Unsupervised Phenotyping of Asthma: Integrating Serum Periostin with Clinical and Inflammatory Profiles

<b>Background/Objectives:</b> Asthma is a heterogeneous inflammatory airway disease. Periostin, a matricellular protein induced by interleukin-13, contributes to airway inflammation and remodeling. This study evaluated serum periostin as a diagnostic biomarker and explored multidimensional phenotypes in adult asthma. <b>Methods:</b> A cross-sectional study included 76 adults, with 25 healthy controls, 25 moderate, and 26 severe asthma patients, classified per Global Initiative for Asthma (GINA)-2020 guidelines. Serum periostin was measured using an enzyme-linked immunosorbent assay (ELISA). Diagnostic accuracy was assessed using receiver operating characteristic (ROC) analysis, Firth-penalized logistic regression, bootstrap calibration (1000 resamples), decision curve analysis (DCA), and gradient boosting machine (GBM) validation. Principal component analysis (PCA) followed by k-means clustering identified distinct phenotypes based on clinical, functional, and inflammatory variables. <b>Results:</b> Asthma patients had higher serum periostin than controls (median 52.9 vs. 32.5 pg/mL; <i>p</i> < 0.01), with excellent diagnostic accuracy (AUC = 0.987; sensitivity = 94.1%, specificity = 100%). Firth regression identified periostin as the only independent predictor of asthma diagnosis (β = 0.387; OR = 1.47; 95% CI 1.23-2.08; <i>p</i> < 0.001). Calibration showed minimal error (MAE = 0.042) and DCA demonstrated clear net benefit. GBM confirmed periostin as the dominant diagnostic predictor. PCA revealed three clusters: Cluster 1: younger, lower periostin, preserved lung function, good symptom control; Cluster 2: intermediate periostin, greater airflow limitation, poorer control; and Cluster 3: highest periostin, elevated systemic inflammation (NLR, PLR, SII), with moderate functional impairment. <b>Conclusions:</b> Serum periostin is a reliable diagnostic biomarker for asthma. Multidimensional clustering highlights clinically relevant phenotypes linked to periostin, inflammatory burden, and lung function, supporting its role in personalized asthma management.