Conventional methods to predict 3D molecular structure are based on iterative stochastic optimization techniques by calculating energy using physics-based electronic structure modeling such as Density Functional Theory (DFT) or Molecular Dynamics (MD) which is computationally expensive and multi-modal by in nature. As a cost-efficient and relatively deterministic alternative, we propose a novel RL-based algorithm to optimize 3D structure of single molecule based on DDPG method. Our model teaches the agent to find the best movement trajectory of each atom to reach the correct structure by guiding their movement using three reward approaches based on DFT calculation. Our experiment shows that our model successfully predicts the most energetically stable 3D structure of small aromatic-hydrocarbon molecules.

Conventional methods to predict 3D structure of molecule are based on iterative stochastic optimization techniques based on energy calculation using physics-based electronic structure modeling such as DFT or MD. Therefore, computing cost of physics-based modeling is significantly depended by the number of iterations to calculate energy until the total energy of structure is converged. As the cost-efficient alternatives, we propose a novel RL-based algorithm to optimize 3D structure of single H2O molecule based on DDPG (Deep Deterministic Policy Gradient) method. To demonstrate the efficiency of our model, we predicted 3D structure of H2O molecule and compared with results from the conventional DFT calculation. Our experiments show that our model succeed to predict 3D structure of H2O molecule which is identical with the results from DFT calculation.