Interpretable Machine Learning Identifies Hub Biomarkers of Renal Fibrosis and Their Potential Medical Applications

Area of Science:

• Nephrology
• Genomics
• Biomarker Discovery

Background:

• Renal fibrosis is a primary driver of chronic kidney disease (CKD).
• Current assessment methods like renal biopsy have limitations due to fibrosis heterogeneity.
• Accurate diagnostic biomarkers and therapeutic targets for renal fibrosis are needed.

Purpose of the Study:

• To identify accurate diagnostic biomarkers for renal fibrosis.
• To discover potential therapeutic targets for renal fibrosis.
• To explore the role of specific genes in renal fibrosis pathogenesis.

Main Methods:

• Analysis of renal fibrosis transcriptomic datasets to identify differentially expressed genes (DEGs).
• Selection of hub genes using LASSO regression and analysis of immune infiltration via CIBERSORT.
• Development and validation of machine learning models (XGBoost, DNN) with SHAP analysis for biomarker identification.
• Experimental validation using histological staining, Western blot, and cellular assays.
• Identification of potential therapeutic drugs through molecular docking.

Main Results:

• 26 fibrosis-related genes were identified, correlating with inflammation and immune infiltration.
• Machine learning models achieved high diagnostic accuracy (XGBoost: 96%, DNN: 92%).
• AGR2 and DOCK2 were identified as top predictors and validated as significantly upregulated in fibrotic processes.
• Potential therapeutic agents like Dexamethasone and Ciclosporin were identified for targeting AGR2.

Conclusions:

• AGR2 and DOCK2 are novel biomarkers for renal fibrosis.
• These genes represent potential therapeutic targets for renal fibrosis.
• The study highlights a dual role for AGR2 and DOCK2 in diagnosis and targeted therapy development.