Using coevolution to predict protein-protein interactions

Prediction of interactions between cell surface proteins by machine learning

Cells detect changes in their external environments or communicate with each other through proteins on their surfaces. These cell surface proteins form a complicated network of interactions in order to fulfill their functions. The interactions between cell surface proteins are highly dynamic and, thus, challenging to detect using traditional experimental techniques. Here, we tackle this challenge using a computational framework. The primary focus of the framework is to develop new tools to identify interactions between domains in the immunoglobulin (Ig) fold, which is the most abundant domain family in cell surface proteins. These interactions could be formed between ligands and receptors from different cells or between proteins on the same cell surface. In practice, we collected all structural data on Ig domain interactions and transformed them into an interface fragment pair library. A high-dimensional profile can then be constructed from the library for a given pair of query protein sequences. Multiple machine learning models were used to read this profile so that the probability of interaction between the query proteins could be predicted. We tested our models on an experimentally derived dataset that contains 564 cell surface proteins in humans. The cross-validation results show that we can achieve higher than 70% accuracy in identifying the PPIs within this dataset. We then applied this method to a group of 46 cell surface proteins in Caenorhabditis elegans. We screened every possible interaction between these proteins. Many interactions recognized by our machine learning classifiers have been experimentally confirmed in the literature. In conclusion, our computational platform serves as a useful tool to help identify potential new interactions between cell surface proteins in addition to current state-of-the-art experimental techniques. The tool is freely accessible for use by the scientific community. Moreover, the general framework of the machine learning classification can also be extended to study the interactions of proteins in other domain superfamilies.

Prediction of Interactions between Cell Surface Proteins by Machine Learning

Cells detect changes of external environments or communicate with each other through proteins on their surfaces. These cell surface proteins form a complicated network of interactions in order to fulfill their functions. The interactions between cell surface proteins are highly dynamic and thus challenging to detect using traditional experimental techniques. Here we tackle this challenge by a computational framework. The primary focus of the framework is to develop new tools to identify interactions between domains in immunoglobulin (Ig) fold, which is the most abundant domain family in cell surface proteins. These interactions could be formed between ligands and receptors from different cells, or between proteins on the same cell surface. In practice, we collected all structural data of Ig domain interactions and transformed them into an interface fragment pair library. A high dimensional profile can be then constructed from the library for a given pair of query protein sequences. Multiple machine learning models were used to read this profile, so that the probability of interaction between the query proteins can be predicted. We tested our models to an experimentally derived dataset which contains 564 cell surface proteins in human. The cross-validation results show that we can achieve higher than 70% accuracy in identifying the PPIs within this dataset. We then applied this method to a group of 46 cell surface proteins in C elegans. We screened every possible interaction between these proteins. Many interactions recognized by our machine learning classifiers have been experimentally confirmed in the literatures. In conclusion, our computational platform serves a useful tool to help identifying potential new interactions between cell surface proteins in addition to current state-of-the-art experimental techniques. The tool is freely accessible for use by the scientific community. Moreover, the general framework of the machine learning classification can also be extended to study interactions of proteins in other domain superfamilies.

Intra-Domain Residue Coevolution in Transcription Factors Contributes to DNA Binding Specificity

