Co-evolutionary Information Research Articles

Evolutionary-guided de novo structure prediction of self-associated transmembrane helical proteins with near-atomic accuracy

How specific protein associations regulate the function of membrane receptors remains poorly understood. Conformational flexibility currently hinders the structure determination of several classes of membrane receptors and associated oligomers. Here we develop EFDOCK-TM, a general method to predict self-associated transmembrane protein helical (TMH) structures from sequence guided by co-evolutionary information. We show that accurate intermolecular contacts can be identified using a combination of protein sequence covariation and TMH binding surfaces predicted from sequence. When applied to diverse TMH oligomers, including receptors characterized in multiple conformational and functional states, the method reaches unprecedented near-atomic accuracy for most targets. Blind predictions of structurally uncharacterized receptor tyrosine kinase TMH oligomers provide a plausible hypothesis on the molecular mechanisms of disease-associated point mutations and binding surfaces for the rational design of selective inhibitors. The method sets the stage for uncovering novel determinants of molecular recognition and signalling in single-spanning eukaryotic membrane receptors.

Protein and RNA Structure Prediction by Integration of Co-Evolutionary Information into Molecular Simulation

Exploring the relationship of structure and function is crucial for the understanding of molecular life. Despite significant progress of experimental methods, the structural characterization of important structures for both proteins and RNA - typically preceding any detailed mechanistic exploration of their function - remains challenging. In recent years, increasingly ubiquitous availability of sequential information and novel statistical analysis has allowed to trace the co-evolution of residues and predict contact maps. These contact maps can be exploited in structure prediction tools. One maximum entropy based approach is called Direct Coupling Analysis (DCA)[1] and, e.g., was found to be sufficient for the blind prediction of a protein complex[2] and its active conformation[3] later confirmed by experiment. We have successfully adapted this method to the specifics of RNA (rnaDCA): While it is comparably simple to predict RNA secondary structure through analyzing possible Watson-Crick base pairings, predicting tertiary contacts has remained - despite significant effort - an elusive task met with limited success. In contrast, our novel rnaDCA is able to extract tertiary contacts from genomic data. We further demonstrate that these tertiary contacts are sufficient to systematically improve tertiary RNA prediction quality[4]. Considering the large gap of known ncRNA sequences to experimentally resolved tertiary structures, we are convinced that this will significantly impact all structural RNA related research.

Using evolution to predict structure.

Researchers use sequence coevolution information to predict the structures of protein complexes.

Sequence-based antigenic change prediction by a sparse learning method incorporating co-evolutionary information.

Rapid identification of influenza antigenic variants will be critical in selecting optimal vaccine candidates and thus a key to developing an effective vaccination program. Recent studies suggest that multiple simultaneous mutations at antigenic sites accumulatively enhance antigenic drift of influenza A viruses. However, pre-existing methods on antigenic variant identification are based on analyses from individual sites. Because the impacts of these co-evolved sites on influenza antigenicity may not be additive, it will be critical to quantify the impact of not only those single mutations but also multiple simultaneous mutations or co-evolved sites. Here, we developed and applied a computational method, AntigenCO, to identify and quantify both single and co-evolutionary sites driving the historical antigenic drifts. AntigenCO achieved an accuracy of up to 90.05% for antigenic variant prediction, significantly outperforming methods based on single sites. AntigenCO can be useful in antigenic variant identification in influenza surveillance.

Coevolutionary information, protein folding landscapes, and the thermodynamics of natural selection.

The energy landscape used by nature over evolutionary timescales to select protein sequences is essentially the same as the one that folds these sequences into functioning proteins, sometimes in microseconds. We show that genomic data, physical coarse-grained free energy functions, and family-specific information theoretic models can be combined to give consistent estimates of energy landscape characteristics of natural proteins. One such characteristic is the effective temperature T(sel) at which these foldable sequences have been selected in sequence space by evolution. T(sel) quantifies the importance of folded-state energetics and structural specificity for molecular evolution. Across all protein families studied, our estimates for T(sel) are well below the experimental folding temperatures, indicating that the energy landscapes of natural foldable proteins are strongly funneled toward the native state.

