Abstract
OBJECTIVE: Cancer-associated fibroblasts (CAFs) are a critical component of the tumor microenvironment and play a significant role in renal cell carcinoma (RCC) progression and treatment response. However, current methods for evaluating CAFs infiltration in RCC are inadequate. This study aims to develop a non-invasive histopathological model based on H&E staining and collagen features to predict CAFs infiltration and investigate its prognostic value and biological relevance. METHODS: We conducted a retrospective analysis using H&E pathological images and clinical data from The Cancer Genome Atlas (TCGA) database. A Pathomics model integrating 465 histopathological features was constructed using machine learning algorithms (n = 354) to predict CAFs infiltration. A preliminary technical validation was performed using multiphoton microscopy-based collagen quantification (n = 25) to assess the correlation between the Pathomics Score and CAF-related fibrotic activity. Enrichment analysis, immune cell infiltration profiling, and mutation analysis were employed to explore the biological mechanisms underlying the model. RESULTS: The Pathomics model demonstrated high predictive accuracy (AUC=0.813) and correlated significantly with collagen deposition (r = 0.66, P < 0.001). High Pathomics scores were independently associated with poor survival (HR=1.80, P = 0.003) and linked to key biological processes, including YAP/TAZ activation, extracellular matrix (ECM) remodeling, immune suppression (e.g., CD276, IDO1), and frequent mutations in VHL and PBRM1 (>40%). CONCLUSION: This study establishes the first H&E-based Pathomics framework for quantifying CAFs infiltration in RCC, providing a cost-effective and non-invasive tool for preliminary risk stratification. The model's strong correlation with collagen features and its ability to reveal underlying molecular mechanisms highlight its potential for potential value in understanding the stromal microenvironment, though further external validation is required for clinical translation.