A machine learning study identified a three-gene panel (Pparg-Ptpn11-Ace2) intersecting pyroptosis and myocardial fibrosis pathways in HFpEF, achieving high apparent discriminative performance but with substantial uncertainty requiring prospective validation before any clinical application.
Key Findings
Results
Six candidate genes—Pparg, Ace2, Smad2, Lcn2, Ptpn11, and Sqstm1—were identified as intersecting pyroptosis- and myocardial fibrosis-related genes in HFpEF through integrated transcriptomic analysis and machine learning feature selection.
Two murine cardiac transcriptomic datasets were integrated after batch correction and intersected with curated pyroptosis- and myocardial fibrosis-related gene sets.
Feature selection used univariate logistic regression, SVM-RFE, and LASSO regression.
Complete separation in the ordinary three-gene logistic fit prompted refitting with Firth bias-reduced logistic regression.
Gene Ontology and KEGG analyses linked these genes to SMAD binding, autolysosome/amphisome activity, cellular senescence, efferocytosis, and renin-angiotensin signaling.
Results
The Pparg-Ptpn11-Ace2 Firth model achieved an apparent AUC of 1.0000 in the 45-sample discovery cohort, with bootstrap correction yielding an AUC of 0.9971.
The discovery cohort comprised 45 samples.
Apparent AUC was 1.0000; bootstrap-corrected AUC was 0.9971 (95% percentile interval, 0.9780–1.0000).
Repeated nested cross-validation produced a mean out-of-fold AUC of 0.9286 (2.5%–97.5% percentile interval, 0.8726–0.9700), revealing additional feature-selection uncertainty.
The authors characterize the panel as 'hypothesis-generating rather than a stable diagnostic model.'
Results
No inferred immune-cell fraction differed significantly between HFpEF and control after false-discovery-rate correction, though exploratory correlations were observed.
CIBERSORT was used to estimate immune-cell fractions from transcriptomic data.
No immune-cell fraction reached statistical significance between HFpEF and control after FDR correction.
Exploratory correlations were observed among deconvolved fractions and with Ptpn11 expression.
These immune-cell fraction findings are characterized as exploratory.
Results
HFD/L-NAME-treated mice developed diastolic dysfunction, reduced exercise capacity, pulmonary congestion, cardiomyocyte hypertrophy, and myocardial fibrosis despite preserved systolic function.
The model used a high-fat diet (HFD) combined with N⍵-nitro-L-arginine methyl ester hydrochloride (L-NAME) treatment.
Systolic function was preserved in this model, consistent with HFpEF phenotype.
Myocardial Ptpn11, Pparg, and Ace2 transcripts and their corresponding proteins were reduced in HFD/L-NAME-treated mice.
The model recapitulated key HFpEF features including pulmonary congestion and cardiomyocyte hypertrophy.
Results
In patients with HFpEF, whole-blood PTPN11 mRNA was lower than in controls and was inversely associated with NT-proBNP and C-reactive protein.
PTPN11 mRNA was measured in whole blood from patients with HFpEF and controls.
PTPN11 expression was lower in HFpEF patients compared to controls.
PTPN11 was inversely associated with N-terminal pro-B-type natriuretic peptide (NT-proBNP) and C-reactive protein (CRP).
The authors explicitly note these are 'unadjusted findings' that 'remain exploratory.'
Results
Re-estimation of the gene panel in a human left-ventricular dataset represented exploratory panel-level replication rather than validation of a locked cross-species model.
The human left-ventricular dataset was used for replication analysis.
The authors explicitly distinguish this as 'panel-level replication rather than validation of a locked cross-species model.'
Transportability, selection stability, and clinical relevance are noted as requiring prospective validation.
The cross-species application introduces additional uncertainty beyond the murine discovery findings.
Results
Functional enrichment analyses linked the six candidate genes to SMAD binding, autolysosome/amphisome activity, cellular senescence, efferocytosis, and renin-angiotensin signaling pathways.
Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were used for functional enrichment.
Pathways identified include SMAD binding, autolysosome/amphisome activity, cellular senescence, efferocytosis, and renin-angiotensin signaling.
These pathways suggest mechanistic links between pyroptosis, fibrotic remodeling, and immune dysregulation in HFpEF.
Ace2's involvement implicates renin-angiotensin system dysregulation in the HFpEF fibrotic-pyroptotic intersection.
What This Means
This research suggests that heart failure with preserved ejection fraction (HFpEF)—a common form of heart failure where the heart pumps normally but is too stiff to fill properly—may involve a molecular intersection between two processes: pyroptosis (an inflammatory form of cell death) and myocardial fibrosis (scarring of the heart muscle). Using mouse heart tissue data and machine learning techniques, the researchers identified six genes of interest and narrowed them down to a three-gene combination (involving Pparg, Ptpn11, and Ace2) that could distinguish HFpEF from healthy hearts with very high accuracy in their dataset. Importantly, the same genes showed reduced activity in both a mouse model of HFpEF and in blood samples from human HFpEF patients, and lower PTPN11 levels in patients were associated with worse heart failure markers and higher inflammation.
However, the study has significant limitations that the authors themselves emphasize. The discovery dataset was small (45 samples), and while the three-gene model performed extremely well statistically, the authors explicitly describe it as 'hypothesis-generating rather than a stable diagnostic model.' The near-perfect apparent accuracy likely reflects overfitting to a small dataset, and more rigorous cross-validation revealed lower but still promising performance. The human data analysis was exploratory and the findings were not adjusted for other clinical factors.
This research suggests that the interaction between inflammatory cell death pathways and heart scarring may be important in HFpEF, and that genes like PTPN11, PPARG, and ACE2 could be worth investigating further as potential biomarkers or therapeutic targets. These findings need to be confirmed in much larger, prospectively designed studies before they could have any clinical application in diagnosing or treating HFpEF.
Gao B, Wang G, Xu Z, Peng M. (2026). Immune and inflammatory features associated with pyroptosis- and myocardial fibrosis-related genes in HFpEF: a machine learning study.. Frontiers in immunology. https://doi.org/10.3389/fimmu.2026.1925067