Explaining and avoiding failure modes in goal-directed generation of small molecules
Abstract Despite growing interest and success in automated in-silico molecular design, questions remain regarding the ability of goal-directed generation algorithms to perform unbiased exploration of novel chemical spaces. A specific phenomenon has recently been highlighted: goal-directed generation...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
BMC
2022-04-01
|
Series: | Journal of Cheminformatics |
Subjects: | |
Online Access: | https://doi.org/10.1186/s13321-022-00601-y |
_version_ | 1811311692745801728 |
---|---|
author | Maxime Langevin Rodolphe Vuilleumier Marc Bianciotto |
author_facet | Maxime Langevin Rodolphe Vuilleumier Marc Bianciotto |
author_sort | Maxime Langevin |
collection | DOAJ |
description | Abstract Despite growing interest and success in automated in-silico molecular design, questions remain regarding the ability of goal-directed generation algorithms to perform unbiased exploration of novel chemical spaces. A specific phenomenon has recently been highlighted: goal-directed generation guided with machine learning models produce molecules with high scores according to the optimization model, but low scores according to control models, even when trained on the same data distribution and the same target. In this work, we show that this worrisome behavior is actually due to issues with the predictive models and not the goal-directed generation algorithms. We show that with appropriate predictive models, this issue can be resolved, and molecules generated have high scores according to both the optimization and the control models. |
first_indexed | 2024-04-13T10:23:25Z |
format | Article |
id | doaj.art-dfdfc3492b0846e8a4ea19633e34131f |
institution | Directory Open Access Journal |
issn | 1758-2946 |
language | English |
last_indexed | 2024-04-13T10:23:25Z |
publishDate | 2022-04-01 |
publisher | BMC |
record_format | Article |
series | Journal of Cheminformatics |
spelling | doaj.art-dfdfc3492b0846e8a4ea19633e34131f2022-12-22T02:50:26ZengBMCJournal of Cheminformatics1758-29462022-04-0114111310.1186/s13321-022-00601-yExplaining and avoiding failure modes in goal-directed generation of small moleculesMaxime Langevin0Rodolphe Vuilleumier1Marc Bianciotto2Molecular Design Sciences - Integrated Drug DiscoveryPASTEUR, Département de chimie, École Normale Supérieure, PSL University, Sorbonne Université, CNRSMolecular Design Sciences - Integrated Drug DiscoveryAbstract Despite growing interest and success in automated in-silico molecular design, questions remain regarding the ability of goal-directed generation algorithms to perform unbiased exploration of novel chemical spaces. A specific phenomenon has recently been highlighted: goal-directed generation guided with machine learning models produce molecules with high scores according to the optimization model, but low scores according to control models, even when trained on the same data distribution and the same target. In this work, we show that this worrisome behavior is actually due to issues with the predictive models and not the goal-directed generation algorithms. We show that with appropriate predictive models, this issue can be resolved, and molecules generated have high scores according to both the optimization and the control models.https://doi.org/10.1186/s13321-022-00601-yDe novo designGenerative modelsGoal-directed generationRecurrent neural networkReinforcement learningFailure modes |
spellingShingle | Maxime Langevin Rodolphe Vuilleumier Marc Bianciotto Explaining and avoiding failure modes in goal-directed generation of small molecules Journal of Cheminformatics De novo design Generative models Goal-directed generation Recurrent neural network Reinforcement learning Failure modes |
title | Explaining and avoiding failure modes in goal-directed generation of small molecules |
title_full | Explaining and avoiding failure modes in goal-directed generation of small molecules |
title_fullStr | Explaining and avoiding failure modes in goal-directed generation of small molecules |
title_full_unstemmed | Explaining and avoiding failure modes in goal-directed generation of small molecules |
title_short | Explaining and avoiding failure modes in goal-directed generation of small molecules |
title_sort | explaining and avoiding failure modes in goal directed generation of small molecules |
topic | De novo design Generative models Goal-directed generation Recurrent neural network Reinforcement learning Failure modes |
url | https://doi.org/10.1186/s13321-022-00601-y |
work_keys_str_mv | AT maximelangevin explainingandavoidingfailuremodesingoaldirectedgenerationofsmallmolecules AT rodolphevuilleumier explainingandavoidingfailuremodesingoaldirectedgenerationofsmallmolecules AT marcbianciotto explainingandavoidingfailuremodesingoaldirectedgenerationofsmallmolecules |