VGAE-MCTS: A New Molecular Generative Model Combining the Variational Graph Auto-Encoder and Monte Carlo Tree Search

Area of Science:

• Computational chemistry
• Artificial intelligence in chemistry

Background:

• Molecular generation is vital for drug discovery, materials science, and chemical exploration.
• Current AI methods, like graph-based molecular generative models, aid in identifying molecules with specific properties.

Purpose of the Study:

• To propose a novel molecular generative model integrating graph-based deep neural networks and reinforcement learning.
• To evaluate the generated molecules for validity, novelty, and optimized physicochemical properties.

Main Methods:

• Development of a new molecular generative model.
• Integration of a graph-based deep neural network with a reinforcement learning technique.
• Evaluation of generated molecules' properties and chemical space exploration.

Main Results:

• The proposed model successfully generated molecules with valid, novel, and optimized physicochemical properties.
• The model demonstrated the ability to explore uncharted regions of chemical space.
• Efficient discovery and design of new molecules were achieved.

Conclusions:

• The novel AI approach holds significant potential to revolutionize drug discovery and materials science.
• This method accelerates scientific innovation by enabling efficient molecular design.
• Exploration of unexplored chemical spaces offers promising prospects for future research.