ResPRE is an algorithm for protein residue-residue contact-map prediction. Starting from a query sequence, multiple sequence alignments (MSAs) are collected from sequence databases. The inverse covariance matrix, or precision matrix, of the MSAs are then used to train the contact models through deep residual convolutional neural networks. The major advantage of ResPRE is in the use of precision matrix which helps to rule out translational noises of covariance correlations, while the the residual network with parallel shortcut layer connections increases the learning efficiency of the deep convolutional neural network training.
yangzhanglabumich.edu | (734) 647-1549 | 100 Washtenaw Avenue, Ann Arbor, MI 48109-2218