Systems-Level Transcriptomics Maps Multilevel Remodeling and Pathway-Selective Translational Alignment Across Murine Models of Cardiometabolic HFpEF

Abstract

Background. Murine cardiometabolic heart failure with preserved ejection fraction (HFpEF) models are selected and generalized from pathway-level transcriptomic summaries, but how reproducibly those signals hold across laboratories and species is unclear. Methods. I re-analyzed ventricular RNA-sequencing, ventricular single-cell data and subgroup-resolved human HFpEF proteomics for db/db plus aldosterone (db/db+Aldo) and high-fat-diet plus L-NAME (HFD+L-NAME) models, quantifying agreement at gene and pathway level with bootstrap intervals, and applied the identical pipeline to five mouse cardiac model pairs, each two independent deposits at matched sample size. Results. For HFD+L-NAME, agreement across 23,838 genes was rho = 0.08 (95% CI 0.06-0.09), rising to 0.45 among 349 differentially expressed genes, whereas the eight-pathway summary gave rho = 0.71 with an interval spanning -0.04 to 0.98. A transverse aortic constriction positive control at matched sample size reached rho = 0.54 at gene level and 0.90 at pathway level, with 2,241 genes significant at four mice per arm. Across five model pairs gene-level agreement ranged nearly nine-fold (0.08 to 0.69) while pathway-level agreement never fell below 0.71. Cross-species concordance was confined to oxidative phosphorylation (142 genes). Independent single-cell projection placed the myofibroblast signature in mural cells rather than fibroblasts. Conclusions. Apparent reproducibility in murine HFpEF transcriptomics is governed by the resolution at which it is measured. Pathway-level correlations were uniformly high, including for models that do not reproduce gene by gene, and cannot certify reproducibility. Reporting gene-level agreement, feature support and interval estimates turns public omics into usable evidence for model selection and 3Rs-informed design.