A generative framework for enhancing drug target interaction prediction in drug discovery

Area of Science:

• Computational chemistry
• Artificial intelligence in drug discovery
• Bioinformatics

Background:

• In silico drug-target interaction (DTI) prediction is crucial for accelerating drug discovery.
• Traditional methods face challenges with biochemical data complexity and scale, limiting accuracy.
• There is a need for advanced computational frameworks to enhance DTI prediction.

Purpose of the Study:

• To present VGAN-DTI, a generative AI framework combining GANs, VAEs, and MLPs for improved DTI prediction.
• To enhance the encoding of molecular features and uncover molecular mechanisms.
• To boost predictive accuracy and reliability in drug discovery.

Main Methods:

• Developed VGAN-DTI, integrating generative adversarial networks (GANs) for candidate generation and variational autoencoders (VAEs) for feature representation optimization.
• Utilized multilayer perceptrons (MLPs) trained on BindingDB data for interaction classification and binding affinity prediction.
• Conducted rigorous ablation studies to validate framework robustness.

Main Results:

• Achieved high performance metrics: 96% accuracy, 95% precision, 94% recall, and 94% F1 score.
• Demonstrated superior performance compared to existing DTI prediction methods.
• Validated the effectiveness and robustness of the VGAN-DTI framework.

Conclusions:

• VGAN-DTI significantly enhances in silico drug-target interaction prediction accuracy.
• The framework optimizes innovation, synthetic feasibility, and predictive accuracy in drug discovery.
• VGAN-DTI advances data-driven pharmaceutical research by ensuring reliable DTI predictions.