Understanding the basis of the DNA-binding specificity of transcription factors (TFs) has been of long-standing interest. Despite extensive efforts to map millions of putative TF binding sequences, identifying the critical determinants for DNA binding specificity remains a major challenge. The coevolution of residues in proteins occurs due to a shared evolutionary history. However, it is unclear how coevolving residues in TFs contribute to DNA binding specificity. Here, we systematically collected publicly available data sets from multiple large-scale high-throughput TF-DNA interaction screening experiments for the major TF families with large numbers of TF members. These families included the Homeobox, HLH, bZIP_1, Ets, HMG_box, ZF-C4, and Zn_clus TFs. We detected TF subclass-determining sites (TSDSs) and showed that the TSDSs were more likely to coevolve with other TSDSs than with non-TSDSs, particularly for the Homeobox, HLH, Ets, bZIP_1, and HMG_box TF families. By in silico modeling, we showed that mutation of the highly coevolving residues could significantly reduce the stability of the TF-DNA complex. The distant residues from the DNA interface also contributed to TF-DNA binding activity. Overall, our study gave evidence that coevolved residues relate to transcriptional regulation and provided insights into the potential application of engineered DNA-binding domains and proteins. IMPORTANCE While unraveling DNA-binding specificity of TFs is the key to understanding the basis and molecular mechanism of gene expression regulation, identifying the critical determinants that contribute to DNA binding specificity remains a major challenge. In this study, we provided evidence showing that coevolving residues in TF domains contributed to DNA binding specificity. We demonstrated that the TSDSs were more likely to coevolve with other TSDSs than with non-TSDSs. Mutation of the coevolving residue pairs (CRPs) could significantly reduce the stability of THE TF-DNA complex, and even the distant residues from the DNA interface contribute to TF-DNA binding activity. Collectively, our study expands our knowledge of the interactions among coevolved residues in TFs, tertiary contacting, and functional importance in refined transcriptional regulation. Understanding the impact of coevolving residues in TFs will help understand the details of transcription of gene regulation and advance the application of engineered DNA-binding domains and protein.

Modeling Protein Complexes and Molecular Assemblies Using Computational Methods

Many biological molecules are assembled into supramolecular complexes that are necessary to perform functions in the cell. Better understanding and characterization of these molecular assemblies are thus essential to further elucidate molecular mechanisms and key protein-protein interactions that could be targeted to modulate the protein binding affinity or develop new binders. Experimental access to structural information on these supramolecular assemblies is often hampered by the size of these systems that make their recombinant production and characterization rather difficult. Computational methods combining both structural data, molecular modeling techniques, and sequence coevolution information can thus offer a good alternative to gain access to the structural organization of protein complexes and assemblies. Herein, we present some computational methods to predict structural models of the protein partners, to search for interacting regions using coevolution information, and to build molecular assemblies. The approach is exemplified using a case study to model the succinate-quinone oxidoreductase heterocomplex.

Detection of significant protein coevolution

MOTIVATION

The evolution of proteins cannot be fully understood without taking into account the coevolutionary linkages entangling them. From a practical point of view, coevolution between protein families has been used as a way of detecting protein interactions and functional relationships from genomic information. The most common approach to inferring protein coevolution involves the quantification of phylogenetic tree similarity using a family of methodologies termed mirrortree. In spite of their success, a fundamental problem of these approaches is the lack of an adequate statistical framework to assess the significance of a given coevolutionary score (tree similarity). As a consequence, a number of ad hoc filters and arbitrary thresholds are required in an attempt to obtain a final set of confident coevolutionary signals.

RESULTS

In this work, we developed a method for associating confidence estimators (P values) to the tree-similarity scores, using a null model specifically designed for the tree comparison problem. We show how this approach largely improves the quality and coverage (number of pairs that can be evaluated) of the detected coevolution in all the stages of the mirrortree workflow, independently of the starting genomic information. This not only leads to a better understanding of protein coevolution and its biological implications, but also to obtain a highly reliable and comprehensive network of predicted interactions, as well as information on the substructure of macromolecular complexes using only genomic information.

AVAILABILITY AND IMPLEMENTATION

The software and datasets used in this work are freely available at: http://csbg.cnb.csic.es/pMT/.

CONTACT

pazos@cnb.csic.es

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Studying tumorigenesis through network evolution and somatic mutational perturbations in the cancer interactome

Cells govern biological functions through complex biological networks. Perturbations to networks may drive cells to new phenotypic states, for example, tumorigenesis. Identifying how genetic lesions perturb molecular networks is a fundamental challenge. This study used large-scale human interactome data to systematically explore the relationship among network topology, somatic mutation, evolutionary rate, and evolutionary origin of cancer genes. We found the unique network centrality of cancer proteins, which is largely independent of gene essentiality. Cancer genes likely have experienced a lower evolutionary rate and stronger purifying selection than those of noncancer, Mendelian disease, and orphan disease genes. Cancer proteins tend to have ancient histories, likely originated in early metazoan, although they are younger than proteins encoded by Mendelian disease genes, orphan disease genes, and essential genes. We found that the protein evolutionary origin (age) positively correlates with protein connectivity in the human interactome. Furthermore, we investigated the network-attacking perturbations due to somatic mutations identified from 3,268 tumors across 12 cancer types in The Cancer Genome Atlas. We observed a positive correlation between protein connectivity and the number of nonsynonymous somatic mutations, whereas a weaker or insignificant correlation between protein connectivity and the number of synonymous somatic mutations. These observations suggest that somatic mutational network-attacking perturbations to hub genes play an important role in tumor emergence and evolution. Collectively, this work has broad biomedical implications for both basic cancer biology and the development of personalized cancer therapy.

