← Back to all posts
Drug firms' secret data supercharge AI protein models
🧬

Drug firms' secret data supercharge AI protein models

Here's something really exciting - pharmaceutical companies are sitting on thousands of protein structures that could transform AI tools like AlphaFold. A consortium of drug firms just trained an improved model on over 20,000 proprietary protein structures, and the results are genuinely impressive compared to models trained only on public data. The thing is, these companies keep this data locked away because it's connected to their drug-development efforts, but now they're realizing that sharing it could help everyone build better AI tools. And that's opening the door to a whole new approach - generating publicly available datasets, like the OpenBind project, that could supercharge protein-folding AI for the entire research community.
Read the original article ↗
Share