Macrophage Polarization Signatures from RNA-seq Identify Therapeutic Exosomal miRNA Payloads for Muscle Regeneration: An Open, Reproducible Computational Framework

Abstract

Background. Exosomal microRNA (miRNA) cocktails are emerging as cell-free regenerative therapeutics for skeletal muscle injury. Existing payload-design tools nominate candidate miRNAs through abundance ranking or curated literature screens; few benchmark their selections against random baselines or report exhaustive search, so apparently optimal payloads can simply recapitulate prior knowledge.Methods. We re-analysed GSE212372, a 2×2 factorial RNA-seq study of mouse tibialis-anterior macrophages (n=12; anti-TNFα neutralising antibody × Ifngr1 wt/null vs. null/null). Differential expression was performed with pyDESeq2 with Benjamini–Hochberg correction. Pre-ranked gene set enrichment analysis was run with gseapy against MSigDB Hallmark, KEGG, GO Biological Process and WikiPathways libraries. Leading-edge genes from pathways at FDR<0.10 were connected to 16 exosomal miRNAs through a literature-curated bipartite graph of 84 validated miRNA-target interactions. Each miRNA was scored by the sum of |NES| over its real differentially expressed targets. Payload-of-four selection was evaluated by greedy top-K, exhaustive enumeration of all (16¦4)=1820 combinations, and comparison against 1000 random combinations using z-score and one-sample t-test. Results. Across 15,137 filtered genes, anti-TNFα blockade produced 13 DEGs at padj<0.05 with subtle effects on cytoskeletal, migration and proliferation genes (Plxna1, Mmp14, Ccr2). Ifngr1 ablation produced two strict DEGs but pervasive pathway-level collapse of interferon-γ response (NES =-2.35, FDR =0.024). GSEA recovered 35/50 Hallmark pathways at FDR<0.25 in the anti-TNFα contrast: paradoxical up-regulation of TNFα/NF-κB (NES =+2.24), IL-6/JAK/STAT3 (+1.84) and interferon-α response (+2.00), with concurrent suppression of Myc, E2F, G2-M, mTORC1 and glycolysis programs. Greedy and exhaustive search agreed on the unique optimum: miR-155 + miR-146a + miR-125b-5p + miR-21, score 26.86, z=3.72, p<10^-300 versus random baseline, 100th percentile. Conclusion. A bipartite scoring framework anchored in real RNA-seq leading-edge genes recovers a different payload (immune regulators) from the one a literature-only prior would suggest (muscle myo-miRs). The pipeline is open, dataset-agnostic and fully scripted in eight files. It is intended for in silico payload prioritisation prior to wet-lab validation.