Publications, GitHub code and database

One of the key aspects in the widespread interest and utility of AlphaFold is the fact that DeepMind decided to share all details, prediction models and code. This open sourcing provides a solid base for various applications, refinements and interpretation of the system. The following information can be directly found online:

1. Official publication

In July, DeepMind published the AlphaFold2 publication in Nature. Together with supplementary material of 61 pages, it contains the full architectural and training details. Additionally, in October, a follow-up paper was published on bioRxiv, called AlphaFold-Multimer, describing refinements of the prediction models for protein interaction modeling.

Jumper, J., Evans, R., Pritzel, A. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021). https://doi.org/10.1038/s41586-021-03819-2

Evans, R., O’Neill, M., Pritzel, A., et al. 2021. Protein complex prediction with AlphaFold-Multimer. bioRxiv (2021). https://doi.org/10.1101/2021.10.04.463034

2. GitHub code

All code necessary to run AlphaFold, along with model parameters, installation instructions, data download instructions and more, are found on their official GitHub page (https://github.com/deepmind/alphafold). Note that this software is very resource-intensive and storage-intensive (almost 2 TB of preferably fast storage is needed), so you will not immediately be running this on your personal computer.

3. AlphaFold Protein Structure Database

In collaboration with EMBL-EBI, a database was set up with AlphaFold predictions for all human proteins, as well as initially some 20 other organisms. By the end of 2021, the database size had already been doubled by covering almost the full SwissProt database. More information at https://alphafold.ebi.ac.uk/.

Tunyasuvunakool, K., Adler, J., Wu, Z. et al. Highly accurate protein structure prediction for the human proteome. Nature 596, 590–596 (2021). https://doi.org/10.1038/s41586-021-03828-1