Sepsis reporting signals associated with endothelin receptor antagonists and IFITM3-Centered interferon-responsive monocyte features: a pharmacovigilance and transcriptomic study.

Background: Whether pulmonary arterial hypertension (PAH)-targeted therapies, particularly endothelin receptor antagonists (ERAs), are associated with disproportionate sepsis reporting in real-world pharmacovigilance data remains insufficiently explored. The monocyte transcriptional states that characterize sepsis-related immune dysregulation and may provide biological context for such reporting signals are also incompletely defined. Methods: We constructed an integrated, hypothesis-generating analytical framework incorporating: (i) FDA Adverse Event Reporting System (FAERS) disproportionality analysis coupled with XGBoost-based modeling for pharmacovigilance signal detection; (ii) single-cell RNA sequencing (scRNA-seq) analysis of peripheral blood mononuclear cells with intercellular communication inference; (iii) weighted gene co-expression network analysis (WGCNA) and cytoHubba-based topological prioritization, combined with an ensemble machine learning framework for diagnostic signature construction; and (iv) molecular docking and 100-nanosecond all-atom molecular dynamics (MD) simulation. Results: FAERS analysis identified Maitentan and ambrisentan as PAH-targeted therapies with positive reporting signals for the MedDRA Preferred Term "Sepsis," with adjusted reporting associations persisting after adjustment for available demographic variables. Sex-stratified analysis showed marked heterogeneity in reporting signals, although these findings may be influenced by the sex distribution of PAH populations and other unmeasured confounders. ScRNA-seq resolved seven monocyte subpopulations, among which the interferon-responsive Mono_IFN subset-marked by IFIT1, ISG15, and IFITM3 expression-occupied a signaling hub position within the IFN-γ communication network and expanded in sepsis-associated states. Systematic comparison of 112 integrated machine learning algorithm combinations based on cytoHubba-prioritized genes yielded a 29-gene diagnostic model with cross-cohort discrimination for sepsis. Transcriptional co-expression analysis nominated IFITM3, a marker of the interferon-responsive monocyte state, as a candidate node connecting the diagnostic signature with interferon-related immune dysregulation. Molecular docking and 100-nanosecond MD simulation suggested a structurally stable riociguat-IFITM3 interaction Conclusion: This integrated pharmacovigilance and transcriptomic study identifies sepsis-reporting signals associated with selected endothelin receptor antagonists and characterizes an IFITM3-associated interferon-responsive monocyte state in sepsis datasets. These findings should be interpreted as reporting associations and transcriptomic hypotheses rather than evidence of causal drug-induced sepsis. The predicted riociguat-IFITM3 interaction provides a computational hypothesis for future experimental validation.