ColabFold is an initiative by Milot Mirdita, Sergey Ovchinnikov and Martin Steinegger, providing more user-friendly access to AlphaFold prediction models. The different Google Colab notebooks can be found at

Its main contributions in comparison with the official notebook are as follows:

  • Use of MMSeqs2 for the MSA search, resulting in a significantly faster computation without losing prediction performance.
  • More customizability options: it is possible to enable or disable templates and relaxation, the number of recycling iterations can be modified, and more.
  • More outputs are generated: visualizations for the pLDDT and PAE of each of the five models are added, a visualization of the MSA sequence coverage is attached, and all models and files are downloaded in the final zip file.

However, despite these contributions, the Google Colab environment and its resource allocation remains a limiting factor.

Mirdita M., Ovchinnikov S., Steinegger M. ColabFold – Making protein folding accessible to all biorxiv, doi: 10.1101/2021.08.15.456425.