DeepMind has predicted the structure of almost every protein known to science


According to DeepMind, its AlphaFold tool has accurately predicted the structure of almost all proteins currently understood by science. The library of more than 200 million proteins is now available to everyone for free from the Alphabet-owned AI lab.

In 2020, DeepMind surprised the scientific community by releasing AlphaFold. One of the "great difficulties" of biology was to comprehend the structure of proteins, which are necessary for life. Scientists had been working on this problem for decades. Understanding their form is essential to comprehending how they work.

Nearly every protein found in the human body is represented in the AlphaFold Protein Structure Database, which DeepMind made accessible last year along with the source code for AlphaFold. The European Molecular Biology Laboratory, a global public research organization that currently manages a sizable collection of protein data, collaborated with us to create the database.

The database gets a significant boost from the most recent data release. In a call with reporters this week, Demis Hassabis, founder and CEO of DeepMind, said that the update includes structures for "plants, bacteria, animals, and many, many other organisms, opening up huge opportunities for AlphaFold to have impact on important issues such as sustainability, fuel, food insecurity, and neglected diseases."

Scientists may find the enhanced database to be a valuable resource in their quest to comprehend illnesses. Additionally, it could hasten biological and medicinal discovery innovation.

According to Jian Peng, a professor of computer science at the University of Illinois Urbana-Champaign who specializes in computational biology, "AlphaFold is arguably the most significant gift from the AI community to the scientific community."

Researchers have already utilized AlphaFold since its debut in 2020 to comprehend proteins that impact honeybee health and create a potent malaria vaccine.

According to Hassabis, the database makes it possible for researchers to seek up the 3D structures of proteins "nearly as quickly as running a keyword Google search."

A tool with 200 million publicly available protein structures would save researchers a ton of time since predicting the structures of proteins takes a lot of time, according to Mohammed AlQuraishi, a systems biologist at Columbia University who is not involved in DeepMind's study.

According to Peng, AlphaFold could also enable researchers to reexamine earlier work in order to comprehend how illnesses develop.

AlQuraishi said that the database "will not address" the fact that for many proteins, "we're interested in knowing how their structure is affected by mutations and natural allelic variation." However, he continued, "the area is evolving quickly, so I anticipate tools to precisely predict protein variations will start to surface shortly."

According to Peng, the quality of AlphaFold's predictions could not be as precise for rarer proteins with less known evolutionary data.

In "digital biology," where "AI and computational approaches might assist to comprehend and simulate essential biological processes," according to Hassabis, the move marks the next step in DeepMind's initiative. Hassabis also serves as the CEO of Isomorphic Labs, a brand-new business that is now a part of Alphabet and focuses on using AI to find novel drugs.

The behavior and interactions of proteins with other proteins are only a couple of the difficulties the business still intends to address in the life sciences, according to Pushmeet Kohli, head of AI for research at DeepMind.

Hassabis stated that he hopes AI will play a "major role in the discovery process for novel treatments and cures," rather than only helping in determining the structure of proteins.