Multi-target-based polypharmacology prediction (mTPP): An approach using virtual screening and machine learning for...

Area of Science:

• Computational chemistry and pharmacology
• Drug discovery and development
• Bioinformatics and machine learning

Background:

• Polypharmacology, the study of drugs acting on multiple targets, is crucial for treating polygenic diseases.
• Existing in-silico methods primarily focus on screening single-target molecules, lacking a comprehensive approach for multi-target drug efficacy.
• A predictive method is urgently needed to understand the relationship between multi-target drug actions and overall therapeutic effectiveness.

Purpose of the Study:

• To develop and validate a novel computational approach, multi-target-based polypharmacology prediction (mTPP), for predicting drug efficacy.
• To explore the relationship between multi-target binding affinities and cellular responses in the context of drug-induced liver injury (DILI).
• To identify potential therapeutic compounds for DILI using the developed mTPP model.

Main Methods:

• Constructed the mTPP model using virtual screening data on target binding strengths and cell proliferation rates against acetaminophen (APAP)-induced liver injury.
• Employed machine learning algorithms including Multi-layer Perceptron (MLP), Support Vector Regression (SVR), Decision Tree Regressor (DTR), and Gradient Boost Regression (GBR).
• Evaluated model performance using test set metrics (R²test = 0.73, EVtest = 0.75) and validated predictions with experimental testing of identified compounds.

Main Results:

• The Gradient Boost Regression (GBR) algorithm demonstrated superior performance in the mTPP model compared to MLP, SVR, and DTR.
• The mTPP model successfully predicted 20 candidate compounds with potential efficacy against DILI.
• Experimental validation confirmed that two predicted candidates, Chelerythrine and Biochanin A, significantly improved the viability of APAP-induced liver injury cells.

Conclusions:

• The developed mTPP model provides a robust computational framework for predicting the efficacy of multi-target drugs.
• This approach holds significant promise for advancing polypharmacology and accelerating the discovery of novel multi-target therapeutics.
• The identified compounds, Chelerythrine and Biochanin A, warrant further investigation as potential treatments for DILI.