Multi-Objective-Conditioned-Return-Blocks Decision Transformer: A Novel Approach for Multi-Objective De Novo Drug Design.

M-GoRT achieves high success in multi-objective drug design, maintaining 80% success rates while enhancing structural diversity and novelty across complex biological targets.

• Why it matters: Designing molecules that meet multiple, often conflicting, properties is crucial for effective drug development but remains a significant challenge due to the complexity of balancing targets.
• What they did: The team developed M-GoRT, a decision transformer architecture with return-to-go blocks conditioned on goal embeddings, trained without exploration or reward weighting, to optimize multiple targets efficiently.
• The result: M-GoRT outperforms existing methods in success rates and structural diversity, demonstrating its potential as a practical tool for multi-property molecular generation in high-dimensional biological tasks.