Toward rationally redesigning bacterial two-component signaling systems using coevolutionary information

A challenge in molecular biology is to distinguish the key subset of residues that allow two-component signaling (TCS) proteins to recognize their correct signaling partner such that they can transiently bind and transfer signal, i.e., phosphoryl group. Detailed knowledge of this information would allow one to search sequence space for mutations that can be used to systematically tune the signal transmission between TCS partners as well as potentially encode a TCS protein to preferentially transfer signals to a nonpartner. Motivated by the notion that this detailed information is found in sequence data, we explore the sequence coevolution between signaling partners to better understand how mutations can positively or negatively alter their ability to transfer signal. Using direct coupling analysis for determining evolutionarily conserved protein-protein interactions, we apply a metric called the direct information score to quantify mutational changes in the interaction between TCS proteins and demonstrate that it accurately correlates with experimental mutagenesis studies probing the mutational change in measured in vitro phosphotransfer. Furthermore, by subtracting from our metric an appropriate null model corresponding to generic, conserved features in TCS signaling pairs, we can isolate the determinants that give rise to interaction specificity and recognition, which are variable among different TCS partners. Our methodology forms a potential framework for the rational design of TCS systems by allowing one to quickly search sequence space for mutations or even entirely new sequences that can increase or decrease our metric, as a proxy for increasing or decreasing phosphotransfer ability between TCS proteins.

From structure to function: the convergence of structure based models and co-evolutionary information

Understanding protein folding and function is one of the most important problems in biological research. Energy landscape theory and the folding funnel concept have provided a framework to investigate the mechanisms associated to these processes. Since protein energy landscapes are in most cases minimally frustrated, structure based models (SMBs) have successfully determined the geometrical features associated with folding and functional transitions. However, structural information is limited, particularly with respect to different functional configurations. This is a major limitation for SBMs. Alternatively, statistical methods to study amino acid co-evolution provide information on residue-residue interactions useful for the study of structure and function. Here, we show how the combination of these two methods gives rise to a novel way to investigate the mechanisms associated with folding and function. We use this methodology to explore the mechanistic aspects of protein translocation in the integral membrane protease FtsH. Dual basin-SBM simulations using the open and closed state of this hexameric motor reveals a functionally important paddling motion in the catalytic cycle. We also find that Direct Coupling Analysis (DCA) predicts physical contacts between AAA and peptidase domains of the motor, which are crucial for the open to close transition. Our combined method, which uses structural information from the open state experimental structure and co-evolutionary couplings, suggests that this methodology can be used to explore the functional landscape of complex biological macromolecules previously inaccessible to methods dependent on experimental structural information. This efficient way to sample the conformational space of large systems creates a theoretical/computational framework capable of better characterizing the functional landscape in large biomolecular assemblies.

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

Recently developed methods have shown considerable promise in predicting residue-residue contacts in protein 3D structures using evolutionary covariance information. However, these methods require large numbers of evolutionarily related sequences to robustly assess the extent of residue covariation, and the larger the protein family, the more likely that contact information is unnecessary because a reasonable model can be built based on the structure of a homolog. Here we describe a method that integrates sequence coevolution and structural context information using a pseudolikelihood approach, allowing more accurate contact predictions from fewer homologous sequences. We rigorously assess the utility of predicted contacts for protein structure prediction using large and representative sequence and structure databases from recent structure prediction experiments. We find that contact predictions are likely to be accurate when the number of aligned sequences (with sequence redundancy reduced to 90%) is greater than five times the length of the protein, and that accurate predictions are likely to be useful for structure modeling if the aligned sequences are more similar to the protein of interest than to the closest homolog of known structure. These conditions are currently met by 422 of the protein families collected in the Pfam database.

