DeepPotential is a method for protein inter-residue geometry and full-length 3D structure prediction. For a query sequence, DeepPotential starts with the collection of deep multiple sequence alignments (MSAs) through whole-genome and metagenome sequence databases. Next, a complimentary set of coevolutionay feature matrices extracted from the selected MSAs and are used to predict geometry maps with deep multi-tasking ResNet. The full-length structure is quickly constructed by PotentialFold through the limited-memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS) algorithm.
yangzhanglabumich.edu | (734) 647-1549 | 100 Washtenaw Avenue, Ann Arbor, MI 48109-2218