Generative AI-Driven Molecular Design: Combining Predictive Models and Reinforcement Learning for Tailored Molecule Generation

Teslim Olayiwola,Miriam Nnadili,Jose Romagnoli,Andrew Okafor,David Akinpelu

DOI: https://doi.org/10.26434/chemrxiv-2023-wcrv3

2023-11-17

Abstract:Molecular design is a critical aspect of various scientific and industrial fields, where the properties of molecules hold significant importance. In this study, a three-fold methodology design is presented that leverages the power of generative artificial intelligence (AI), predictive modeling, and reinforcement learning to create tailored molecules with desired properties. This model synergistically combines deep learning techniques with Self-Referencing Embedded Strings (SELFIES) molecular representation to build a generative model which generates valid molecules and a graphical neural network model that accurately forecasts molecular properties. The Variational Autoencoder (VAE) coupled with reinforcement learning, helps refine molecule generation based on targeted attributes. Data from an experimental study involving surfactants was used to test the framework. Saliency maps for the generated surfactants were produced to identify the features explaining the property values. The results showed that the proposed framework can effectively produce valid molecules within the set property threshold value. This approach not only streamlines molecular design for surfactant systems but also augurs transformative advancements across different scientific and industrial landscapes.

Chemistry

What problem does this paper attempt to address?

The paper aims to address the problem of how to use artificial intelligence (AI) to generate molecules with specific attributes. The research proposes a triple approach that combines predictive models, reinforcement learning, and generative AI to create customized molecules. Specifically, it uses deep learning techniques along with SELFIES (a string representation that guarantees the generation of valid molecules) to build a generative model, and predicts molecular properties through variational autoencoders (VAE) and graph neural networks (GNN). Molecular generation is optimized based on target attributes through reinforcement learning. The experiments demonstrate that this framework can effectively generate valid molecules within set thresholds, particularly in designing surfactant molecules with low critical micelle concentration (CMC). This approach not only simplifies the process of designing specific molecules but also indicates potential transformations in various scientific and industrial fields.

Generative AI-Driven Molecular Design: Combining Predictive Models and Reinforcement Learning for Tailored Molecule Generation

Machine learning-aided generative molecular design

Molecular Design in Synthetically Accessible Chemical Space via Deep Reinforcement Learning

Mol-AIR: Molecular Reinforcement Learning with Adaptive Intrinsic Rewards for Goal-directed Molecular Generation

Advances in Deep Learning-Based 3D Molecular Generative Models

Rethinking Molecular Design: Integrating Latent Variable and Auto-Regressive Models for Goal Directed Generation

Diversity oriented Deep Reinforcement Learning for targeted molecule generation

LS-MolGen: Ligand-and-Structure Dual-driven Deep Reinforcement Learning for Target-specific Molecular Generation Improves Binding Affinity and Novelty

Evaluation of Reinforcement Learning in Transformer-based Molecular Design

LS-MolGen: Ligand-and-Structure Dual-Driven Deep Reinforcement Learning for Target-Specific Molecular Generation Improves Binding Affinity and Novelty.

Molecular De Novo Design through Deep Reinforcement Learning

Generative artificial intelligence for small molecule drug design

Human-in-the-loop active learning for goal-oriented molecule generation

Guided Multi-objective Generative AI to Enhance Structure-based Drug Design

Large Language Models Open New Way of AI-Assisted Molecule Design for Chemists

Generative Organic Electronic Molecular Design Informed by Quantum Chemistry

Multi-Objective Molecular Design in Constrained Latent Space

High-throughput property-driven generative design of functional organic molecules

Scalable Fragment-Based 3D Molecular Design with Reinforcement Learning

Adaptive language model training for molecular design

Unlocking comprehensive molecular design across all scenarios with large language model and unordered chemical language