Symmetry-Aware Actor-Critic for 3D Molecular Design
This work is significant for researchers and engineers in materials science and drug discovery, as it offers a new method to automate and accelerate the search for novel 3D molecular structures.
This paper addresses the challenge of 3D molecular design using deep reinforcement learning by proposing a novel actor-critic architecture. It leverages rotational symmetries through a spherical harmonics series expansion, enabling the generation of molecular structures previously unattainable and improving generalization and quality.
Automating molecular design using deep reinforcement learning (RL) has the potential to greatly accelerate the search for novel materials. Despite recent progress on leveraging graph representations to design molecules, such methods are fundamentally limited by the lack of three-dimensional (3D) information. In light of this, we propose a novel actor-critic architecture for 3D molecular design that can generate molecular structures unattainable with previous approaches. This is achieved by exploiting the symmetries of the design process through a rotationally covariant state-action representation based on a spherical harmonics series expansion. We demonstrate the benefits of our approach on several 3D molecular design tasks, where we find that building in such symmetries significantly improves generalization and the quality of generated molecules.