Generative deep learning enables the discovery of a potent and selective RIPK1 inhibitor
Problem
Retrieving hit/lead compounds with novel scaffolds for a given biological target in early drug development is hard, and generative models' capacity to produce wet-lab-validated, target-specific molecules was largely unverified.
Why it matters
Novel-scaffold leads for targets like RIPK1 could seed new treatments for inflammatory diseases.
What the AI did
A generative deep learning model (conditional RNN) generated tailor-made virtual compound libraries for RIPK1; virtual screening and wet-lab bioactivity testing yielded inhibitor RI-962.
Results
RI-962: potent, selective RIPK1 inhibitor with a previously unreported scaffold; potent in vitro protection from necroptosis and good in vivo efficacy in two inflammatory models.
Impact
Demonstrates generative deep learning can produce wet-lab-validated, novel-scaffold lead compounds for a chosen target.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Yueshan Li, Liting Zhang, Yifei Wang, Jun Zou, Ruicheng Yang, Xinling Luo, Chengyong Wu, Wei Min Yang et al. |
| Citations | 101 (OpenAlex, at import time) |
| References | 69 |
| Model or method | distribution-learning conditional recurrent neural network (cRNN) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-022-34692-w |
| Open access | gold |
| Keywords | generative deep learning, drug discovery, RIPK1, inhibitor, cRNN, necroptosis |
Importance rating
significant — thousands to millions affected
Abstract calls hit/lead retrieval 'important but challenging'; wet-lab-validated novel-scaffold generation for a chosen target is a genuinely new capability for drug discovery, but only one target (RIPK1) is demonstrated.