Biotech labs have initiated to use of a type of generative AI known as a diffusion model to generate designs for previously unseen types of protein. Separately, two labs announced programmes that use diffusion models to generate more precise designs for novel proteins than ever before. Baker and his colleagues demonstrate in an online preprint paper that their model can generate precise designs for novel proteins that can then be brought to life in the lab.
These protein generators can be programmed to generate protein designs with specific properties, such as shape, size, or function. In effect, this enables the development of new proteins to perform specific tasks on demand. What distinguishes this work is the generation of proteins based on desired constraints says Ava Amini, a biophysicist at Microsoft Research in Cambridge, Massachusetts.
Why are proteins the main focus?
Proteins are the basic building blocks of all living systems. Proteins play a role in people getting sick. Proteins are thus ideal drug targets. Furthermore, many of today’s newest drugs are protein-based. The primary objective of protein generation is to add a nearly infinite number of computer-designed proteins to that list.
RoseTTAFold Diffusion makes use of information from a separate neural network trained to predict protein structure about how the pieces of a protein fit together. They can produce proteins with varying degrees of symmetry, such as circular, triangular, or hexagonal proteins. Because the shape of a protein determines its function, the ability to generate different structures on demand is critical. It’s one thing to create bizarre designs on a computer. However, the ultimate goal is to convert these designs into actual proteins.
Future aspirations
Baker sees the main benefit as the creation of a new protein that binds to the parathyroid hormone, which regulates calcium levels in the blood. We basically just gave the model the hormone and told it to make a protein that binds to it, and it created this protein design out of thin air, Baker says. Grigoryan acknowledges that developing new proteins is only the first of many steps. At the end of the day, what matters is whether or not we can create medicines that work. Protein-based drugs must be mass-produced, then tested in the lab, and finally in humans.
