DrugGEN Paper on Generative AI for Drug Design Published in Nature Machine Intelligence

A paper detailing “DrugGEN,” a novel system for designing target-specific drug candidate molecules using generative AI, has been published in Nature Machine Intelligence, one of the world’s leading scientific journals for artificial intelligence.

The research was conducted by a team from the Hacettepe Biological Data Science Lab., led by Prof. Dr. Tunca Doğan from Hacettepe University’s Department of Computer Engineering and Head of the Institute of Informatics, Health Informatics A.B.D.

The DrugGEN system presents a novel architecture that, for the first time, integrates Generative Adversarial Networks (GANs) with graph transformer-based deep learning methods. This combined architecture leverages both the powerful feedback mechanism of GANs and the sophisticated modeling capabilities of graph transformers to automatically design de novo drug candidate molecules conditioned on a specific protein target.

Contribution to Cancer Research

In the study, the team targeted the AKT1 protein, which is associated with many different types of cancer. The AI designed thousands of potential molecules, five of which were synthesized and tested in the laboratory. Two of these molecules successfully suppressed the target protein at a significant level. Computational analyses, including attention maps and molecular dynamics, confirmed that the model was able to capture the intended target-specific interactions.

Why It Matters

The study demonstrates that generative AI not only provides speed and cost advantages in the drug discovery pipeline but can also identify completely novel molecule candidates that have not been discovered before. This capability promises significant contributions to the development of new treatments for diseases.

In the interest of scientific transparency and public benefit, all code, pre-trained models, datasets, and result outputs have been made available via open access.

The research was led by Hacettepe University with participation from METU and Gazi University, and was supported by the TÜBİTAK 2247 National Leading Researchers Program. The study’s completion entirely within Turkey highlights the country’s capability for high-quality research at the intersection of generative AI and health.

Links

• Article Link: https://www.nature.com/articles/s42256-025-01082-y
• Free Full-Text Access: https://rdcu.be/eGoHv
• Source Code, Data, Models: https://github.com/HUBioDataLab/DrugGEN
• Demo: https://huggingface.co/spaces/HUBioDataLab/DrugGEN