DockingGA: enhancing targeted molecule generation using transformer neural network and genetic algorithm with docking...

Area of Science:

• Computational chemistry and cheminformatics
• Artificial intelligence in drug discovery
• Molecular modeling and simulation

Background:

• Generative molecular models explore chemical space for novel molecules.
• Traditional methods like genetic algorithms lack integration with docking simulations.
• Existing models often require extensive post-generation optimization for drug candidacy.

Purpose of the Study:

• To introduce DockingGA, a hybrid model combining Transformer networks and genetic algorithms.
• To enhance the generation of molecules with improved binding affinity to specific biological targets.
• To address limitations of traditional generative models in drug discovery pipelines.

Main Methods:

• Utilized Self-referencing Chemical Structure Strings for molecular representation.
• Integrated genetic algorithms with Transformer neural networks for molecular optimization.
• Employed docking simulations to guide the generative process and assess binding affinity.

Main Results:

• DockingGA demonstrated superior performance across top 1, 10, and 100 generated molecules in docking results.
• Achieved 100% novelty in generated molecules, ensuring unique chemical entities.
• Generated molecules exhibited favorable and appropriate physicochemical properties, suitable for drug development.

Conclusions:

• DockingGA represents a significant advancement in generative molecular modeling for drug discovery.
• The model's ability to optimize binding affinity and physicochemical properties streamlines the path to candidate drugs.
• This innovation opens new avenues for applying AI in practical drug discovery and development.