Metabolic Profiling Combined With Machine Learning for Assessing Clonal Hematopoiesis Status in Chronic Immune Thrombocytopenia: An Exploratory Analysis.

Abstract

Clonal hematopoiesis (CH) has emerged as a potential contributor to the pathogenesis of immune thrombocytopenia (ITP), yet the clinical and metabolic features of ITP patients harboring CH-associated mutations remain poorly characterized. We retrospectively analyzed a genomic discovery cohort of 1817 patients with chronic ITP, identifying CH in 196 individuals (10.8%), with DNMT3A as the most frequently mutated gene. In a metabolomic profiling cohort of 539 patients (492 ITP-non-CH, 47 ITP-CH), we characterized plasma metabolic signatures using untargeted LC-MS/MS and systematically evaluated the discriminative capacity of integrated metabolic-clinical features through multiple machine learning algorithms. Differential analysis identified 14 plasma metabolites between ITP-non-CH and ITP-CH patients. The cohort was randomly partitioned into a training set (n = 377) and an independent test set (n = 162). Among eight evaluated algorithms, logistic regression with balanced class weighting and 14 metabolites + 3 clinical features (sex, age, PLT) achieved the optimal discriminative balance, with a test AUC of 0.721 (95% CI: 0.57-0.87). Our study reveals distinct hematologic and metabolic dysregulation between ITP-CH and ITP-non-CH, and highlights the potential of integrative hematologic-metabolic signatures for improving disease stratification in ITP.

Abstract as supplied by the indexing source. © the authors / publisher.