This observational study analyzed stool samples from 33 patients hospitalized with ST-elevation myocardial infarction (STEMI), of whom 7 presented with primary ventricular tachyarrhythmias (primary VT/VF) and 26 did not. Samples were collected in two time windows: the acute phase (≤7 days after presentation) and the recovery phase (30–60 days). All participants provided written informed consent and the study was approved by the Institutional Review Board of China Medical University Hospital (CMUH111-REC3-195).
Full-length 16S rRNA sequencing was performed on stool specimens. Taxonomic features derived from sequencing were fed into a machine learning (ML) pipeline designed to identify reproducible features associated with primary VT/VF. The pipeline integrated univariate differential abundance testing with multivariate classification to prioritize taxa that were both statistically significant and informative for prediction. Predicted functional profiles were generated using PICRUSt2 to infer metabolic pathways and enzyme abundances from taxonomic composition.
Core biomarkers in this analysis were defined as taxa that met two criteria: (1) univariate significance with a false discovery rate (FDR) < 0.05, and (2) importance in multivariate classification within the ML pipeline. This dual requirement was used to emphasize reproducibility and potential utility for distinguishing patients with primary VT/VF from those without.
Five taxa identified in the acute phase were reproducibly associated with primary VT/VF, each meeting the FDR < 0.05 threshold and showing multivariate importance. These taxa were Clostridium aldenense, Enterocloster bolteae, bacterium NLAE-zl-G101, Alistipes shahii, and Roseburia sp. In addition, differential abundance analysis indicated enrichment of Bacteroides fragilis, Bacteroides thetaiotaomicron, and E. bolteae in patients with primary VT/VF.
The authors characterize the gut signature linked to primary VT/VF as notable for Lachnoclostridium-related taxa and persistent enrichment of certain Bacteroides species. The specific taxa above were observed during the acute-phase sampling; the report does not provide additional temporal dynamics beyond the acute and recovery sampling windows other than noting these acute-phase associations.
A reduced-feature support vector machine (SVM) model built from the reproducible taxa achieved an area under the receiver operating characteristic curve (AUC) of 0.846, with a 95% confidence interval of 0.676 to 0.984. Reported performance metrics for this classifier included 85.7% sensitivity and 76.9% specificity for discriminating patients with primary VT/VF from those without in this dataset. The study frames this model as an exploratory classifier derived from a small cohort.
Beyond the five core acute-phase taxa, the analysis highlighted persistent enrichment of Bacteroides species, including B. fragilis and B. thetaiotaomicron, in the group with primary VT/VF. The report notes enrichment of E. bolteae by differential abundance testing as well. These observations support a pattern of community composition differences between STEMI patients who present with primary VT/VF and those who do not.
Functional prediction using PICRUSt2 suggested that microbial metabolic functions related to fucose degradation and purine catabolism were enriched in patients with primary VT/VF. Predicted enzyme-level increases included L-fucose mutarotase and xanthine dehydrogenase. These findings are derived from inferred metagenomic content rather than direct metabolomic or metatranscriptomic measurement.
The authors describe these results as exploratory and emphasize the need for validation in larger, independent cohorts and for mechanistic studies to establish causality. The cohort size was limited (33 patients, 7 with primary VT/VF), and functional inferences were predictions rather than direct measurements. The authors declared no competing interests. Ethical approval and written informed consent were obtained according to the Declaration of Helsinki. The data supporting the findings are available from the corresponding author upon reasonable request, subject to institutional and ethical restrictions.