Generative and reinforcement learning approaches for the automated de novo design of bioactive compounds

Maria V. Korshunova, Niles Huang, Stephen J. Capuzzi, Dmytro S. Radchenko, Olena Savych, Yurii S. Moroz, Carrow I. Wells, Timothy Mark Willson et al.

Communications Chemistry 2022 article cited by 101

Source Open-access PDF More in Science & Research

importance 6/10 partially solved Computational Drug Discovery Methods drug discovery stage: research

Problem

Deep generative models with reinforcement learning for designing bioactive molecules suffer from sparse rewards, since most generated molecules are predicted inactive.

Why it matters

Automated design of novel bioactive compounds could substantially speed up costly drug discovery.

What the AI did

A deep generative recurrent neural network enhanced by reinforcement learning and several proposed technical tricks was used to design EGFR inhibitors, which were then experimentally validated for potency.

Results

designed EGFR inhibitors experimentally validated for potency; no numeric values in abstract

Impact

Proof-of-concept demonstrates AI-designed compounds can be experimentally active, with expected improvement in finding novel bioactive compounds.

Record

Venue	Communications Chemistry
Published	2022
Work type	article
Authors	Maria V. Korshunova, Niles Huang, Stephen J. Capuzzi, Dmytro S. Radchenko, Olena Savych, Yurii S. Moroz, Carrow I. Wells, Timothy Mark Willson et al.
Citations	101 (OpenAlex, at import time)
References	44
Model or method	deep generative recurrent neural network with reinforcement learning
Organization	not stated in source
Primary source	https://doi.org/10.1038/s42004-022-00733-0
Open access	gold
Keywords	generative models, reinforcement learning, de novo design, EGFR inhibitors, drug discovery

Importance rating

6 / 10

significant — thousands to millions affected

Automated design of bioactive compounds could speed drug discovery; abstract claims proposed solutions expected to substantially improve success rate of finding novel bioactive compounds, but only proof-of-concept with EGFR inhibitors, no scale stated.

Basis: inferred. Judged from the source text, which gives no explicit scale. Stated scale: not stated in source. Method and limits.

More in Science & Research

#	Paper	Year	Citations	Importance
1	Accelerating drug target inhibitor discovery with a deep generative foundation model Vijil Chenthamarakshan, Samuel C. Hoffman, David Owen, Petra Lukacik,...	2023	51	8
2	Discovery of senolytics using machine learning Vanessa Smer-Barreto, Andrea Quintanilla, Richard J. R. Elliott, John...	2023	130	8
3	Deep Learning Pipeline for Accelerating Virtual Screening in Drug Discovery Fatima Noor, Muhammad Junaid, Atiah H. Almalki, Mohammed Almaghrabi, S...	2024	54	8
4	Matrix Factorization-based Technique for Drug Repurposing Predictions Gaia Ceddia, Pietro Pinoli, Stefano Ceri, Marco Masseroli	2020	56	8
5	A Systematic Review of Deep Learning Methodologies Used in the Drug Discovery Process with Emphasis on In Vivo Validation Nikoletta-Maria Koutroumpa, Konstantinos D. Papavasileiou, Anastasios...	2023	35	8