A Deep-Learning View of Chemical Space Designed to Facilitate Drug Discovery
importance 6/10
ai assisted
Computational Drug Discovery Methods drug discovery stage: research
Problem
Designing small molecules with desired properties relies on human intuition, slowing drug discovery design cycles.
Why it matters
Faster molecular design could shorten drug discovery cycles and reduce costs.
What the AI did
DESMILES, a deep neural network, modifies input molecules toward desired properties, augmenting human design in drug discovery.
Results
77% lower failure rate vs state-of-the-art models on a published dopamine receptor D2 benchmark.
Impact
Benchmark improvement demonstrated on one published task; broader drug-discovery impact not shown.
Record
| Venue | Journal of Chemical Information and Modeling |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Paul Maragakis, Hunter M. Nisonoff, Brian Cole, David Elliot Shaw |
| Citations | 41 (OpenAlex, at import time) |
| References | 35 |
| Model or method | DESMILES (deep neural network) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1021/acs.jcim.0c00321 |
| Open access | hybrid |
| Keywords | deep learning, molecular design, drug discovery, DESMILES, benchmark |
Importance rating
6
significant — thousands to millions affected
Abstract states deep learning could augment human intuition and expedite drug discovery projects; demonstrated on a D2 benchmark with 77% lower failure rate.