The gospel of scientists, two AI algorithms for predicting protein structure are open source

On July 15, London-based DeepMind released the open source version of its deep learning neural network AlphaFold 2, and published its method in the journal Nature

At the same time, the research team at the University of Washington School of Medicine was inspired by AlphaFold 2 and developed its own protein structure prediction tool

Jinbo Xu, a computational biologist at the University of Chicago, believes that the open source nature of these tools means that the scientific community should be able to develop more powerful and useful software based on existing advances

Proteins are made up of a long string of amino acids, and when they are folded into a 3D shape, they determine the function of these proteins in the cell

Last year, DeepMind caused an uproar in the scientific community.

DeepMind has been tight-lipped about its work, only introducing AlphaFold 2 in a short speech on CASP

David Baker, a biochemist at the University of Washington in Seattle, said: “There is a lot of pessimism in academia

As a result, Baker and his colleagues started brainstorming ideas to find ways to replicate the success of AlphaFold 2

RoseTTaFold performed almost as well as AlphaFold 2, and it was much better than other entries in the protein prediction competition

In addition, DeepMind has simplified AlphaFold 2

Although the source code of AlphaFold 2 is freely available, it may not be particularly useful for researchers without professional skills

In addition to providing RoseTTaFold code for free, Baker's team also set up a server where researchers can insert protein sequences and get predicted structures

As RoseTTaFold and AlphaFold 2 both begin to provide code for free, researchers will be able to build on these two advancements

##

Jumper, J.

Baek, M.