Machine Learning and Single-Cell Analysis Identify Molecular Features of IPF-Associated Fibroblast Subtypes and Their Implications on IPF Prognosis.

Abstract

Idiopathic pulmonary fibrosis (IPF) is a devastating lung disease of unknown cause, and the involvement of fibroblasts in its pathogenesis is well recognized. However, a comprehensive understanding of fibroblasts' heterogeneity, their molecular characteristics, and their clinical relevance in IPF is lacking. In this study, we aimed to systematically classify fibroblast populations, uncover the molecular and biological features of fibroblast subtypes in fibrotic lung tissue, and establish an IPF-associated, fibroblast-related predictive model for IPF. Herein, a meticulous analysis of scRNA-seq data obtained from lung tissues of both normal and IPF patients was conducted to identify fibroblast subpopulations in fibrotic lung tissues. In addition, hdWGCNA was utilized to identify co-expressed gene modules associated with IPF-related fibroblasts. Furthermore, we explored the prognostic utility of signature genes for these IPF-related fibroblast subtypes using a machine learning-based approach. Two predominant fibroblast subpopulations, termed IPF-related fibroblasts, were identified in fibrotic lung tissues. Additionally, we identified co-expressed gene modules that are closely associated with IPF-fibroblasts by utilizing hdWGCNA. We identified gene signatures that hold promise as prognostic markers in IPF. Moreover, we constructed a predictive model specifically focused on IPF-fibroblasts which can be utilized to assess disease prognosis in IPF patients. These findings have the potential to improve disease prediction and facilitate targeted interventions for patients with IPF.