Chemical creativity in the design of synthetic chemical entities with druglike properties has been the domain of medicinal chemists. At the same time, constructive machine learning models have been shown to autonomously sample drug-like molecules from chemical space without the need for explicit design rules. A machine learning method that combines a rule-based approach with a machine learning model was trained on synthetic routes described in chemical patent literature. This unique combination enables a balance between ligand-similarity based generation of innovative compounds by scaffold hopping and forward-synthetic feasibility of the designs. Prospective results demonstrate the capability of this hybrid machine learning model to capture implicit chemical knowledge from chemical reaction data and suggest feasible syntheses of new chemical matter. We will present various applications of molecular de novo design with machine intelligence, and discuss the advantages and limitations of these design concepts.
Button, A., Merk, D., Hiss, J. A., Schneider, G. Nat. Mach. Intell. 2019, 1, 307–315.
Schneider, G. Nat. Mach. Intell. 2019, 1, 128–130.
Schneider, G., Clark, D. Angew. Chem. Int. Ed. 2019, 58, 10792–10803.
Schneider, G. Nat. Rev. Drug Discov. 2018, 17, 97–113.