Internal organization of large protein families: relationship between the sequence, structure, and function-based clustering

The protein universe can be organized in families that group proteins sharing common ancestry. Such families display variable levels of structural and functional divergence, from homogenous families, where all members have the same function and very similar structure, to very divergent families, where large variations in function and structure are observed. For practical purposes of structure and function prediction, it would be beneficial to identify sub-groups of proteins with highly similar structures (iso-structural) and/or functions (iso-functional) within divergent protein families. We compared three algorithms in their ability to cluster large protein families and discuss whether any of these methods could reliably identify such iso-structural or iso-functional groups. We show that clustering using profile-sequence and profile-profile comparison methods closely reproduces clusters based on similarities between 3D structures or clusters of proteins with similar biological functions. In contrast, the still commonly used sequence-based methods with fixed thresholds result in vast overestimates of structural and functional diversity in protein families. As a result, these methods also overestimate the number of protein structures that have to be determined to fully characterize structural space of such families. The fact that one can build reliable models based on apparently distantly related templates is crucial for extracting maximal amount of information from new sequencing projects.     

Mapping protein family interactions: intramolecular and intermolecular protein family interaction repertoires in the PDB and yeast

In the postgenomic era, one of the most interesting and important challenges is to understand protein interactions on a large scale. The physical interactions between protein domains are fundamental to the workings of a cell: in multi-domain polypeptide chains, in multi-subunit proteins and in transient complexes between proteins that also exist independently. To study the large-scale patterns and evolution of interactions between protein domains, we view interactions between protein domains in terms of the interactions between structural families of evolutionarily related domains. This allows us to classify 8151 interactions between individual domains in the Protein Data Bank and the yeast Saccharomyces cerevisiae in terms of 664 types of interactions, between protein families. At least 51 interactions do not occur in the Protein Data Bank and can only be derived from the yeast data. The map of interactions between protein families has the form of a scale-free network, meaning that most protein families only interact with one or two other families, while a few families are extremely versatile in their interactions and are connected to many families. We observe that almost half of all known families engage in interactions with domains from their own family. We also see that the repertoires of interactions of domains within and between polypeptide chains overlap mostly for two specific types of protein families: enzymes and same-family interactions. This suggests that different types of protein interaction repertoires exist for structural, functional and regulatory reasons. Copyright 12001 Academic Press.     

Hierarchical classification of glycoside hydrolases

This review deals with structural and functional features of glycoside hydrolases, a widespread group of enzymes present in almost all living organisms. Their catalytic domains are grouped into 120 amino acid sequence-based families in the international classification of the carbohydrate-active enzymes (CAZy database). At a higher hierarchical level some of these families are combined in 14 clans. Enzymes of the same clan have common evolutionary origin of their genes and share the most important functional characteristics such as composition of the active center, anomeric configuration of cleaved glycosidic bonds, and molecular mechanism of the catalyzed reaction (either inverting, or retaining). There are now extensive data in the literature concerning the relationship between glycoside hydrolase families belonging to different clans and/or included in none of them, as well as information on phylogenetic protein relationship within particular families. Summarizing these data allows us to propose a multilevel hierarchical classification of glycoside hydrolases and their homologs. It is shown that almost the whole variety of the enzyme catalytic domains can be brought into six main folds, large groups of proteins having the same three-dimensional structure and the supposed common evolutionary origin.     

Predicting substrate specificity of adenylation domains of nonribosomal peptide synthetases and other protein properties by latent semantic indexing

