A promising start

Interesting science news this week about Google's company "DeepMind" who have been using a neural network (AI software) to attempt to solve the Biological riddle that is protein folding. We have been able to discover the molecules that make up proteins for many years now, but this simplistic one dimensional view, i.e. essentially a list indicating the order in which particular amino acids are linked together, only tells part of the story. Proteins are three dimensional structures, not flat lists, and their shapes often dictate what they do or don't do, computing the three dimensional shape from a flat list has been somewhat of a holy grail of Biology and Medicine for decades. DeepMind took a large database of protein shapes and trained a specially designed neural net using the data, the network then predicted the 3d structures of proteins with a roughly 80% accuracy, importantly, it only takes minutes to find the answers whereas traditional methods (such as x-ray crystallography) may take years. Of course this isn't the end of the puzzle, even with the huge resources that Google has and the outrageous cash that they pay their AI gurus, their tech isn't able to handle the computation of complexes and interactions of proteins that work together so there's still some way to go, but it's a promising start!