Practical aspects of protein co-evolution

Co-evolution is a fundamental aspect of Evolutionary Theory. At the molecular level, co-evolutionary linkages between protein families have been used as indicators of protein interactions and functional relationships from long ago. Due to the complexity of the problem and the amount of genomic data required for these approaches to achieve good performances, it took a relatively long time from the appearance of the first ideas and concepts to the quotidian application of these approaches and their incorporation to the standard toolboxes of bioinformaticians and molecular biologists. Today, these methodologies are mature (both in terms of performance and usability/implementation), and the genomic information that feeds them large enough to allow their general application. This review tries to summarize the current landscape of co-evolution-based methodologies, with a strong emphasis on describing interesting cases where their application to important biological systems, alone or in combination with other computational and experimental approaches, allowed getting new insight into these.

Generation of divergent uroplakin tetraspanins and their partners during vertebrate evolution: identification of novel uroplakins

BACKGROUND

The recent availability of sequenced genomes from a broad array of chordates (cephalochordates, urochordates and vertebrates) has allowed us to systematically analyze the evolution of uroplakins: tetraspanins (UPK1a and UPK1b families) and their respective partner proteins (UPK2 and UPK3 families).

RESULTS

We report here: (1) the origin of uroplakins in the common ancestor of vertebrates, (2) the appearance of several residues that have statistically significantly positive dN/dS ratios in the duplicated paralogs of uroplakin genes, and (3) the existence of strong coevolutionary relationships between UPK1a/1b tetraspanins and their respective UPK2/UPK3-related partner proteins. Moreover, we report the existence of three new UPK2/3 family members we named UPK2b, 3c and 3d, which will help clarify the evolutionary relationships between fish, amphibian and mammalian uroplakins that may perform divergent functions specific to these different and physiologically distinct groups of vertebrates.

CONCLUSIONS

Since our analyses cover species of all major chordate groups this work provides an extremely clear overall picture of how the uroplakin families and their partner proteins have evolved in parallel. We also highlight several novel features of uroplakin evolution including the appearance of UPK2b and 3d in fish and UPK3c in the common ancestor of reptiles and mammals. Additional studies of these novel uroplakins should lead to new insights into uroplakin structure and function.

Predicting protein-protein interaction by the mirrortree method: possibilities and limitations

Molecular co-evolution analysis as a sequence-only based method has been used to predict protein-protein interactions. In co-evolution analysis, Pearson's correlation within the mirrortree method is a well-known way of quantifying the correlation between protein pairs. Here we studied the mirrortree method on both known interacting protein pairs and sets of presumed non-interacting protein pairs, to evaluate the utility of this correlation analysis method for predicting protein-protein interactions within eukaryotes. We varied metrics for computing evolutionary distance and evolutionary span of the species analyzed. We found the differences between co-evolutionary correlation scores of the interacting and non-interacting proteins, normalized for evolutionary span, to be significantly predictive for proteins conserved over a wide range of eukaryotic clades (from mammals to fungi). On the other hand, for narrower ranges of evolutionary span, the predictive power was much weaker.