Successful genome mining is dependent on accurate prediction of protein function from sequence. This often involves dividing protein families into functional subtypes (e.g., with different substrates). In many cases, there are only a small number of known functional subtypes, but in the case of the adenylation domains of nonribosomal peptide synthetases (NRPS), there are >500 known substrates. Latent semantic indexing (LSI) was originally developed for text processing but has also been used to assign proteins to families. Proteins are treated as ‘‘documents’’ and it is necessary to encode properties of the amino acid sequence as ‘‘terms’’ in order to construct a term-document matrix, which counts the terms in each document. This matrix is then processed to produce a document-concept matrix, where each protein is represented as a row vector. A standard measure of the closeness of vectors to each other (cosines of the angle between them) provides a measure of protein similarity. Previous work encoded proteins as oligopeptide terms, i.e. counted oligopeptides, but used no information regarding location of oligopeptides in the proteins. A novel tokenization method was developed to analyze information from multiple alignments. LSI successfully distinguished between two functional subtypes in five well-characterized families. Visualization of different ‘‘concept’’ dimensions allows exploration of the structure of protein families. LSI was also used to predict the amino acid substrate of adenylation domains of NRPS. Better results were obtained when selected residues from multiple alignments were used rather than the total sequence of the adenylation domains. Using ten residues from the substrate binding pocket performed better than using 34 residues within 8 Å of the active site. Prediction efficiency was somewhat better than that of the best published method using a support vector machine.     

Structure-Based Phylogenetic Analysis of the Lipocalin Superfamily

Lipocalins constitute a superfamily of extracellular proteins that are found in all three kingdoms of life. Although very divergent in their sequences and functions, they show remarkable similarity in 3-D structures. Lipocalins bind and transport small hydrophobic molecules. Earlier sequence-based phylogenetic studies of lipocalins highlighted that they have a long evolutionary history. However the molecular and structural basis of their functional diversity is not completely understood. The main objective of the present study is to understand functional diversity of the lipocalins using a structure-based phylogenetic approach. The present study with 39 protein domains from the lipocalin superfamily suggests that the clusters of lipocalins obtained by structure-based phylogeny correspond well with the functional diversity. The detailed analysis on each of the clusters and sub-clusters reveals that the 39 lipocalin domains cluster based on their mode of ligand binding though the clustering was performed on the basis of gross domain structure. The outliers in the phylogenetic tree are often from single member families. Also structure-based phylogenetic approach has provided pointers to assign putative function for the domains of unknown function in lipocalin family. The approach employed in the present study can be used in the future for the functional identification of new lipocalin proteins and may be extended to other protein families where members show poor sequence similarity but high structural similarity.     

Optimal contact map alignment of protein-protein interfaces

The long-standing problem of constructing protein structure alignments is of central importance in computational biology. The main goal is to provide an alignment of residue correspondences, in order to identify homologous residues across chains. A critical next step of this is the alignment of protein complexes and their interfaces. Here, we introduce the program CMAPi, a two-dimensional dynamic programming algorithm that, given a pair of protein complexes, optimally aligns the contact maps of their interfaces: it produces polynomial-time near-optimal alignments in the case of multiple complexes. We demonstrate the efficacy of our algorithm on complexes from PPI families listed in the SCOPPI database and from highly divergent cytokine families. In comparison to existing techniques, CMAPi generates more accurate alignments of interacting residues within families of interacting proteins, especially for sequences with low similarity. While previous methods that use an all-atom based representation of the interface have been successful, CMAPi's use of a contact map representation allows it to be more tolerant to conformational changes and thus to align more of the interaction surface. These improved interface alignments should enhance homology modeling and threading methods for predicting PPIs by providing a basis for generating template profiles for sequence-structure alignment.     

Expanding the nitrogen regulatory protein superfamily: Homology detection at below random sequence identity

Nitrogen regulatory (PII) proteins are signal transduction molecules involved in controlling nitrogen metabolism in prokaryots. PII proteins integrate the signals of intracellular nitrogen and carbon status into the control of enzymes involved in nitrogen assimilation. Using elaborate sequence similarity detection schemes, we show that five clusters of orthologs (COGs) and several small divergent protein groups belong to the PII superfamily and predict their structure to be a (betaalphabeta)(2) ferredoxin-like fold. Proteins from the newly emerged PII superfamily are present in all major phylogenetic lineages. The PII homologs are quite diverse, with below random (as low as 1%) pairwise sequence identities between some members of distant groups. Despite this sequence diversity, evidence suggests that the different subfamilies retain the PII trimeric structure important for ligand-binding site formation and maintain a conservation of conservations at residue positions important for PII function. Because most of the orthologous groups within the PII superfamily are composed entirely of hypothetical proteins, our remote homology-based structure prediction provides the only information about them. Analogous to structural genomics efforts, such prediction gives clues to the biological roles of these proteins and allows us to hypothesize about locations of functional sites on model structures or rationalize about available experimental information. For instance, conserved residues in one of the families map in close proximity to each other on PII structure, allowing for a possible metal-binding site in the proteins coded by the locus known to affect sensitivity to divalent metal ions. Presented analysis pushes the limits of sequence similarity searches and exemplifies one of the extreme cases of reliable sequence-based structure prediction. In conjunction with structural genomics efforts to shed light on protein function, our strategies make it possible to detect homology between highly diverse sequences and are aimed at understanding the most remote evolutionary connections in the protein world.     

