significance (false discovery rate [FDR] <0.05) and multivariate classification importance. Functional pathways and enzymes were predicted using PICRUSt2. ResultsFive acute-phase taxa were reproducibly associated with primary VT/VF: Clostridium aldenense, Enterocloster bolteae, bacterium NLAE-zl-G101, Alistipes shahii, and Roseburia sp. (all FDR <0.05). A reduced-feature support vector machine model achieved an area under the receiver operating characteristic curve of 0.846 (95% confidence interval, 0.676-0.984), with 85.7% sensitivity and 76.9% specificity. Differential abundance analysis showed enrichment of Bacteroides fragilis, Bacteroides thetaiotaomicron, and E. bolteae. Functional analysis demonstrated enrichment of fucose degradation and purine catabolism, with increased predicted abundances of L-fucose mutarotase and xanthine dehydrogenase. ConclusionsPrimary VT/VF in STEMI was associated with a distinct gut microbiome-metabolic signature characterized by Lachnoclostridium-related taxa, persistent Bacteroides enrichment, and enhanced predicted fucose and purine metabolism. These exploratory findings warrant validation in larger independent cohorts and mechanistic studies. Clinical PerspectiveWhat Is New? O_LIIn this exploratory cohort of 33 patients with STEMI, including 7 with primary VT/VF, an integrated microbiome and machine-learning framework identified five acute-phase microbial taxa associated with primary VT/VF and a coordinated predicted metabolic signature involving fucose and purine pathways. C_LI What Are the Clinical Implications? O_LIBecause stool sampling occurred after the arrhythmic event and the VT/VF subgroup was small, these findings should be considered hypothesis-generating rather than predictive; independent prospective validation is required before microbiome-derived markers can be considered for clinical risk stratification. C_LI