Advances in De Novo Drug Design: From Conventional to Machine Learning Methods
Problem
De novo drug design, generating novel molecular structures from atomic building blocks, is difficult and costly with conventional structure- or ligand-based methods.
Why it matters
Drug discovery is slow and expensive, and AI-generated molecules could reduce cost and time to new medicines.
What the AI did
This review article summarizes how deep reinforcement learning with recurrent neural networks, convolutional neural networks, GANs, and autoencoders has been employed to develop novel de novo drug design approaches.
Results
not stated in source
Impact
Widely cited review (352 citations) summarizing AI's positive impact on drug discovery; no original experiments.
Record
| Venue | International Journal of Molecular Sciences |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Varnavas D. Mouchlis, Antreas Afantitis, Angela Serra, Michele Fratello, Anastasios G. Papadiamantis, Vassilis Aidinis, Iseult Lynch, Dario Greco et al. |
| Citations | 352 (OpenAlex, at import time) |
| References | 162 |
| Model or method | deep reinforcement learning with RNNs, CNNs, GANs, autoencoders (reviewed) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3390/ijms22041676 |
| Open access | gold |
| Keywords | de novo drug design, machine learning, reinforcement learning, generative models, review |
Importance rating
significant — thousands to millions affected
Drug discovery is a major industry affecting millions; abstract frames AI as positively impacting the process but states no scale or results.