Predicting Cardiotoxicity of Molecules Using Attention-Based Graph Neural Networks.

Area of Science:

• Drug discovery and development
• Computational toxicology
• Pharmacology

Background:

• Drug development is significantly hampered by toxicity concerns, leading to high failure rates and increased costs.
• Drug-induced cardiotoxicity is a severe adverse effect, particularly problematic for cancer therapeutics.
• Existing computational methods for predicting cardiotoxicity have limitations in performance and interpretability.

Purpose of the Study:

• To develop a more effective computational framework for predicting molecular cardiotoxicity.
• To improve the accuracy and interpretability of cardiotoxicity assessments in drug discovery.
• To provide a user-friendly tool for researchers to evaluate potential drug cardiotoxicity.

Main Methods:

• Utilized an attention-based graph neural network (GNN) architecture.
• Developed a novel computational framework for molecular cardiotoxicity prediction.
• Validated model performance against existing computational approaches.

Main Results:

• The proposed attention-based GNN framework demonstrated superior performance in predicting cardiotoxicity compared to other methods.
• Experimental results confirmed the stability and reliability of the developed model.
• The framework successfully identified potential cardiotoxic molecules.

Conclusions:

• The developed computational framework offers a more effective solution for predicting drug-induced cardiotoxicity.
• This approach can aid in de-risking drug candidates early in the development pipeline.
• An accessible online web server has been created to facilitate the use of this predictive model by researchers.