A year ago, DeepMind caused a stir with the announcement of the ability to solve the main mystery in biology - to predict the structure of any protein. At the same time, DeepMind created an open database with data on 350 thousand proteins, the shape of which was predicted by the AlphaFold package. Today, the company announced the discovery of the forms of almost every protein known to terrestrial science - more than 200 million proteins from all spheres of life known on Earth. This is a real revolution in biology.

Knowledge of the spatial shape of a protein can help find the perfect cure for diseases and make many other discoveries in biology. Before the advent of AI algorithms, scientists experimentally determined the shape of proteins, which is very, very difficult and time consuming. The algorithm proposed by DeepMind determines the spatial shape of one protein from 10 to 20 seconds. Thanks to this, the company was able to bring the database of spatial forms of proteins from 350 thousand to more than 200 million in a year.

It should be clarified that predicting the shape of a protein does not mean being 100% accurate. However, AlphaFold showed significant accuracy in determining shapes, which is enough to get started. All the routine work was done by a computer, and it took all the scientists of the world before that 50 years to unravel about only 10% of protein structures.

https://www.science.org/content/article/new-era-digital-biology-ai-reveals-structures-nearly-all-known-proteins