Machine Learning-Driven Discovery of TRIM Genes as Diagnostic Biomarkers for Idiopathic Pulmonary Fibrosis.

BACKGROUND Idiopathic pulmonary fibrosis (IPF) is a progressive lung disease with limited effective treatments and significant challenges in early diagnosis. Identifying reliable biomarkers is crucial for improving diagnostic accuracy and patient outcomes. MATERIAL AND METHODS We analyzed TRIM family gene expression in IPF patients and healthy controls using GSE93606, GSE33566, and GSE38958 datasets. Consensus clustering and WGCNA identified IPF subtypes and hub genes. Machine learning models (RF, GLM, SVM, XGB) were built to identify key disease genes. A nomogram for clinical prediction was developed and validated. Peripheral blood samples from IPF patients and healthy controls were used to validate gene expression via qPCR. RESULTS TRIM family genes were significantly differentially expressed between IPF patients and healthy controls. Two distinct IPF subtypes (C1 and C2) were identified, each exhibiting unique biological functions and signaling pathways. The RF model outperformed other machine learning models, identifying TNIK, NCL, ROPN1L, MTR, and HNRNPH1 as key disease-characteristic genes. The nomogram demonstrated good predictive accuracy (AUC: 0.741, 95% CI: 0.556-0.897). qPCR validation confirmed increased expression of 4 genes in IPF patients, except for ROPN1L, which showed decreased expression. CONCLUSIONS This study identifies and validates TRIM family genes as potential biomarkers for IPF diagnosis using clinical samples. The findings support the integration of these biomarkers into diagnostic workflows, potentially enhancing early diagnosis and personalized treatment strategies for IPF patients. Further research is needed to explore the prognostic value and underlying mechanisms of these genes.