Machine Learning Integrates Bulk and Single-Nucleus RNA Sequence to Explore Apoptosis-Related Gene in Myocardial Infarction.

Fibroblasts demonstrated significantly elevated apoptotic activity compared with other cardiac cell types following myocardial infarction.

• Single-nucleus RNA-seq dataset GSE270788 was used for single-cell level analysis of apoptotic signaling in MI progression.
• AUCell scoring was employed to assess apoptosis-related gene (ARG) activity across different cardiac cell populations.
• 397 apoptosis-related genes were downloaded from GeneCards with a relevance score >7.
• Dramatic variation in ARG activity was observed across cardiac cells after MI, with fibroblasts identified as the dominant cell type for apoptotic activity post-MI.

TNFRSF12A was identified as the major molecule regulating ARG activation in myocardial infarction through integration of single-nucleus and bulk RNA sequencing data.

• Integration combined snRNA-seq (GSE270788) and bulk RNA-seq (GSE21610) datasets.
• Candidate genes were identified through AUCell scoring and correlation analysis.
• Three machine learning algorithms—Random Forest, Boruta, and LASSO—were employed to refine and identify hub genes.
• TNFRSF12A was confirmed as the key hub gene regulating ARG activation in MI.

Experimental validation confirmed upregulation of TNFRSF12A in both an MI animal model and TGF-β-induced NIH3T3 cells.

• Animal experiments were used to validate hub gene expression in an in vivo MI model.
• Cell experiments used TGF-β-induced NIH3T3 fibroblast cells as an in vitro model.
• TNFRSF12A expression was upregulated in both experimental systems, supporting the computational findings.
• The study proposes TNFRSF12A has potential as a diagnostic biomarker and therapeutic target in MI.

A multiomics machine learning pipeline integrating 397 apoptosis-related genes from GeneCards was used to systematically identify candidate hub genes in MI.

• Apoptosis-related genes were sourced from GeneCards with a relevance score cutoff >7, yielding 397 ARGs.
• AUCell scoring and correlation analysis were applied to identify candidate genes from single-nucleus data.
• Random Forest, Boruta, and LASSO algorithms were sequentially employed to refine the candidate gene list down to hub genes.
• This integrative approach bridged single-cell resolution with bulk transcriptomic findings to characterize apoptotic signaling in MI.