Google’s Deepmind new AI predicts folding of proteins

The 13th Critical Assessment of Structure Prediction (competition) took place on 2nd of December 2018. Deepmind showed their latest project AlphaFold during this event.

How AlphaFold works ?

First of all this program gets a sequence of amino acids which later on are analysed thanks to neural network and data base. Then it predicts angels of connection of amino acids and the distance between them. At last it creates 3D model (like the ones above ).

In the contest the AI competed with human teams and made the most accurate prediction 25 times ( out of 43 tests ).

Thanks to predicting how proteins fold, in future we will understand diseases such as Alzheimer’s, Parkinson’s, Huntington’s or  cystic fibrosis and maybe we will figure out how to cure them. We will also gain more knowleage how different proteins could react to one another  or what physical properties they have.

Project AlphaFold is based on previous Alphas (e.g. AlphaGo) which were mainly AI created to beat humans at games and to learn how AI works. Now this AI will help scientists create new ideas, for example how to improve cures, develop new theories or create new materials made of proteins.

AI just a few years ago was something futuristic but now rapidly developes into everyday solution in different fields of our life.



The following two tabs change content below.

Kaczmarek Robert


Latest posts by Kaczmarek Robert (see all)

One thought on “Google’s Deepmind new AI predicts folding of proteins

  1. Avatar Willoński Artur says:

    Good to see AI implemented into biotechnology sector. It is incredible that, today AI open all doors that were closed only because of human imperfections. These makes a great hopes for the future of molecular sciences.

Leave a Reply