Prospective de novo drug design with deep interactome learning
Problem
Designing new drug-like molecules from scratch with specific bioactivity, synthesizability and structural novelty is a core challenge in drug discovery.
Why it matters
Generating tailored compound libraries for human drug targets could accelerate development of new medicines.
What the AI did
Combined graph neural networks and chemical language models in an interactome-based deep learning approach for zero-shot ligand- and structure-based molecule generation, applied to PPAR-gamma.
Results
Top-ranking AI-designed PPAR-gamma ligands were chemically synthesized and computationally, biophysically and biochemically characterized.
Impact
Demonstrated prospective AI-designed, synthesized and characterized ligands for a human drug target.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Kenneth Atz, Leandro Cotos, Clemens Isert, M. Håkansson, Dorota Focht, Mattis Hilleke, David F. Nippa, Michael Iff et al. |
| Citations | 95 (OpenAlex, at import time) |
| References | 117 |
| Model or method | Graph neural networks plus chemical language models (deep interactome learning) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-024-47613-w |
| Open access | gold |
| Keywords | drug design, de novo generation, graph neural networks, PPAR gamma, zero-shot |
Importance rating
significant — thousands to millions affected
Core challenge in drug discovery; abstract claims enabling creation of innovative bioactive molecules for human drug targets, but gives no scale numbers.