|
Listen to this story
|
Life sciences have undergone a radical change as a result of the artificial intelligence-powered protein model prediction. The Netherlands Cancer Institute’s biochemists present AlphaFill, an algorithm that ‘transplants’ missing small molecules and ions from experimentally determined structures to protein-prediction models using sequence and structure similarity.
None of the existing protein models in the AlphaFold protein structure database contain coordinates for small molecules necessary for their molecular structure or function. For example, haemoglobin does not contain bound heme. Additionally, there are no ligands or cofactors, and none of the ATPases or kinases is associated with ADP or ATP.
AlphaFold and RoseTTAFold models predict domain structures almost accurately, but flexible parts of the protein—such as loops or intrinsically disordered regions—are predicted with lower accuracy. The artificial intelligence prediction models have learned the inherent rules of protein folding based on extensive training on experimentally resolved structures. However, many proteins do not occur in nature without their cofactor: myoglobin or haemoglobin need a heme to fold. Predicted structural models exclusively account for the 20 canonical amino-acid residues and do not predict the coordinates for small molecules, ligands and cofactors.
AlphaFill is an updated version of the models in the AlphaFold database that transplant small molecules and ions that have been experimentally observed in homologous protein structures. The AlphaFill procedure has been applied to all AlphaFold models and validated against experimental structures to produce a new resource, the AlphaFill databank, which is intended to make it simple for life scientists to develop novel hypotheses for protein function and frame pertinent research questions.
Click here to learn more about AlphaFill.
Besides AlphaFill, Chinese scientist Li Ziqing and Westlake University’s School of Engineering recently claimed to have developed an AI model ProtMD that claims to predict the dynamic structural changes of protein molecules. It claims that it can predict conformation proteins formed in different physiological environments. The algorithm of ProtMD is said to calculate the movements of a protein at the atomic level, generating data based on molecular dynamics.
Last week, Generate Biomedicines, and David Baker‘s Group came up with Chroma and RoseTTAFold Diffusion, which are the new protein-synthesis methods. They used text-to-image diffusion models for the same.
