Donate to Science & Enterprise

S&E on Mastodon

S&E on LinkedIn

S&E on Flipboard

Please share Science & Enterprise

Alphabet Forms AI Drug Discovery Company

Artificial intelligence graphic

(Seanbatty, Pixabay)

5 Nov. 2021. Alphabet, the parent company of Google, is forming a new enterprise using artificial intelligence techniques to discover new protein-based drugs. Isomorphic Labs, in London, is spun-off from DeepMind, an A.I. research subsidiary of Alphabet, which began focusing on life sciences in recent years.

Isomorphic Labs plans to apply DeepMind’s advances in protein folding to design new therapies. Proteins are complex chains of amino acids created by cells with instructions from an individual’s genetic code. A protein’s functions are determined by its three-dimensional structure, which in turn is the result of folding patterns of those amino acid chains. Errors in those folding patterns can result in toxic or malfunctioning proteins, and cause disease.

DeepMind began in London in 2010, with the goal of solving difficult problems with human-level intelligence. This refined level of intelligence, sometimes called artificial general intelligence, seeks to create models human brain activity that incorporate higher-level understanding, such as creativity and problem solving. DeepMind is based on deep learning algorithms generated from a neural network that mimics functions of neurons in the human brain. Google acquired DeepMind in 2014, which operates today as an Alphabet company.

“Biology can be thought of as an information processing system”

For protein research, DeepMind built its AlphaFold platform to understand, model, and predict protein structure and design. At a scientific community problem-solving experiment in protein design in 2020, AlphaFold demonstrated its ability to accurately predict a protein’s eventual structure from a sequence of amino acids, to atomic levels of detail. As described in a paper in the journal Nature in July 2021, AlphaFold could predict a protein’s detailed structure, even if no similar folding pattern existed. Moreover, AlphaFold completed these tasks much faster that before. In a companion paper, also published in Nature, the DeepMind team applied AlphaFold to model the human collection of proteins known as proteome, predicting 98.5 percent of all human proteins.

DeepMind makes the AlphaFold source code available to the public on GitHub. The AlphaFold protein structure database is accessible through the European Bioinformatics Institute.

Isomorphic Labs plans to apply what it calls digital biology to drug discovery, designing therapeutic proteins with A.I., as well as medicinal chemistry, biophysics, and engineering. Demis Hassabis, founder and CEO of Isomorphic Labs says in a company blog post:

At its most fundamental level, I think biology can be thought of as an information processing system, albeit an extraordinarily complex and dynamic one. Taking this perspective implies there may be a common underlying structure between biology and information science — an isomorphic mapping between the two — hence the name of the company.

He adds, “But just as mathematics turned out to be the right description language for physics, biology may turn out to be the perfect type of regime for the application of AI.” Hassabis is also co-founder and CEO of DeepMind, and expects to hand over the reins to a permanent CEO later on.

More from Science & Enterprise:

*     *     *

Comments are closed.