Citrus tristeza virus: Evolution of Complex and Varied Genotypic Groups

Amongst the Closteroviridae, Citrus tristeza virus (CTV) is almost unique in possessing a number of distinct and characterized strains, isolates of which produce a wide range of phenotype combinations among its different hosts. There is little understanding to connect genotypes to phenotypes, and to complicate matters more, these genotypes are found throughout the world as members of mixed populations within a single host plant. There is essentially no understanding of how combinations of genotypes affect symptom expression and disease severity. We know little about the evolution of the genotypes that have been characterized to date, little about the biological role of their diversity and particularly, about the effects of recombination. Additionally, genotype grouping has not been standardized. In this study we utilized an extensive array of CTV genomic information to classify the major genotypes, and to determine the major evolutionary processes that led to their formation and subsequent retention. Our analyses suggest that three major processes act on these genotypes: (1) ancestral diversification of the major CTV lineages, followed by (2) conservation and co-evolution of the major functional domains within, though not between CTV genotypes, and (3) extensive recombination between lineages that have given rise to new genotypes that have subsequently been retained within the global population. The effects of genotype diversity and host-interaction are discussed, as is a proposal for standardizing the classification of existing and novel CTV genotypes.

The origins of the evolutionary signal used to predict protein-protein interactions

Background

The correlation of genetic distances between pairs of protein sequence alignments has been used to infer protein-protein interactions. It has been suggested that these correlations are based on the signal of co-evolution between interacting proteins. However, although mutations in different proteins associated with maintaining an interaction clearly occur (particularly in binding interfaces and neighbourhoods), many other factors contribute to correlated rates of sequence evolution. Proteins in the same genome are usually linked by shared evolutionary history and so it would be expected that there would be topological similarities in their phylogenetic trees, whether they are interacting or not. For this reason the underlying species tree is often corrected for. Moreover processes such as expression level, are known to effect evolutionary rates. However, it has been argued that the correlated rates of evolution used to predict protein interaction explicitly includes shared evolutionary history; here we test this hypothesis.

Results

In order to identify the evolutionary mechanisms giving rise to the correlations between interaction proteins, we use phylogenetic methods to distinguish similarities in tree topologies from similarities in genetic distances. We use a range of datasets of interacting and non-interacting proteins from Saccharomyces cerevisiae. We find that the signal of correlated evolution between interacting proteins is predominantly a result of shared evolutionary rates, rather than similarities in tree topology, independent of evolutionary divergence.

Conclusions

Since interacting proteins do not have tree topologies that are more similar than the control group of non-interacting proteins, it is likely that coevolution does not contribute much to, if any, of the observed correlations.

A census of human soluble protein complexes

Cellular processes often depend on stable physical associations between proteins. Despite recent progress, knowledge of the composition of human protein complexes remains limited. To close this gap, we applied an integrative global proteomic profiling approach, based on chromatographic separation of cultured human cell extracts into more than one thousand biochemical fractions that were subsequently analyzed by quantitative tandem mass spectrometry, to systematically identify a network of 13,993 high-confidence physical interactions among 3,006 stably associated soluble human proteins. Most of the 622 putative protein complexes we report are linked to core biological processes and encompass both candidate disease genes and unannotated proteins to inform on mechanism. Strikingly, whereas larger multiprotein assemblies tend to be more extensively annotated and evolutionarily conserved, human protein complexes with five or fewer subunits are far more likely to be functionally unannotated or restricted to vertebrates, suggesting more recent functional innovations.

Coevolution reveals a network of human proteins originating with multicellularity

Protein interaction networks play central roles in biological systems, from simple metabolic pathways through complex programs permitting the development of organisms. Multicellularity could only have arisen from a careful orchestration of cellular and molecular roles and responsibilities, all properly controlled and regulated. Disease reflects a breakdown of this organismal homeostasis. To better understand the evolution of interactions whose dysfunction may be contributing factors to disease, we derived the human protein coevolution network using our MatrixMatchMaker algorithm and using the Orthologous MAtrix project (OMA) database as a source for protein orthologs from 103 eukaryotic genomes. We annotated the coevolution network using protein–protein interaction data, many functional data sources, and we explored the evolutionary rates and dates of emergence of the proteins in our data set. Strikingly, clustering based only on the topology of the coevolution network partitions it into two subnetworks, one generally representing ancient eukaryotic functions and the other functions more recently acquired during animal evolution. That latter subnetwork is enriched for proteins with roles in cell–cell communication, the control of cell division, and related multicellular functions. Further annotation using data from genetic disease databases and cancer genome sequences strongly implicates these proteins in both ciliopathies and cancer. The enrichment for such disease markers in the animal network suggests a functional link between these coevolving proteins. Genetic validation corroborates the recruitment of ancient cilia in the evolution of multicellularity.