Use of designed sequences in protein structure recognition

BMC Bioinformatics, 2012

Background: Identification of protein structural cores requires isolation of sets of proteins all sharing a same subset of structural motifs. In the context of an ever growing number of available 3D protein structures, standard and automatic clustering algorithms require adaptations so as to allow for efficient identification of such sets of proteins. Results: When considering a pair of 3D structures, they are stated as similar or not according to the local similarities of their matching substructures in a structural alignment. This binary relation can be represented in a graph of similarities where a node represents a 3D protein structure and an edge states that two 3D protein structures are similar. Therefore, classifying proteins into structural families can be viewed as a graph clustering task. Unfortunately, because such a graph encodes only pairwise similarity information, clustering algorithms may include in the same cluster a subset of 3D structures that do not share a common substructure. In order to overcome this drawback we first define a ternary similarity on a triple of 3D structures as a constraint to be satisfied by the graph of similarities. Such a ternary constraint takes into account similarities between pairwise alignments, so as to ensure that the three involved protein structures do have some common substructure. We propose hereunder a modification algorithm that eliminates edges from the original graph of similarities and gives a reduced graph in which no ternary constraints are violated. Our approach is then first to build a graph of similarities, then to reduce the graph according to the modification algorithm, and finally to apply to the reduced graph a standard graph clustering algorithm. Such method was used for classifying ASTRAL-40 non-redundant protein domains, identifying significant pairwise similarities with Yakusa, a program devised for rapid 3D structure alignments.

Proteins: Structure, Function, and Bioinformatics, 2008

We present a comprehensive evaluation of a new structure mining method called PB-ALIGN. It is based on the encoding of protein structure as 1D sequence of a combination of 16 short structural motifs or protein blocks (PBs). PBs are short motifs capable of representing most of the local structural features of a protein backbone. Using derived PB substitution matrix and simple dynamic programming algorithm, PB sequences are aligned the same way amino acid sequences to yield structure alignment. PBs are short motifs capable of representing most of the local structural features of a protein backbone. Alignment of these local features as sequence of symbols enables fast detection of structural similarities between two proteins. Ability of the method to characterize and align regions beyond regular secondary structures, for example, N and C caps of helix and loops connecting regular structures, puts it a step ahead of existing methods, which strongly rely on secondary structure elements. PB-ALIGN achieved efficiency of 85% in extracting true fold from a large database of 7259 SCOP domains and was successful in 82% cases to identify true super-family members. On comparison to 13 existing structure comparison/mining methods, PB-ALIGN emerged as the best on general ability test dataset and was at par with methods like YAKUSA and CE on nontrivial test dataset. Furthermore, the proposed method performed well when compared to flexible structure alignment method like FATCAT and outperforms in processing speed (less than 45 s per database scan). This work also establishes a reliable cut-off value for the demarcation of similar folds. It finally shows that global alignment scores of unrelated structures using PBs follow an extreme value distribution. PB-ALIGN is freely available on web server called Protein Block Expert (PBE) at http://bioinformatics.univ-reunion.fr/PBE/. Proteins 2008; 71:920-937. V V C 2007 Wiley-Liss, Inc.

Nucleic Acids Research, 2002

Members of a superfamily of proteins could result from divergent evolution of homologues with insignificant similarity in the amino acid sequences. A superfamily relationship is detected commonly after the threedimensional structures of the proteins are determined using X-ray analysis or NMR. The SUPFAM database described here relates two homologous protein families in a multiple sequence alignment database of either known or unknown structure. The present release (1.1), which is the first version of the SUPFAM database, has been derived by analysing Pfam, which is one of the commonly used databases of multiple sequence alignments of homologous proteins. The first step in establishing SUPFAM is to relate Pfam families with the families in PALI, which is an alignment database of homologous proteins of known structure that is derived largely from SCOP. The second step involves relating Pfam families which could not be associated reliably with a protein superfamily of known structure. The profile matching procedure, IMPALA, has been used in these steps. The first step resulted in identification of 1280 Pfam families (out of 2697, i.e. 47%) which are related, either by close homologous connection to a SCOP family or by distant relationship to a SCOP family, potentially forming new superfamily connections. Using the profiles of 1417 Pfam families with apparently no structural information, an all-against-all comparison involving a sequence-profile match using IMPALA resulted in clustering of 67 homologous protein families of Pfam into 28 potential new superfamilies. Expansion of groups of related proteins of yet unknown structural information, as proposed in SUPFAM, should help in identifying 'priority proteins' for structure determination in structural genomics initiatives to expand the coverage of structural information in the protein sequence space. For example, we could assign 858 distinct Pfam domains in 2203 of the gene products in the genome of Mycobacterium tubercolosis. Fifty-one of these Pfam families of unknown structure could be clustered into 17 potentially new superfamilies forming good targets for structural genomics. SUPFAM database can be accessed at http://pauling.mbu.iisc.ernet.in/~supfam.

