PROTEIN DISTANCE CONSTRAINTS PREDICTED BY NEURAL NETWORKS AND PROBABILITY DENSITY-FUNCTIONS

We predict interatomic C-alpha distances by two independent data drive n methods, The first method uses statistically derived probability dis tributions of the pairwise distance between two amino acids, whilst th e latter method consists of a neural network prediction approach equip ped with windows taking the context of the two residues into account. These two methods are used to predict whether distances in independent test sets were above or below given thresholds, We investigate which distance thresholds produce the most information-rich constraints and, in turn, the optimal performance of the two methods. The predictions are based on a data set derived using a new threshold which defines wh en sequence similarity implies structural similarity. We show that dis tances in proteins are predicted more accurately by neural networks th an by probability density functions. We show that the accuracy of the predictions can be further increased by using sequence profiles, A thr eading method based on the predicted distances is presented. A homepag e with software, predictions and data related to this paper is availab le at http://www.cbs.dtu.dk/services/CPHmodels/.