DeepMind AI predicts 3D protein shapes
3 Dec 2018 by qwqcdhhcbc.d qwqcdhhcbc.d
A British technology company has used AI to predict the 3D structure of a protein based on its genetic sequence.
DeepMind used its AlphaFold AI system focused on modelling target shapes from scratch, without using previously solved proteins as templates.
The company trained a neural network to predict physical properties of a protein structure, including distances between pairs of amino acids and the angles between chemical bonds that connect those amino acids.
“The success of our first foray into protein folding is indicative of how machine learning systems can integrate diverse sources of information to help scientists come up with creative solutions to complex problems at speed.” DeepMind said in a blog post.
“We are especially excited about how it might improve our understanding of the body and how it works, enabling scientists to design new, effective cures for diseases more efficiently.”
The network predicted a separate distribution of distances between every pair of residues in a protein, which were then combined into a score that estimates how accurate a proposed protein structure is.
A protein’s shape is fundamental to diagnosing and treating diseases believed to be caused by misfolded proteins, such as Alzheimer’s, Parkinson’s, Huntington’s, and cystic fibrosis.
AlphaFold has been in development for the past two years and builds on prior research in using genomic data to predict protein structure.
DeepMind came first in the Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP), the biennial global competition for assessing protein-folding techniques.
