HP-MoleQ: An Effective Predictive Model for High-Throughput Screening of Food-Derived Hepatoprotective Compounds

Area of Science:

• Natural Product Chemistry
• Computational Chemistry
• Hepatology

Background:

• Hepatoprotective natural compounds are crucial for functional foods and preventing liver disorders.
• Existing methods for discovering these compounds can be inefficient and unreliable.

Purpose of the Study:

• To develop a novel predictive framework, HP-MoleQ, for efficient and reliable discovery of hepatoprotective compounds.
• To validate the efficacy of top-ranked compounds identified by HP-MoleQ in cellular models of liver injury.

Main Methods:

• Development of HP-MoleQ, integrating a pretrained Transformer and an uncertainty quantification module.
• Screening the FoodB database using HP-MoleQ and ranking compounds by confidence intervals.
• Experimental validation of three prioritized compounds (1,2,6-trigalloyl-beta-D-glucopyranose, hamamelitannin, and 6-hydroxyluteolin 7-glucoside) in cellular models.

Main Results:

• HP-MoleQ demonstrated superior performance compared to existing models in screening hepatoprotective compounds.
• The top 20 candidates included phenolic glucopyranosides, furanosides, and flavonoid glycosides.
• The three validated compounds significantly attenuated markers of alcohol-induced hepatotoxicity and fatty liver disease in vitro.

Conclusions:

• The HP-MoleQ platform is effective for high-confidence screening and identification of potential hepatoprotective agents.
• The validated compounds exhibit significant hepatoprotective effects, supporting their potential therapeutic applications.
• This study highlights the synergy between computational prediction and experimental validation in natural product drug discovery.