Accurate de novo structure prediction of large transmembrane protein domains using fragment-assembly and correlated mutation analysis

Abstract

A new de novo protein structure prediction method for transmembrane proteins (FILM3) is described that is able to accurately predict the structures of large membrane proteins domains using an ensemble of two secondary structure prediction methods to guide fragment selection in combination with a scoring function based solely on correlated mutations detected in multiple sequence alignments. This approach has been validated by generating models for 28 membrane proteins with a diverse range of complex topologies and an average length of over 300 residues with results showing that TM-scores > 0.5 can be achieved in almost every case following refinement using MODELLER. In one of the most impressive results, a model of mitochondrial cytochrome c oxidase polypeptide I was obtained with a TM-score > 0.75 and an rmsd of only 5.7 Å over all 514 residues. These results suggest that FILM3 could be applicable to a wide range of transmembrane proteins of as-yetunknown 3D structure given sufficient homologous sequences.structural bioinformatics | protein modeling | compressed sensing | amino acid contacts A lpha-helical transmembrane proteins (TMPs) constitute roughly 30% of a typical genome and play critical roles in a diverse range of biological processes whereas many are also important drug targets. Despite the recent increase in the number of solved TMP crystal structures, coverage of TMP fold space remains sparse, particularly at high resolutions, with close to 300 unique structures deposited as of 2011 (1). Computational methods to predict TMP structure are therefore vital in helping to further our knowledge of the structure and function of these proteins.To date, TMP structure prediction has been dominated by topology prediction. Machine learning-based predictors, trained and validated using topology data derived from structural data combined with evolutionary information, now achieve prediction accuracies in the range 80-90% (2, 3). Another approach, based on an experimental scale of position-specific amino acid contributions to membrane insertion free energy, achieves similar accuracy suggesting that predicting TMP topology from first principles is an achievable goal (4).As with globular proteins, predicting the structure of TMPs by homology modeling is very effective particularly when TMP-specific methods are used (5); however, the paucity of solved structures means that homology modeling can only be applied to a minority of TMP families. With this in mind, a small number of de novo modeling approaches, which attempt to build 3D models for TMPs without the use of homology to known structures, have also been developed.FILM (6), a modification of the globular protein structure prediction method FRAGFOLD (7, 8), attempts to assemble folds from supersecondary structural fragments taken from a library of highly resolved protein structures using simulated annealing. FILM differs from FRAGFOLD in the addition of a membrane environment potential, derived from the statistical analysis of 640 transmembrane helices, by measuri...

Assessing the utility of coevolution-based residue–residue contact predictions in a sequence- and structure-rich era

“…Recent studies (4,11,12) have shown that for targets with deep alignments, the predicted contacts are sufficiently accurate to predict the 3D structures of proteins. However, because of the steady increase in the structures deposited in the Protein Data Bank (PDB; ref.…”

“…S8)]. Predicted contacts are likely most useful for membrane proteins (12,15), protein assemblies (16), and other systems where high-resolution structural information is currently sparse.…”

Computation and Functional Studies Provide a Model for the Structure of the Zinc Transporter hZIP4

“…Residue-Residue Contacts Predicted from Co-evolutionMembrane structures have been modeled using contact prediction based on co-evolutionary patterns in large multiple sequence alignments of homologous proteins (32,33). To see whether such an analysis was possible for hZIP4, we constructed an alignment over the conserved transmembrane region by querying the UniProt database with HHblits (19,34).…”

Charting the Signal Trajectory in a Light-Oxygen-Voltage Photoreceptor by Random Mutagenesis and Covariance Analysis

“…Second, it can be challenging to untangle direct interactions between residue pairs from indirect interactions mediated by a third intermediary residue that directly interacts with the first two residues. The latter aspect is efficiently addressed by advanced covariance analysis methods (30) such as the direct-coupling analysis (31,32) and the protein-sparse-inverse-covariance (PSICOV) (24,33) strategy.…”