Former Google subsidiary DeepMind has created 3D models of complex proteins, a task long viewed as one of the "core challenges in biology" and hugely significant to medication development.
In a blog released by DeepMind, the AI company explained that due to the fact that "proteins are large, complex molecules," protein folding has been a big problem for biologists. The structure and how the amino acids which make up the protein is "folded" also largely determines how it will function and interact with other molecules. This not only includes actions such as how our muscles will work or how we turn food into fuel but misfolding is what are believed to cause ailments such as Parkinson's, Alzheimer's, Huntington's and cystic fibrosis.
"But figuring out the 3D shape of a protein purely from its genetic sequence is a complex task that scientists have found challenging for decades," said the blog and this is exactly what DeepMind's AI system AlphaFold has managed to achieve. Due to the increase in data available to AI scientists due to the dramatic fall in the cost of genetic sequencing, "biologists are turning to AI methods as an alternative to this long and laborious process for difficult proteins".
DeepMind has been working on this problem for the last two years and has made "unprecedented progress in the ability of computational methods to predict protein structure". The firm has been utilizing deep neural networks to model the 3D structure of proteins and predicting its properties.
"The properties our networks predict are: (a) the distances between pairs of amino acids and (b) the angles between chemical bonds that connect those amino acids," explained the blog. "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."