EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation
- PDF / 3,341,784 Bytes
- 19 Pages / 595.276 x 790.866 pts Page_size
- 32 Downloads / 273 Views
RESEARCH ARTICLE
Journal of Cheminformatics Open Access
EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation Jules Leguy1 , Thomas Cauchy2* , Marta Glavatskikh1,2 , Béatrice Duval1 and Benoit Da Mota1*
Abstract The objective of this work is to design a molecular generator capable of exploring known as well as unfamiliar areas of the chemical space. Our method must be flexible to adapt to very different problems. Therefore, it has to be able to work with or without the influence of prior data and knowledge. Moreover, regardless of the success, it should be as interpretable as possible to allow for diagnosis and improvement. We propose here a new open source generation method using an evolutionary algorithm to sequentially build molecular graphs. It is independent of starting data and can generate totally unseen compounds. To be able to search a large part of the chemical space, we define an original set of 7 generic mutations close to the atomic level. Our method achieves excellent performances and even records on the QED, penalised logP, SAscore, CLscore as well as the set of goal-directed functions defined in GuacaMol. To demonstrate its flexibility, we tackle a very different objective issued from the organic molecular materials domain. We show that EvoMol can generate sets of optimised molecules having high energy HOMO or low energy LUMO, starting only from methane. We can also set constraints on a synthesizability score and structural features. Finally, the interpretability of EvoMol allows for the visualisation of its exploration process as a chemically relevant tree. Keywords: Chemical space exploration, Organic molecular materials Introduction One of the main objectives of chemical research is to find a molecule that has desired properties for a given application. However, the molecular space being immeasurable, one needs to define strategies to efficiently explore its relevant parts. Even an incomplete enumeration of the chemical space limited to 17 heavy atoms (C, N, O, S and halogens) already leads to more than 160 billion compounds [1]. To tackle this problem, we will see that many methods have been proposed, adapting recent advances in deep learning and in reinforcement learning, or using *Correspondence: thomas.cauchy@univ‑angers.fr; benoit.damota@univ‑angers.fr 1 Laboratoire LERIA, UNIV Angers, SFR MathSTIC, 2 Bd Lavoisier, 49045 Angers, France 2 Laboratoire MOLTECH-Anjou, UMR CNRS 6200, UNIV Angers, SFR MATRIX, 2 Bd Lavoisier, 49045 Angers, France
more classical optimisation methods such as evolutionary algorithms. Actually, fully automated de novo molecular generation is a subject that has regained considerable attention [2, 3]. In the 1990–2000s, evolutionary algorithms were already used for de novo molecular generation [4]. To limit the number of steps and to improve the likeliness of the solutions, they were commonly based on the combination of fragments rather than mutating the molecules at atomic level. The interest in evolutionary algor
Data Loading...
