Alphabet lets AI take the lead role in new drug discovery startup
Last year, DeepMind produced a stunning advance in scientific research, demonstrating how its artificial intelligence could be used to predict 3D structures of unique proteins in a solution to a 50-year science problem. This has moved parent company Alphabet to spin-off a new venture named Isomorphic Labs that will leverage the technology for the purposes of drug discovery and biomedical breakthroughs.
The problem at the heart of DeepMind's major breakthrough begins with one-dimensional chains of amino acids that can fold into an almost endless number of intricate 3D structures. Predicting what that unique structure will look like based on the amino acid chain alone is known as the protein folding problem, and it's one scientists had been working to solve since the 1970s.
DeepMind's AI system designed to tackle this specific problem is called AlphaFold. After training it on publicly available protein structures already determined through scientific experiments, the AI was able to solve protein structures that scientists had been working on for many years, with the results described as "astonishingly accurate."
This was heralded as a new era in biological research, as the tool can help scientists identify malfunctioning proteins with great expediency and, with them, the underlying causes of certain diseases, which could greatly accelerate the development of novel drugs. Earlier this year, we saw the technology prove its worth further by predicting structures for nearly every protein in the human body.
Alphabet is now building on this progress by launching Isomorphic Labs, a name it says is inspired by the notion that biology and information science could share a common underlying structure. Just like mathematics is the language used to describe physics, AI may be the ideal interpreter for biology, says DeepMind and Isomorphic Labs founder Demis Hassabis.
The spin-off will partner with pharmaceutical companies and is on a mission to reimagine drug discovery from the ground up with an AI-first approach, according to Hassabis. Further, it is hoped to help us better model and understand some of the fundamental mechanisms of life, and perhaps "find cures for some of humanity’s most devastating diseases."
Source: Isomorphic Labs