Exhaustive enumeration of protein domain families

Domains are considered as the basic units of protein folding, evolution, and function. Decomposing each protein into modular domains is thus a basic prerequisite for accurate functional classification of biological molecules. Here, we present ADDA, an automatic algorithm for domain decomposition and clustering of all protein domain families. We use alignments derived from an all-on-all sequence comparison to define domains within protein sequences based on a global maximum likelihood model. In all, 90% of domain boundaries are predicted within 10% of domain size when compared with the manual domain definitions given in the SCOP database. A representative database of 249,264 protein sequences were decomposed into 450,462 domains. These domains were clustered on the basis of sequence similarities into 33,879 domain families containing at least two members with less than 40% sequence identity. Validation against family definitions in the manually curated databases SCOP and PFAM indicates almost perfect unification of various large domain families while contamination by unrelated sequences remains at a low level. The global survey of protein-domain space by ADDA confirms that most large and universal domain families are already described in PFAM and/or SMART. However, a survey of the complete set of mobile modules leads to the identification of 1479 new interesting domain families which shuffle around in multi-domain proteins. The data are publicly available at ftp://ftp.ebi.ac.uk/pub/contrib/heger/adda.     

InterPro as a new tool for whole genome analysis. A comparative analysis of Mycobacterium tuberculosis, Bacillus subtilis and Escherichia coli as a case study

InterPro was developed as a new integrated documentation resource for protein families, domains and functional sites to rationalize the complementary efforts of the PROSITE, PRINTS, Pfam and ProDom database projects and has applications in computational functional classification of newly determined sequences lacking biochemical characterization and in comparative genome analysis. InterPro contains over 3500 entries, with more than 1000000 hits in SWISS-PROT and TrEMBL. The database is accessible for text- and sequence-based searches at http://www.ebi.ac.uk/interpro/. InterPro was used for whole proteome analysis of the pathogenic microorganism, Mycobacterium tuberculosis, and comparison with the predicted protein coding sequences of the complete genomes of Bacillus subtilis and Escherichia coli. 64.8% of the M. tuberculosis proteins in the proteome matched InterPro entries, and these could be classified according to function. The comparison with B. subtilis and E. coli provided information on the most common protein families and domains, and the most highly represented families in each organism. InterPro thus provides a useful tool for global views of whole proteomes and their compositions.     

Structural Analysis of Metal Sites in Proteins: Non-heme Iron Sites as a Case Study

In metalloproteins, the protein environment modulates metal properties to achieve the required goal, which can be protein stabilization or function. The analysis of metal sites at the atomic level of detail provided by protein structures can thus be of benefit in functional and evolutionary studies of proteins. In this work, we propose a structural bioinformatics approach to the study of metalloproteins based on structural templates of metal sites that include the PDB coordinates of protein residues forming the first and the second coordination sphere of the metal. We have applied this approach to non-heme iron sites, which have been analyzed at various levels. Templates of sites located in different protein domains have been compared, showing that similar sites can be found in unrelated proteins as the result of convergent evolution. Templates of sites located in proteins of a large superfamily have been compared, showing possible mechanisms of divergent evolution of proteins to achieve different functions. Furthermore, template comparisons have been used to predict the function of uncharacterized proteins, showing that similarity searches focused on metal sites can be advantageously combined with typical whole-domain comparisons. Structural templates of metal sites, finally, may constitute the basis for a systematic classification of metalloproteins in databases.     

