BiRLNN: Bidirectional Reinforcement-Learning Neural Network for Constrained Molecular Design

We present BiRLNN, a bidirectional molecular design framework that combines recurrent neural networks with reinforcement learning to optimize drug-like properties of generated compounds. Unlike traditional unidirectional models, our approach generates molecular sequences synchronously in both forward and backward directions, enabling more balanced exploration of chemical space while maintaining constraint requirements during molecular design. We employ Self-Referencing Embedded Strings representations to ensure 100% syntactic validity of generated molecules. To guide generation towards desirable pharmacological targets, we implement a multi-objective reward function based on quantitative estimate of drug-likeness and synthetic accessibility, and apply policy gradient-based reinforcement learning for fine-tuning. Simulation results show that our model covers a much larger part of the chemical space compared to unidirectional ones, allowing it to explore regions that contain molecules unreachable by the latter. Moreover, the reinforcement learning process successfully steers the generation process toward desirable compound classes with improved reward metrics. Our results demonstrate that BiRLNN offers a robust and flexible strategy for navigating chemical space in multi-objective drug design tasks.