The methodological architecture of the study deserves closer examination, because the reliability of any shared molecular signature depends heavily on how the underlying data were assembled and interrogated. The investigators drew on publicly available gene expression datasets deposited in the Gene Expression Omnibus, applying standardized preprocessing pipelines to harmonize samples across studies that were originally generated by different laboratories, on different microarray or sequencing platforms, and with different clinical phenotyping conventions. Cross-dataset integration of this kind is a well-established strategy in computational genomics, but it introduces well-known challenges: batch effects, differences in tissue composition, and variable proportions of contaminating cell types can all masquerade as genuine biological signals. The fact that the authors identified six concordantly upregulated and eleven concordantly downregulated differentially expressed genes shared between IPAH and malignant pleural mesothelioma suggests that their harmonization procedures were sufficiently rigorous to survive the noise inherent in meta-analytic transcriptomics, yet the modest number of overlapping genes is itself informative. It implies that while the two diseases do share a transcriptional core, the majority of their differential expression programs remain disease-specific, consistent with the idea that IPAH is not simply a malignancy in disguise but a vascular disorder that has co-opted selected cancer-like programs.

The enrichment patterns of the shared genes add biological texture to this picture. The concordantly upregulated genes clustered around cell cycle regulation, a finding that resonates with decades of histopathological observation in pulmonary arterial hypertension. Pulmonary arterial smooth muscle cells and endothelial cells within remodeled vessels of IPAH patients are known to proliferate in an apparently uncontrolled fashion, forming the concentric laminar lesions and plexiform structures that pathologists have long recognized. That these proliferating cells share a transcriptional signature of cell cycle activation with mesothelioma tumor cells provides molecular corroboration for the cancer hypothesis first articulated by Voelkel and colleagues in the late 1990s. The downregulated arm of the shared signature, enriched for neutrophil chemotaxis, is equally telling. It points to a perturbation of innate immune trafficking, suggesting that IPAH vasculature may fail to recruit or retain neutrophils in the manner expected of inflamed tissue, or alternatively that the peripheral blood compartment sampled in these datasets reflects a redistribution of these cells away from circulation and into the remodeled pulmonary vasculature itself.

The convergence of three independent machine learning algorithms on CXCR2, S100A9, and LCN2 strengthens the case that these genes are not artifacts of any single feature-selection approach. Least absolute shrinkage and selection operator regression, random forest classification, and support vector machine recursive feature elimination each operate on fundamentally different mathematical principles, and genes that survive all three filters are less likely to be statistical flukes. CXCR2 is a chemokine receptor whose ligands include interleukin-8 and related CXC chemokines, and it sits at the interface of inflammation and cell migration. Its concordant downregulation, together with the observed decrease in neutrophil infiltration in IPAH samples, suggests a coordinated suppression of neutrophil-directed chemotactic signaling. S100A9, a calcium-binding protein released abundantly by activated neutrophils and monocytes, functions as a danger-associated molecular pattern and has been implicated in sterile inflammation across multiple cardiovascular diseases. LCN2, also known as lipocalin-2 or neutrophil gelatinase-associated lipocalin, participates in iron sequestration and metabolic reprogramming, and its correlation with metabolic pathways in this study aligns with the growing recognition that Warburg-like aerobic glycolysis characterizes proliferating pulmonary vascular cells in IPAH just as it characterizes many tumors.

The immune infiltration analysis performed by single-sample gene set enrichment adds a cellular dimension to the transcriptomic findings. The reported increase in Th17 cells alongside decreased neutrophils in IPAH paints a picture of adaptive immune skewing concurrent with innate immune depletion from the sampled compartment. Th17 cells, characterized by their production of interleukin-17, are known to promote fibrosis and vascular remodeling in several disease contexts, and their expansion in IPAH is consistent with the inflammatory phenotype that has been described in advanced pulmonary vascular disease. The correlation between CXCR2 and S100A9 expression and neutrophil infiltration provides internal consistency: if these genes are expressed by or regulate neutrophil biology, their abundance should track with neutrophil representation in the sample, and the observed correlations satisfy that expectation. The separate correlation of LCN2 with metabolic reprogramming rather than with immune cell fractions suggests that this gene may serve a dual role, bridging inflammatory signaling and the metabolic rewiring that underpins the proliferative phenotype of remodeled pulmonary vasculature.

