Guided multi-objective generative AI for drug design

This article discusses how SandboxAQ is utilizing generative artificial intelligence (AI) and cloud computing on AWS to accelerate computer-aided drug design (CADD) and drug discovery. They developed IDOLPro, a guided multi-objective generative AI system, to efficiently generate novel drug compounds with optimized properties like high binding affinity to target proteins and good synthesizability.

Specifically, the article covers:

• The challenges in traditional virtual screening methods for CADD and how generative AI can explore uncharted chemical space
• The approach of IDOLPro, which uses gradients and optimization to traverse the latent space of a generative model and find optimal drug candidate molecules
• How IDOLPro simultaneously optimizes binding affinity and synthesizability, producing better results than other generative tools on benchmark datasets
• The AWS architecture and services used, including ParallelCluster, EC2 instances, S3, RDS, and MLFlow for experiment tracking
• Performance benchmarks showing IDOLPro generating molecules with better binding affinities than experimentally verified ligands
• The capability of IDOLPro for lead optimization, producing optimized molecules with similar scaffolds as reference ligands
• Conclusion and future plans to incorporate additional properties like toxicity and solubility into the multi-objective optimization