Google’s DeepMind AI unit has solved one of the holy grails of biology – the protein folding problem.
This is where, given a sequence of amino acids, a computer will be able to predict the final shape of the protein it codes for.
The company won the 14th Critical Assessment of Protein Structure Prediction, or CASP, contest by showing it can accurately predict the final structure of 100 sequences, with results in many cases as good as structures determined by actual physical experiments.
“We have been stuck on this one problem – how do proteins fold up – for nearly 50 years,” said Professor John Moult, CASP chair and co-founder, in a DeepMind blog post. “To see DeepMind produce a solution for this, having worked personally on this problem for so long and after so many stops and starts, wondering if we’d ever get there, is a very special moment.”
CASP has been running since 1994 but it was only in recent years with the advent of AI that real progress has been made.
Google’s Alphafold has been the ultimate expression of that strategy, using 128 of Google TPUv3 cores to crack the code.
The technology, which is not quite perfected yet for the hardest problems, can ultimately be used for problems such as intelligent drug design and other hard biology problems.
See Google’s presentation on their victory below and read all the details at their site here.