The natural history of Get3‐like chaperones

Get3 in yeast or TRC40 in mammals is an ATPase that, in eukaryotes, is a central element of the GET or TRC pathway involved in the targeting of tail‐anchored proteins. Get3 has also been shown to possess chaperone holdase activity. A bioinformatic assessment was performed across all domains of life on functionally important regions of Get3 including the TRC40‐insert and the hydrophobic groove essential for tail‐anchored protein binding. We find that such a hydrophobic groove is much more common in bacterial Get3 homologs than previously appreciated based on a directed comparison of bacterial ArsA and yeast Get3. Furthermore, our analysis shows that the region containing the TRC40‐insert varies in length and methionine content to an unexpected extent within eukaryotes and also between different phylogenetic groups. In fact, since the TRC40‐insert is present in all domains of life, we suggest that its presence does not automatically predict a tail‐anchored protein targeting function. This opens up a new perspective on the function of organellar Get3 homologs in plants which feature the TRC40‐insert but have not been demonstrated to function in tail‐anchored protein targeting. Our analysis also highlights a large diversity of the ways Get3 homologs dimerize. Thus, based on the structural features of Get3 homologs, these proteins may have an unexplored functional diversity in all domains of life.      

Protein families and TRIBES in genome sequence space

Accurate detection of protein families allows assignment of protein function and the analysis of functional diversity in complete genomes. Recently, we presented a novel algorithm called TribeMCL for the detection of protein families that is both accurate and efficient. This method allows family analysis to be carried out on a very large scale. Using TribeMCL, we have generated a resource called TRIBES that contains protein family information, comprising annotations, protein sequence alignments and phylogenetic distributions describing 311 257 proteins from 83 completely sequenced genomes. The analysis of at least 60 934 detected protein families reveals that, with the essential families excluded, paralogy levels are similar between prokaryotes, irrespective of genome size. The number of essential families is estimated to be between 366 and 426. We also show that the currently known space of protein families is scale free and discuss the implications of this distribution. In addition, we show that smaller families are often formed by shorter proteins and discuss the reasons for this intriguing pattern. Finally, we analyse the functional diversity of protein families in entire genome sequences. The TRIBES protein family resource is accessible at http://www.ebi.ac.uk/research/cgg/tribes/.     

An efficient algorithm for large-scale detection of protein families

Detection of protein families in large databases is one of the principal research objectives in structural and functional genomics. Protein family classification can significantly contribute to the delineation of functional diversity of homologous proteins, the prediction of function based on domain architecture or the presence of sequence motifs as well as comparative genomics, providing valuable evolutionary insights. We present a novel approach called TRIBE-MCL for rapid and accurate clustering of protein sequences into families. The method relies on the Markov cluster (MCL) algorithm for the assignment of proteins into families based on precomputed sequence similarity information. This novel approach does not suffer from the problems that normally hinder other protein sequence clustering algorithms, such as the presence of multi-domain proteins, promiscuous domains and fragmented proteins. The method has been rigorously tested and validated on a number of very large databases, including SwissProt, InterPro, SCOP and the draft human genome. Our results indicate that the method is ideally suited to the rapid and accurate detection of protein families on a large scale. The method has been used to detect and categorise protein families within the draft human genome and the resulting families have been used to annotate a large proportion of human proteins.     

