Sequence-based Prediction of Metamorphic Behavior in Proteins

Abstract

An increasing number of proteins have been demonstrated in recent years to adopt multiple three-dimensional folds with different functions. These metamorphic proteins are characterized by having two or more folds with significant differences in their secondary structure, where each fold is stabilized by a distinct local environment. So far, about 90 metamorphic proteins have been identified in the Protein Databank (PDB), but we and others hypothesize that a far greater number of metamorphic proteins remain undiscovered. In this work, we introduce a computational model to predict metamorphic behavior in proteins using only knowledge of the sequence. In this model, secondary structure prediction programs are used to calculate diversity indices, which are measures of uncertainty in predicted secondary structure at each position in the sequence; these are then used to assign protein sequences as likely to be metamorphic vs. monomorphic (i.e. having just one-fold). We constructed a reference dataset to train our classification method, which includes a novel compilation of 140 likely-monomorphic proteins, and a set of 201 metamorphic proteins taken from the literature. Our model is able to classify proteins as metamorphic vs. monomorphic with a Matthews correlation coefficient of [~]0.4 and true positive / true negative rates of [~]65% / 80%, suggesting that it is possible to predict metamorphic behavior in proteins using only sequence information. Statement of SignificanceThis paper introduces the diversity index as a descriptor to distinguish metamorphic proteins, which possess multiple stable folds, from monomorphic proteins that possess only one fold. The diversity index is designed to measure uncertainty in computationally predicted secondary structure, which we hypothesize is elevated for metamorphic proteins. We tested our hypothesis by training a binary classifier using the diversity index and an annotated dataset of metamorphic and monomorphic proteins, and found an optimal Matthews correlation coefficient of 0.4, supporting the hypothesis and demonstrating for the first time that it is possible to predict metamorphic behavior in proteins using only sequence information. The sequence-based classifier has broader applicability compared to methods that rely on making comparison to experimentally measured structures.