Drug Companies’ Data Boosts AI Protein Models
For drug discovery, protein-folding models like AlphaFold face a data shortage. To improve these AI tools, pharmaceutical companies are sharing proprietary datasets of proteins interacting with drugs. A consortium trained an open-source model on over 20,000 such structures and found it outperformed public models by significantly improving prediction accuracy. This suggests that making more of…
1 source primary source
Key points
- Pharmaceutical companies share proprietary protein-ligand interaction data
- OpenFold3 model trained on more than 20,000 internal structures outperforms public models
- AISB Network fine-tunes OpenFold3 on company-specific datasets
Read the original at Nature Machine Learning primary source Open source ↗
Topics · follow one to build your own front page
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us.
More in Generative AI & Models
All →- scTransMIL links single-cell transcriptomics to patient cancer phenotypes · 1 src
- Elon Musk's Long-Standing Warnings on AI Gain Attention · 1 src
- AI Lawsuits Pile Up: Meta Settles $18B, DOJ Upholds Fair Use · 1 src
- OpenAI’s GPT‑6 Astra Claims AGI, Quickly Replicates Existing Games · 2 src
- OpenAI’s GPT‑6 Astra Faces Post‑Launch Quality Drop Claims · 1 src
Comments
via GitHub Discussions