tareensoft
Computer scientists have developed an artificial intelligence system that can accurately predict three-dimensional protein structures using significantly less data than previous methods. The new algorithm, called DeepFold, requires only partial sequence information to generate reliable structural models of complex proteins. “This represents a major advance in computational biology,” says Dr. James Wong of the Molecular Computing Institute. The system could accelerate drug discovery by allowing researchers to quickly determine the structures of therapeutic targets even when complete sequence data isn’t available. Initial tests show DeepFold achieves 85% accuracy with just 30% of the data required by earlier systems.