The drug-repurposing component of the study, while preliminary, illustrates how computational target identification can be coupled to structure-based screening. Navarixin, a clinically investigated CXCR2 antagonist, emerged as the strongest binder to CXCR2 in molecular docking, with a predicted binding free energy of minus 10.1 kilocalories per mole. Molecular dynamics simulations subsequently confirmed that the docked pose remained stable over the simulated trajectory, with the ligand maintaining its contacts within the receptor binding pocket. These in silico results are encouraging but must be interpreted with appropriate caution. Docking scores are approximations that depend on the quality of the protein structure, the force field parameters, and the sampling algorithm, and a favorable docking score does not guarantee pharmacological activity, selectivity, or clinical efficacy. Nevertheless, the identification of an existing CXCR2 antagonist with predicted high affinity provides a concrete, testable hypothesis: if CXCR2 signaling contributes to the immune dysregulation observed in IPAH, then pharmacological blockade of this receptor might modulate disease-relevant pathways in ways that current vasodilator therapies cannot.

The qRT-PCR validation in human peripheral blood samples adds an experimental anchor to the computational pipeline. Confirming significant downregulation of all three candidate biomarker genes in an independent cohort of IPAH patients demonstrates that the transcriptomic signal detected in silico is reproducible in a real-world clinical population, at least for the genes and the sample type tested. Peripheral blood, however, is a surrogate compartment, and the expression levels measured there may not directly reflect what is happening within the pulmonary vasculature itself. Future studies that examine these genes in lung tissue, pulmonary artery specimens, or isolated vascular cells from IPAH patients would provide a more direct test of whether the shared signature is operative at the site of disease.

The statistical rigor applied to the classification model is noteworthy and sets a standard for computational biomarker studies. Achieving an area under the receiver operating characteristic curve of 0.909 in the training set and 0.875 in an external validation set is promising, but the authors went further by employing leave-one-out cross-validation, which yielded an AUC of 0.879, and bootstrap optimism correction, which produced an optimism-adjusted AUC of 0.896. These additional metrics guard against the overfitting that plagues many machine learning studies in biomedicine, where apparent performance in the training data inflates expectations that are not borne out in independent samples. The narrow gap between the training and corrected performance figures suggests that the three-gene signature is reasonably robust, though the limited size of the external validation cohort remains a constraint that only larger, multicentre studies can overcome.

The ROTS sensitivity analysis, which confirmed that all 1710 differentially expressed genes in the mesothelioma dataset were reproducible and directionally concordant at a false discovery rate threshold below 0.05 despite an imbalanced case-to-control ratio, addresses a common criticism of transcriptomic meta-analyses. Imbalanced group sizes can inflate false positive rates and produce unstable rankings, and demonstrating that the full differential expression profile withstands a rank-based reproducibility test strengthens confidence that the downstream analyses were built on a solid foundation.

Looking forward, several avenues of investigation follow naturally from these findings. Functional experiments using pulmonary arterial smooth muscle cells or endothelial cells, in which CXCR2, S100A9, or LCN2 are knocked down or overexpressed, could establish whether these genes are causal drivers of the proliferative and inflammatory phenotype or merely passive markers of it. Animal models of pulmonary hypertension, such as the Sugen-hypoxia rat or the monocrotaline mouse, could be used to test whether pharmacological CXCR2 blockade alters the course of vascular remodeling. Proteomic studies could determine whether the transcriptomic findings translate to measurable changes in circulating protein levels, which would be a prerequisite for developing a clinically deployable biomarker panel. And longitudinal studies tracking these gene expression signatures over the natural history of IPAH could reveal whether they carry prognostic information or respond to therapeutic intervention. Until such work is completed, the present study should be understood as a well-executed hypothesis-generating exercise that has identified a plausible molecular bridge between two seemingly disparate diseases and has laid the groundwork for the experimental and clinical studies needed to determine whether that bridge can bear the weight of therapeutic translation.

Subject of Research: Shared transcriptomic biomarkers between idiopathic pulmonary arterial hypertension and malignant pleural mesothelioma identified by machine learning and molecular docking

Article Title: Machine learning, molecular docking, and experimental validation identify shared molecular signatures between idiopathic pulmonary arterial hypertension and malignant pleural mesothelioma: a retrospective study

Article References: Machine learning, molecular docking, and experimental validation identify shared molecular signatures between idiopathic pulmonary arterial hypertension and malignant pleural mesothelioma: a retrospective study. (n.d.). https://doi.org/10.3389/fgene.2026.1894027

Image Credits: AI Generated

DOI: 10.3389/fgene.2026.1894027

Keywords: idiopathic pulmonary arterial hypertension, malignant pleural mesothelioma, machine learning, molecular docking, biomarkers, transcriptomics, CXCR2, S100A9, LCN2, immune infiltration, molecular dynamics, drug repurposing