Learning to Identify Physiological and Adventitious Metal-Binding Sites in the Three-Dimensional Structures of Proteins by Following the Hints of a Deep Neural Network

Abstract

Thirty-eight percent of protein structures in the Protein Data Bank contain at least one metal ion. However, not all these metal sites are biologically relevant. Cations present as impurities during sample preparation or in the crystallization buffer can cause the formation of protein–metal complexes that do not exist in vivo. We implemented a deep learning approach to build a classifier able to distinguish between physiological and adventitious zinc-binding sites in the 3D structures of metalloproteins. We trained the classifier using manually annotated sites extracted from the MetalPDB database. Using a 10-fold cross validation procedure, the classifier achieved an accuracy of about 90%. The same neural classifier could predict the physiological relevance of non-heme mononuclear iron sites with an accuracy of nearly 80%, suggesting that the rules learned on zinc sites have general relevance. By quantifying the relative importance of the features describing the input zinc sites from the network perspective and by analyzing the characteristics of the MetalPDB datasets, we inferred some common principles. Physiological sites present a low solvent accessibility of the aminoacids forming coordination bonds with the metal ion (the metal ligands), a relatively large number of residues in the metal environment (≥20), and a distinct pattern of conservation of Cys and His residues in the site. Adventitious sites, on the other hand, tend to have a low number of donor atoms from the polypeptide chain (often one or two). These observations support the evaluation of the physiological relevance of novel metal-binding sites in protein structures.

Structural Bioinformatics and Deep Learning of Metalloproteins: Recent Advances and Applications

“…Discrimination of physiological and adventitious zinc-binding sites in MPs using a recurrent neural network (RNN) [98] Table 1 summarizes the resources and applications mentioned in this contribution, in the order in which they were described.…”

“…In a very recent application, our own research team developed a DL classifier that can discriminate physiological and adventitious zinc-binding sites in the 3D structures of MPs with an accuracy of about 90% [ 98 ]. In order to develop the tool, we trained a recurrent neural network (RNN) using a dataset of 1944 physiological and 3352 adventitious zinc-binding sites extracted from MetalPDB and manually annotated.…”

Prediction of Ca2+binding site in proteins with a fast and accurate method based on statistical mechanics and analysis of crystal structures

“…Validation software such as PROCHECK 74 , MolProbity 75 , and Coot 76 and servers like CheckMyBlob 77 and CheckMyMetal 78 facilitate the automated assessment of structural parameters. Moreover, the biological context and functional relevance are crucial when analyzing protein structures 33,79-80 .…”

Structural Bioinformatics and Deep Learning of Metalloproteins: Recent Advances and Applications

“…In a very recent application, our own research team developed a DL classifier that can discriminate physiological and adventitious zinc-binding sites in the 3D structures of MPs with an accuracy of about 90% [98]. In order to develop the tool, we trained a recurrent neural network (RNN) using a dataset of 1944 physiological and 3352 adventitious zincbinding sites extracted from MetalPDB and manually annotated.…”