SURF’s UP! — Protein classification by surface comparisons

Large-scale genome sequencing and structural genomics projects generate numerous sequences and structures for 'hypothetical' proteins without functional characterizations. Detection of homology to experimentally characterized proteins can provide functional clues, but the accuracy of homology-based predictions is limited by the paucity of tools for quantitative comparison of diverging residues responsible for the functional divergence. SURF'S UP! is a web server for analysis of functional relationships in protein families, as inferred from protein surface maps comparison according to the algorithm. It assigns a numerical score to the similarity between patterns of physicochemical features(charge, hydrophobicity) on compared protein surfaces. It allows recognizing clusters of proteins that have similar surfaces, hence presumably similar functions. The server takes as an input a set of protein coordinates and returns files with "spherical coordinates" of proteins in a PDB format and their graphical presentation, a matrix with values of mutual similarities between the surfaces, and the unrooted tree that represents the clustering of similar surfaces, calculated by the neighbor-joining method. SURF'S UP! facilitates the comparative analysis of physicochemical features of the surface, which are the key determinants of the protein function. By concentrating on coarse surface features, SURF'S UP! can work with models obtained from comparative modelling. Although it is designed to analyse the conservation among homologs, it can also be used to compare surfaces of non-homologous proteins with different three-dimensional folds, as long as a functionally meaningful structural superposition is supplied by the user. Another valuable characteristic of our method is the lack of initial assumptions about the functional features to be compared. SURF'S UP! is freely available for academic researchers at http://asia.genesilico.pl/surfs_up/.     

A computational model of the LGI1 protein suggests a common binding site for ADAM proteins

Mutations of human leucine-rich glioma inactivated (LGI1) gene encoding the epitempin protein cause autosomal dominant temporal lateral epilepsy (ADTLE), a rare familial partial epileptic syndrome. The LGI1 gene seems to have a role on the transmission of neuronal messages but the exact molecular mechanism remains unclear. In contrast to other genes involved in epileptic disorders, epitempin shows no homology with known ion channel genes but contains two domains, composed of repeated structural units, known to mediate protein-protein interactions.A three dimensional in silico model of the two epitempin domains was built to predict the structure-function relationship and propose a functional model integrating previous experimental findings. Conserved and electrostatic charged regions of the model surface suggest a possible arrangement between the two domains and identifies a possible ADAM protein binding site in the β-propeller domain and another protein binding site in the leucine-rich repeat domain. The functional model indicates that epitempin could mediate the interaction between proteins localized to different synaptic sides in a static way, by forming a dimer, or in a dynamic way, by binding proteins at different times.The model was also used to predict effects of known disease-causing missense mutations. Most of the variants are predicted to alter protein folding while several other map to functional surface regions. In agreement with experimental evidence, this suggests that non-secreted LGI1 mutants could be retained within the cell by quality control mechanisms or by altering interactions required for the secretion process.     

Comparison of sequence profiles. Strategies for structural predictions using sequence information

Distant homologies between proteins are often discovered only after three-dimensional structures of both proteins are solved. The sequence divergence for such proteins can be so large that simple comparison of their sequences fails to identify any similarity. New generation of sensitive alignment tools use averaged sequences of entire homologous families (profiles) to detect such homologies. Several algorithms, including the newest generation of BLAST algorithms and BASIC, an algorithm used in our group to assign fold predictions for proteins from several genomes, are compared to each other on the large set of structurally similar proteins with little sequence similarity. Proteins in the benchmark are classified according to the level of their similarity, which allows us to demonstrate that most of the improvement of the new algorithms is achieved for proteins with strong functional similarities, with almost no progress in recognizing distant fold similarities. It is also shown that details of profile calculation strongly influence its sensitivity in recognizing distant homologies. The most important choice is how to include information from diverging members of the family, avoiding generating false predictions, while accounting for entire sequence divergence within a family. PSI-BLAST takes a conservative approach, deriving a profile from core members of the family, providing a solid improvement without almost any false predictions. BASIC strives for better sensitivity by increasing the weight of divergent family members and paying the price in lower reliability. A new FFAS algorithm introduced here uses a new procedure for profile generation that takes into account all the relations within the family and matches BASIC sensitivity with PSI-BLAST like reliability.     

