Quantum Chemistry Driven Molecular Inverse Design with Data-free Reinforcement Learning

The inverse design of molecules has challenged chemists for decades. In the past years, machine learning and artificial intelligence have emerged as new tools to generate molecules tailoring desired properties, but with the limit of relying on models that are pretrained on large datasets. Here, we present a data-free generative model based on reinforcement learning and quantum mechanics calculations. To improve the generation, our software is based on a five-model reinforcement learning algorithm designed to mimic the syntactic rules of an original ASCII encoding based on the SMILES one, and here reported. The reinforcement learning generator is rewarded by on-the-fly quantum mechanics calculations within a computational routine addressing conformational sampling. We demonstrate that our software successfully generates new molecules with desired properties finding optimal solutions for problems with known solutions and (sub)optimal molecules for unexplored chemical (sub)spaces, jointly showing significant speed-up to a reference baseline.