Integrating machine learning and neural networks for new diagnostic approaches to idiopathic pulmonary fibrosis and immune infiltration research

Abstract

Background Idiopathic pulmonary fibrosis (IPF) is an interstitial lung disease with a fatal outcome, known for its rapid progression and unpredictable clinical course. However, the tools available for diagnosing and treating IPF are quite limited. This study aims to identify and screen potential biomarkers for IPF diagnosis, thereby providing new diagnostic approaches. Methods We choosed datasets from the Gene Expression Omnibus (GEO) database, including samples from both IPF patients and healthy controls. For the training set, we combined two gene array datasets (GSE24206 and GSE10667) and utilized GSE32537 as the test set. We identified differentially expressed genes (DEGs) between IPF and normal tissues and determined IPF-related modules using Weighted Gene Co-expression Network Analysis (WGCNA). Subsequently, we employed two machine learning strategies to screen potential diagnostic biomarkers. Candidate biomarkers were quantitatively evaluated using Receiver Operating Characteristic (ROC) curves to identify key diagnostic genes, followed by the construction of a nomogram. Further validation of the expression of these genes through transcriptomic sequencing data from IPF and normal group animal models. Next, we conducted immune infiltration analysis, single-gene Gene Set Enrichment Analysis (GSEA), and targeted drug prediction. Finally, we created an artificial neural network model specifically for IPF. Results We identified ASPN, COMP, and GPX8 as candidate biomarker genes for IPF, all of which exhibited Area Under the Curve (AUC) above 0.90. These genes were validated by RT-qPCR. Immune infiltration analysis revealed that specific immune cell types are closely related to IPF, suggesting that these immune cells may play a significant role in the pathogenesis of IPF. Conclusion ASPN, COMP, and GPX8 have been identified as potential diagnostic genes for IPF, and the most relevant immune cell types have been determined. Our research results propose potential biomarkers for diagnosing IPF and present new pathways for investigating its pathogenesis and devising novel therapeutic approaches.