Alphabet’s DeepMind AI has won a protein-folding contest by creating an algorithm that was able to predict the complex 3D shapes into which proteins can be folded. Understanding how proteins are shaped means understanding the fundamental molecules of life, since almost all functions in the human body can be traced back to the shape and movement of proteins.
What does this mean?
A protein consists of a chain of different amino acids that are folded in an energy-efficient way. Diseases such as Alzheimer’s, Parkinson’s and diabetes result from a misfolding of proteins. Understanding the shapes of proteins and how they come to be misfolded would thus provide insight into diseases and is an important step in preventing/curing these diseases. Earlier, we wrote about how citizen science was a novel way to predict protein structures in a faster way than scientists were able to. To predict these protein 3D structures, unusual computing power is needed and DeepMind’s success shows that great gains have been made in this domain: it took the program only hours to predict the shapes. Enhanced computing power and AI are leading to new discoveries and cheaper ways to predict diseases.
Quicker and better insight into proteins can serve a wide range of applications both in and beyond healthcare. For instance, understanding specific proteins could enable us to apply their sequence-snipping capabilities to modify DNA-strings (e.g. CRISPR) or to design new proteins to perform different tasks (breaking down plastic pollution in the environment).