Deciphering the Protein Folding Code through Zipping Steps

Earlier experiments suggest that the evolutionary information (conservation of amino acids and coevolution between amino acids) encoded in protein sequences is necessary and sufficient to specify the fold of a protein family. However, there is no computational work to quantify the effect of such evolutionary information on the folding process. Here we simulate a repertoire of native and artificial WW domain sequences using a physics-based protein structure search method called ZAM (Zipping and Assembly method), which samples conformational space effectively towards native-like conformations through zipping and assembly search mechanism. We explore the sequence-structure relationship for WW domains and find that the coevolution information has a remarkable influence on local contacts of N-terminal β-turn of WW domains. This turn would not form correctly in the absence of such information. Moreover, through maximum likelihood approach, we identify five local contacts that play a critical role in folding. using the contact probability of those five local contacts at the early stage of folding, we built a classification model. This enables us to predict the foldability of a WW sequence with 81% accuracy. Based on this classification model, we re-design the unfoldable WW domain sequences and make them foldable by introducing a few mutations that lead to stabilization of these critical contacts.

Structure-based rebuilding of coevolutionary information reveals functional modules in rhodopsin structure

Correlated mutation analysis (CMA) has been used to investigate protein functional sites. However, CMA has suffered from low signal-to-noise ratio caused by meaningless phylogenetic signals or structural constraints. We present a new method, Structure-based Correlated Mutation Analysis (SCMA), which encodes coevolution scores into a protein structure network. A path-based network model is adapted to describe information transfer between residues, and the statistical significance is estimated by network shuffling. This model intrinsically assumes that residues in physical contact have a more reliable coevolution score than distant residues, and that coevolution in distant residues likely arises from a series of contacting and coevolving residues. In addition, coevolutionary coupling is statistically controlled to remove the structural effects. When applied to the rhodopsin structure, the SCMA method identified a much higher percentage of functional residues than the typical coevolution score (61% vs. 22%). In addition, statistically significant residues are used to construct the coevolved residue–residue subnetwork. The network has one highly connected node (retinal bound Lys296), indicating that Lys296 can induce and regulate most other coevolved residues in a variety of locations. The coevolved network consists of a few modular clusters which have distinct functional roles. This article is part of a Special Issue entitled: Computational Methods for Protein Interaction and Structural Prediction.

Efficient algorithms for reconstructing gene content by co-evolution.

BackgroundIn a previous study we demonstrated that co-evolutionary information can be utilized for improving the accuracy of ancestral gene content reconstruction. To this end, we defined a new computational problem, the Ancestral Co-Evolutionary (ACE) problem, and developed algorithms for solving it.ResultsIn the current paper we generalize our previous study in various ways. First, we describe new efficient computational approaches for solving the ACE problem. The new approaches are based on reductions to classical methods such as linear programming relaxation, quadratic programming, and min-cut. Second, we report new computational hardness results related to the ACE, including practical cases where it can be solved in polynomial time.Third, we generalize the ACE problem and demonstrate how our approach can be used for inferring parts of the genomes of non-ancestral organisms. To this end, we describe a heuristic for finding the portion of the genome ('dominant set’) that can be used to reconstruct the rest of the genome with the lowest error rate. This heuristic utilizes both evolutionary information and co-evolutionary information.We implemented these algorithms on a large input of the ACE problem (95 unicellular organisms, 4,873 protein families, and 10, 576 of co-evolutionary relations), demonstrating that some of these algorithms can outperform the algorithm used in our previous study. In addition, we show that based on our approach a ’dominant set’ cab be used reconstruct a major fraction of a genome (up to 79%) with relatively low error-rate (e.g. 0.11). We find that the ’dominant set’ tends to include metabolic and regulatory genes, with high evolutionary rate, and low protein abundance and number of protein-protein interactions.ConclusionsThe ACE problem can be efficiently extended for inferring the genomes of organisms that exist today. In addition, it may be solved in polynomial time in many practical cases. Metabolic and regulatory genes were found to be the most important groups of genes necessary for reconstructing gene content of an organism based on other related genomes.

