Regularized logistic regression-based gene selection for functional analysis of heterogeneous scRNA-seq macrophage populations in muscle aging and regeneration.

DisTwiner, a subgroup-aware regularized logistic regression method, selects macrophage genes with more heterogeneous fold-change and correlation patterns across subpopulations while matching Elastic Net classification performance.

• Why it matters: Standard feature selection can miss subgroup-specific transcriptional structure, which matters in complex tissues where distinct cell states contribute differently to aging and regeneration.
• What they did: DisTwiner extends twiner correlation-based regularization for binary classification with multiple annotated groups by computing pairwise angular scores between feature-correlation profiles and using them as feature-specific weights in Elastic Net logistic regression.
• The result: In simulated data, weighting choices affected penalty dispersion (minimum-based exponential weighting greatest), and in young vs aged regenerating murine skeletal muscle macrophages, DisTwiner matched Elastic Net and exceeded Random Forest, while enrichment highlighted aging-associated interferon-response signatures in specific macrophage populations.