Platelet gene signatures were able to differentiate patients with pulmonary artery stenosis from those without, achieving an AUC of 0.946 in the training set.
Can platelet gene signatures accurately detect pulmonary artery stenosis and differentiate its subtypes in patients with pulmonary hypertension?
Platelet-derived RNA signatures can accurately identify pulmonary artery stenosis and differentiate between CTEPH and fibrosing mediastinitis in patients with pulmonary hypertension.
Pulmonary artery stenosis (PAS) is a major cause of pulmonary hypertension (PH). The advancement of non-invasive biomarkers to identify PAS in high-risk individuals has the potential to enhance the precision of clinical evaluations related to PH. This study aimed to present evidence that gene expression data within blood platelets could be valuable for detecting PAS in patients with PH. Platelets were isolated from 241 PH patients and 98 healthy controls for RNA sequencing. Differentially expressed genes (DEGs) were identified between PAS and non-PAS, and between chronic thromboembolic pulmonary hypertension (CTEPH) and PH caused by fibrosing mediastinitis (FM-PH). Three machine learning algorithms—random forest (RF), extreme gradient boosting (XGBoost), and Boruta—were applied to select platelet genes of discriminative capability. Genes identified by all three algorithms were used for subsequent model construction. Fourteen predictive models were trained and validated using repeated fivefold cross-validation. Functional enrichment and gene set enrichment analyses (GSEA) were performed. Compared with the non-PAS group, PH patients with PAS exhibited 244 upregulated and 1,051 downregulated genes in platelets. GSEA revealed upregulation of pathways including platelet activation, fluid shear stress and atherosclerosis, and Rap1 signaling, alongside downregulation of PI3K-Akt and mTOR signaling in patients with PAS. Six platelet RNAs were identified by RF, XGBoost, and Boruta for differentiating PAS from non-PAS. The RF model, with NOTCH1 contributing most significantly (highest mean decrease in Gini index), achieved the highest area under the curves (AUCs) of 0.946, 0.862, and 0.749 in the training, internal, and external validation sets, respectively. Within PAS, 669 genes were upregulated and 697 downregulated in CTEPH versus FM-PH. Pathways including vascular smooth muscle contraction and blood vessel remodeling were positively enriched in platelets from patients with CTEPH compared to FM-PH. Two genes, PPP1CA and MAPRE1, were shared across all three feature selection algorithms for discriminating between CTEPH and FM-PH. The RF model based on these genes achieved the highest AUCs of 0.960, 0.925, and 0.961 across the training, internal, and external validation sets. Platelet-derived biomarkers are potentially useful in identifying PAS and differentiating its subtypes in individuals with PH.
Jin et al. (Fri,) conducted a other in Pulmonary Hypertension (n=333). Platelet gene signatures vs. Healthy controls was evaluated on Differentially expressed genes between PAS and non-PAS. Platelet gene signatures were able to differentiate patients with pulmonary artery stenosis from those without, achieving an AUC of 0.946 in the training set.