New methods to measure residues coevolution in proteins

BackgroundThe covariation of two sites in a protein is often used as the degree of their coevolution. To quantify the covariation many methods have been developed and most of them are based on residues position-specific frequencies by using the mutual information (MI) model.ResultsIn the paper, we proposed several new measures to incorporate new biological constraints in quantifying the covariation. The first measure is the mutual information with the amino acid background distribution (MIB), which incorporates the amino acid background distribution into the marginal distribution of the MI model. The modification is made to remove the effect of amino acid evolutionary pressure in measuring covariation. The second measure is the mutual information of residues physicochemical properties (MIP), which is used to measure the covariation of physicochemical properties of two sites. The third measure called MIBP is proposed by applying residues physicochemical properties into the MIB model. Moreover, scores of our new measures are applied to a robust indicator conn(k) in finding the covariation signal of each site.ConclusionsWe find that incorporating amino acid background distribution is effective in removing the effect of evolutionary pressure of amino acids. Thus the MIB measure describes more biological background information for the coevolution of residues. Besides, our analysis also reveals that the covariation of physicochemical properties is a new aspect of coevolution information.

Commentary on a New Era of Investigating 3D Structure-Based Human?Virus Protein Network Dynamics

Virology is a fundamental research field for viruses that encompasses studies on minimum replication units, rapid evolution, and clinical applications. Similar to the other biological systems, structural study of viral system is based on three distinct phases of understanding: (I) modeling of steady-state network; (II) involvement of spatial confinements to the model; (III) involvement of timing asymmetry to the model. Most molecular biological understandings are limited to phase (I), rendering later phases to be investigated by future systems biology. Even in phase (I), complexities of protein–protein network system make the researchers puzzled over examining actual underlying principles in the observed data sets. One type of solution for such an abstraction of the principles is to model a protein–protein interaction network based on graph theory; a theory of pair-wise relations between objects of a certain collection. Graph is constituted of nodes (for protein of interests in this case) and edges (for a certain protein–protein interaction), sometimes accompanied by different evaluation weights for an expected dynamics. Scale-free network, in which a small number of nodes act as hubs and the other nodes possess extremely small numbers of links, is supposed to be a characteristic of actual protein–protein interaction network in biological system. This type of network is prominent when the network encounters adding-on system for growth and selection bias for linking nodes act as hubs. However, it is not clear whether an input and output of such an abstracted system represents actual dynamics of original biological system, and what biological significances of parameters characterizing the networks are, i.e., density, transitivity, reciprocity, centrality etc. Progressing of the type of study requires further information to analyze the network elegantly. In the article that deals with underlying principles of human–virus protein interaction network (Proc. Natl. Acad. Sci. U.S.A. 108, 10538–10543), Drs Franzosa and Xia have reported the results from 3D surface structure-based protein interaction network system. Based on 3039 endogenous interactions among 2435 human proteins as well as 53 exogenous interactions among 50 viral proteins from 36 viral species and their 50 human target proteins, they developed combined method of BLAST homology evaluation algorithm and MSMS program based on solvent-accessible surface area (SASA). The unique properties of this model case study are (i) detection of convergent evolution (mimicking evolution of viral protein surfaces to non-homological host human protein surfaces, in contrast to host gene duplication resulted in overlapping protein surface interactions) and evolutionary “arms race” (fast evolving trait of exogenous interactions due to cutting-off and catching-up interactions between human target proteins and viral proteins) due to rapid evolution and highly antagonistic manner of viral proteins and their target proteins, in contrast to cooperative interaction of endogenous human protein interactions; (ii) higher resolution of data based on more information from 3D surface structure interface. The original idea of using 3D surface information to network analysis is from the yeast study by Kim et al. (2006). In contrast to Kim's work, the patterns of evolution became much clearer based on fast evolving antagonistic viral proteins of absolutely different origins from host human. Compared to endogenous interactions, exogenous interactions exhibit transient structure, increase in number of interacting proteins, higher tendency of regulatory function, and faster evolution. Addition of BLAST information elucidates co-evolutionary information such as convergent evolution and evolutionary arms race. Although the researchers are still struggling in phase (I), the application of 3D structural information to a graph of protein–protein network is very useful for investigating the evolutionary dynamics of the network of interests applying huge sum of mechanical interactive information to a simple graph of the interaction network. Additional investigation by commonly used BLAST algorithm will be useful expecting clinical application of drug design to highly conserved protein–protein interaction, instead of fast evolving highly adaptive region for target viruses.

