GuacaMol: Benchmarking Models for de Novo Molecular Design

• DOI: 10.1021/acs.jcim.8b00839
• Publication Date: 2019-03-19T14:28:42Z
• Authors: Nathan Brown; Marco Fiscato; Marwin H.S. Segler; Alain C. Vaucher
• Keywords: Benchmarking; Python; Benchmark (surveying); Chemical space

De novo design seeks to generate molecules with required property profiles by virtual design-make-test cycles. With the emergence of deep learning and neural generative models in many application areas, for molecular based on networks appeared recently show promising results. However, new have not been profiled consistent tasks, comparative studies well-established algorithms only seldom performed. To standardize assessment both classical de design, we propose an evaluation framework, GuacaMol, a suite standardized benchmarks. The benchmark tasks encompass measuring fidelity reproduce distribution training sets, ability novel molecules, exploration exploitation chemical space, variety single multiobjective optimization tasks. benchmarking open-source Python code leaderboard can be found https://benevolent.ai/guacamol .