Patterns of ribosomal RNA evolution in salamanders

Sequence comparisons are presented for four segments of the large subunit of ribosomal RNA, including divergent domains D7a and D7b, portions of the large divergent domains D2, D3, and D8, and evolutionarily conservative sequences flanking divergent domains. These results resolve phylogenetic relationships among exemplars of seven families of salamanders and the three amphibian orders. Phylogenetic analysis confirms the prediction that divergent domains feature the highest relative rates of base substitution and length variation within the ribosome, but the divergent domains evolve more slowly than nuclear noncoding DNA and the silent sites of structural genes. Base substitutions demonstrate approximately twice as many transitions as transversions and an uneven distribution among sites within the divergent domains but no apparent bias in base composition. Length mutations are primarily small insertions and deletions, with deletions predominating. The divergent domains appear to be a good source of phylogenetic information for evolutionary events occurring approximately 100-200 million years ago.      

Automated protein subfamily identification and classification

Function prediction by homology is widely used to provide preliminary functional annotations for genes for which experimental evidence of function is unavailable or limited. This approach has been shown to be prone to systematic error, including percolation of annotation errors through sequence databases. Phylogenomic analysis avoids these errors in function prediction but has been difficult to automate for high-throughput application. To address this limitation, we present a computationally efficient pipeline for phylogenomic classification of proteins. This pipeline uses the SCI-PHY (Subfamily Classification in Phylogenomics) algorithm for automatic subfamily identification, followed by subfamily hidden Markov model (HMM) construction. A simple and computationally efficient scoring scheme using family and subfamily HMMs enables classification of novel sequences to protein families and subfamilies. Sequences representing entirely novel subfamilies are differentiated from those that can be classified to subfamilies in the input training set using logistic regression. Subfamily HMM parameters are estimated using an information-sharing protocol, enabling subfamilies containing even a single sequence to benefit from conservation patterns defining the family as a whole or in related subfamilies. SCI-PHY subfamilies correspond closely to functional subtypes defined by experts and to conserved clades found by phylogenetic analysis. Extensive comparisons of subfamily and family HMM performances show that subfamily HMMs dramatically improve the separation between homologous and non-homologous proteins in sequence database searches. Subfamily HMMs also provide extremely high specificity of classification and can be used to predict entirely novel subtypes. The SCI-PHY Web server at http://phylogenomics.berkeley.edu/SCI-PHY/ allows users to upload a multiple sequence alignment for subfamily identification and subfamily HMM construction. Biologists wishing to provide their own subfamily definitions can do so. Source code is available on the Web page. The Berkeley Phylogenomics Group PhyloFacts resource contains pre-calculated subfamily predictions and subfamily HMMs for more than 40,000 protein families and domains at http://phylogenomics.berkeley.edu/phylofacts/.     

A protein map and its application

Graphical representation of gene sequences provides a simple way of viewing, sorting, and comparing various gene structures. Here we first report a two-dimensional graphical representation for protein sequences. With this method, we constructed the moment vectors for protein sequences, and mathematically proved that the correspondence between moment vectors and protein sequences is one-to-one. Therefore, each protein sequence can be represented as a point in a map, which we call protein map, and cluster analysis can be used for comparison between the points. Sixty-six proteins from five protein families were analyzed using this method. Our data showed that for proteins in the same family, their corresponding points in the map are close to each other. We also illustrate the efficiency of this approach by performing an extensive cluster analysis of the protein kinase C family. These results indicate that this protein map could be used to mathematically specify the similarity of two proteins and predict properties of an unknown protein based on its amino acid sequence.     

Structure-based function inference using protein family-specific fingerprints

We describe a method to assign a protein structure to a functional family using family-specific fingerprints. Fingerprints represent amino acid packing patterns that occur in most members of a family but are rare in the background, a nonredundant subset of PDB; their information is additional to sequence alignments, sequence patterns, structural superposition, and active-site templates. Fingerprints were derived for 120 families in SCOP using Frequent Subgraph Mining. For a new structure, all occurrences of these family-specific fingerprints may be found by a fast algorithm for subgraph isomorphism; the structure can then be assigned to a family with a confidence value derived from the number of fingerprints found and their distribution in background proteins. In validation experiments, we infer the function of new members added to SCOP families and we discriminate between structurally similar, but functionally divergent TIM barrel families. We then apply our method to predict function for several structural genomics proteins, including orphan structures. Some predictions have been corroborated by other computational methods and some validated by subsequent functional characterization.