Reconstructing ancestral gene content by coevolution

Inferring the gene content of ancestral genomes is a fundamental challenge in molecular evolution. Due to the statistical nature of this problem, ancestral genomes inferred by the maximum likelihood (ML) or the maximum-parsimony (MP) methods are prone to considerable error rates. In general, these errors are difficult to abolish by using longer genomic sequences or by analyzing more taxa. This study describes a new approach for improving ancestral genome reconstruction, the ancestral coevolver (ACE), which utilizes coevolutionary information to improve the accuracy of such reconstructions over previous approaches. The principal idea is to reduce the potentially large solution space by choosing a single optimal (or near optimal) solution that is in accord with the coevolutionary relationships between protein families. Simulation experiments, both on artificial and real biological data, show that ACE yields a marked decrease in error rate compared with ML or MP. Applied to a large data set (95 organisms, 4873 protein families, and 10,000 coevolutionary relationships), some of the ancestral genomes reconstructed by ACE were remarkably different in their gene content from those reconstructed by ML or MP alone (more than 10% in some nodes). These reconstructions, while having almost similar likelihood/parsimony scores as those obtained with ML/MP, had markedly higher concordance with the coevolutionary information. Specifically, when ACE was implemented to improve the results of ML, it added a large number of proteins to those encoded by LUCA (last universal common ancestor), most of them ribosomal proteins and components of the F(0)F(1)-type ATP synthase/ATPases, complexes that are vital in most living organisms. Our analysis suggests that LUCA appears to have been bacterial-like and had a genome size similar to the genome sizes of many extant organisms.

Integration of Evolutionary Features for the Identification of Functionally Important Residues in Major Facilitator Superfamily Transporters

The identification of functionally important residues is an important challenge for understanding the molecular mechanisms of proteins. Membrane protein transporters operate two-state allosteric conformational changes using functionally important cooperative residues that mediate long-range communication from the substrate binding site to the translocation pathway. In this study, we identified functionally important cooperative residues of membrane protein transporters by integrating sequence conservation and co-evolutionary information. A newly derived evolutionary feature, the co-evolutionary coupling number, was introduced to measure the connectivity of co-evolving residue pairs and was integrated with the sequence conservation score. We tested this method on three Major Facilitator Superfamily (MFS) transporters, LacY, GlpT, and EmrD. MFS transporters are an important family of membrane protein transporters, which utilize diverse substrates, catalyze different modes of transport using unique combinations of functional residues, and have enough characterized functional residues to validate the performance of our method. We found that the conserved cores of evolutionarily coupled residues are involved in specific substrate recognition and translocation of MFS transporters. Furthermore, a subset of the residues forms an interaction network connecting functional sites in the protein structure. We also confirmed that our method is effective on other membrane protein transporters. Our results provide insight into the location of functional residues important for the molecular mechanisms of membrane protein transporters.

