Biomarker-driven drug repurposing for NAFLD-associated hepatocellular carcinoma using machine learning integrated...

Area of Science:

• Biomarker discovery
• Genomics
• Computational biology

Background:

• Non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH) incidence is rising with obesity and diabetes.
• NASH is a significant cause of hepatocellular carcinoma (HCC), a deadly cancer with poor survival rates.
• Biomarkers are crucial for early screening, monitoring, and drug development in NAFLD/NASH patients.

Purpose of the Study:

• To identify robust gene biomarkers for the progression of NAFLD-related HCC using transcriptomic data.
• To evaluate machine learning models for accurate disease stage classification.
• To discover potential drug candidates for repurposing against identified biomarkers.

Main Methods:

• An ensemble feature selection framework was applied to transcriptomic data.
• Seven machine learning algorithms were assessed for disease stage classification, with DISCR being the most accurate.
• Cox regression survival analysis and molecular docking were used to validate biomarkers and screen drug candidates.

Main Results:

• Ten top genes were identified through ensemble feature selection, with eight validated as potential biomarkers via survival analysis.
• Key genes like ABAT, ABCB11, MBTPS1, and ZFP1 were highlighted, involved in metabolic and cellular processes.
• Eighty-one candidate drugs were identified, with Diosmin, Esculin, Lapatinib, and Phenelzine showing promise through molecular docking.

Conclusions:

• A consensus approach enhances biomarker identification accuracy for NAFLD-associated HCC.
• Validated biomarkers and repurposed drugs offer potential for early intervention strategies.
• This study provides a foundation for improved therapeutic options and patient outcomes in NAFLD-related HCC.