AlphaFold 3 is out! As expected expands coverage to small molecules and nucleic acids. And replaces the structure module with a diffusion-based one. Unfortunately no code or model weights--just a web server for a limited set of ligands:
@MoAlQuraishi
10 searchers per day on a webserver, that should not even be allowed for a method's paper. I get the limitations for commercial use but the gated use through a server with such low volume usage totally sucks for most academic research applications.
@MoAlQuraishi
Not releasing weights was expected but not releasing even code is pretty bad. What is the point of publishing in a scientific journal then, as opposed to the existing white paper? This is pretty close to a paid promotion imho
@MoAlQuraishi
All algorithms seem to be well described in the Supplementary Information document. Looking forward to see the OpenFold implementation soon
@MoAlQuraishi
How can I reproduce AF3 results when only 10 predictions are allowed on the Alphafold Server and the source code is not provided? Science? Fiction?