July 23, 2025

A Proteomics-Driven Machine Learning Tool for Distinguishing ET from pre-PMF

Key Points

The study compares diagnostic accuracy between proteomic profiling and clinical variables for essential thrombocythemia.
A 9-protein classifier achieved an AUC of 0.895, indicating robust discrimination between ET and pre-PMF.
Logistic regression was utilized to assess clinical predictors in a cohort of 440 patients for model development.
The proteomic approach shows promise for early-stage diagnosis of myeloproliferative neoplasms, enhancing clinical decision-making.

Abstract

Abstract Essential thrombocythemia (ET) and prefibrotic primary myelofibrosis (pre-PMF) are phenotypically similar but biologically distinct myeloproliferative neoplasms (MPNs), making accurate diagnosis critical yet challenging. We conducted a retrospective study, including 434 ET and 91 pre-PMF patients. Clinical predictors were evaluated using logistic regression in 440 patients. A subset of 85 patients underwent data-independent acquisition (DIA)–based proteomic profiling of FFPE bone marrow samples. Diagnostic models were developed using clinical, proteomic, and combined features, with performance assessed via nested cross-validation. A 9-protein classifier was constructed using random forest followed by support vector machine–recursive feature elimination (SVM-RFE). The proteomic model achieved superior diagnostic performance compared to the clinical model (AUC = 0.849 vs. 0.499) and was comparable to the combined model (AUC = 0.845). The 9-protein panel showed robust discrimination (AUC = 0.895), with top proteins (ARHGEF19, CAST, SFTPA2) performing well in JAK2V617F⁺ (AUC = 0.971) and CALR⁺ (AUC = 0.768) subsets. Proteomic profiling thus outperforms conventional clinical variables and offers a reproducible, molecular-based tool for distinguishing ET from pre-PMF, with potential utility across molecular subtypes in early-phase MPN diagnosis.

AI से पूछें

Bookmark