1SA1-03 共進化情報によるタンパク質間相互作用予測(1SA1 タンパク質間相互作用研究の手法としての共進化解析,第47回日本生物物理学会年会)

1SA1-03 共進化情報によるタンパク質間相互作用予測(1SA1 タンパク質間相互作用研究の手法としての共進化解析,第47回日本生物物理学会年会)

Improving Protein–Protein Interaction Prediction Based on Phylogenetic Information Using a Least‐Squares Support Vector Machine

Predicting protein-protein interactions has become a key step of reverse-engineering biological networks to better understand cellular functions. The experimental methods in determining protein-protein interactions are time-consuming and costly, which has motivated vigorous development of computational approaches for predicting protein-protein interactions. A set of recently developed bioinformatics methods utilizes coevolutionary information of the interacting partners (e.g., as exhibited in the form of correlations between distance matrices, where, for each protein, a matrix stores the pairwise distances between the protein and its orthologs in a group of reference genomes). We proposed a novel method to account for the intra-matrix correlations in improving predictive accuracy. The distance matrices for a pair of proteins are transformed and concatenated into a phylogenetic vector. A least-squares support vector machine is trained and tested on pairs of proteins, represented as phylogenetic vectors, whose interactions are known. The intra-matrix correlations are accounted for by introducing a weighted linear kernel, which determines the dot product of two phylogenetic vectors. The performance, measured as receiver operator characteristic (ROC) score in cross-validation experiments, shows significant improvement of our method (ROC score 0.928) over that obtained by Pearson correlations (0.659).

Phylogenetic tree information aids supervised learning for predicting protein-protein interaction based on distance matrices.

BackgroundProtein-protein interactions are critical for cellular functions. Recently developed computational approaches for predicting protein-protein interactions utilize co-evolutionary information of the interacting partners, e.g., correlations between distance matrices, where each matrix stores the pairwise distances between a protein and its orthologs from a group of reference genomes.ResultsWe proposed a novel, simple method to account for some of the intra-matrix correlations in improving the prediction accuracy. Specifically, the phylogenetic species tree of the reference genomes is used as a guide tree for hierarchical clustering of the orthologous proteins. The distances between these clusters, derived from the original pairwise distance matrix using the Neighbor Joining algorithm, form intermediate distance matrices, which are then transformed and concatenated into a super phylogenetic vector. A support vector machine is trained and tested on pairs of proteins, represented as super phylogenetic vectors, whose interactions are known. The performance, measured as ROC score in cross validation experiments, shows significant improvement of our method (ROC score 0.8446) over that of using Pearson correlations (0.6587).ConclusionWe have shown that the phylogenetic tree can be used as a guide to extract intra-matrix correlations in the distance matrices of orthologous proteins, where these correlations are represented as intermediate distance matrices of the ancestral orthologous proteins. Both the unsupervised and supervised learning paradigms benefit from the explicit inclusion of these intermediate distance matrices, and particularly so in the latter case, which offers a better balance between sensitivity and specificity in the prediction of protein-protein interactions.

Prediction of Protein-Protein Interactions by Using Co-Evolutionary Information

A distance matrix is a set of genetic distances between all possible pairs of proteins under consideration, and is used to construct a phylogenetic tree by the distance matrix method. Pazos and Valencia (2001) have developed a method to predict protein-protein interaction by evaluating the similarity of the distance matrices, under the assumption that the phylogenetic trees of interacting proteins resemble each other through co-evolution. It is known, however, that the prediction includes many false positives. We postulated that the cause of the false positives is the background similarity of the phylogenetic relationship of the source organisms. We have developed a method to exclude such information from the distance matrices with a projection operator. The number of false positives was drastically reduced from the prediction by evaluating the similarity between the residuals after the projection operation.

S3c2-3 Prediction of protein-protein interaction with co-evolutionary information(S3-c2: "Structural Bioinformatics: Molecular structures as the basis of understanding protein network systems",Symposia,Abstract,Meeting Program of EABS & BSJ 2006)

S3c2-3 Prediction of protein-protein interaction with co-evolutionary information(S3-c2: "Structural Bioinformatics: Molecular structures as the basis of understanding protein network systems",Symposia,Abstract,Meeting Program of EABS & BSJ 2006)