Protein Science, 1992

The availability of fast and robust algorithms for protein structure comparison provides an opportunity to produce a database of three-dimensional comparisons, called families of structurally similar proteins (FSSP). The database currently contains an extended structural family for each of 154 representative (below 30% sequence identity) protein chains. Each data set contains: the search structure; all its relatives with 70–30% sequence identity, aligned structurally; and all other proteins from the representative set that contain substructures significantly similar to the search structure. Very close relatives (above 70% sequence identity) rarely have significant structural differences and are excluded. The alignments of remote relatives are the result of pairwise all-against-all structural comparisons in the set of 154 representative protein chains. The comparisons were carried out with each of three novel automatic algorithms that cover different aspects of protein structure similarity. The user of the database has the choice between strict rigid-body comparisons and comparisons that take into account interdomain motion or geometrical distortions; and, between comparisons that require strictly sequential ordering of segments and comparisons, which allow altered topology of loop connections or chain reversals. The data sets report the structurally equivalent residues in the form of a multiple alignment and as a list of matching fragments to facilitate inspection by three-dimensional graphics. If substructures are ignored, the result is a database of structure alignments of full-length proteins, including those in the twilight zone of sequence similarity. The database makes explicitly visible architectural similarities in the known part of the universe of protein folds and may be useful for understanding protein folding and for extracting structural modules for protein design. The data sets are available via Internet.

European journal of cancer (Oxford, England : 1990), 1990

Methods to predict the three-dimensional structure of a protein from its sequence are reviewed. The approaches to derive information about the local conformation from the local sequence include hydrophobic@ plots, secondary structure prediction and the identification of short, functional sequence motifs. The most reliable method of tertiary structure prediction is model building from the experimentally determined coordinates of a protein with an homologous sequence. This approach is illustrated by a prediction of the three-dimensional structure of human cytochrome P450-IAl. If no known homologous structure is available, then the only approach is to suggest models for the tertiary fold of proteins by packing together predicted secondary structures. A threedimensional model for the dimerisation of the transmembrane cw-helices of neu, a tyrosine kinase growth factor receptor, is described. In general, structure prediction can suggest approaches for regulating protein activity that may lead to new pharmaceutical therapies for cancer.

The FASEB Journal, 1996

With the advent of genome sequencing projects, the amino acid sequences of thousands of proteins are determined every year. Each of these protein sequences must be identifIed with its function and its 3-dimensional structure for us to gain a full understanding of the molecular biology of organisms. To meet this challenge, new methods are being developed for fold recognition, the computational assignment of newly determined amino acid sequences to 3-dimensional protein structures. These methods start with a library of known 3-dimensional target protein structures. The new probe sequence is then aligned to each target protein structure in the library and the compatibility of the sequence for that structure is scored. If a target structure is found to have a significantly high compatibility score, it is assumed that the probe sequence folds in much the same way as the target structure. The fundamental assumptions of this approach are that many different sequences fold in similar ways and there is a relatively high probability that a new sequence possesses a previously observed fold. We review various approaches to fold recognition and break down the process into its main steps: creation of a library of target folds; representation of the folds; alignment of the probe sequence to a target fold using a sequence-to-structure compatibility scoring function; and assessment of significance of compatibility. We emphasize that even though this new field of fold recognition has made rapid progress, technical problems remain to be solved in most of the steps. Standard benchmarks may help identify the problem steps and find solutions to the problems.

Protein Science, 2002

The elucidation of the domain content of a given protein sequence in the absence of determined structure or significant sequence homology to known domains is an important problem in structural biology. Here we address how successfully the delineation of continuous domains can be accomplished in the absence of sequence homology using simple baseline methods, an existing prediction algorithm (Domain Guess by Size), and a newly developed method (DomSSEA). The study was undertaken with a view to measuring the usefulness of these prediction methods in terms of their application to fully automatic domain assignment. Thus, the sensitivity of each domain assignment method was measured by calculating the number of correctly assigned top scoring predictions. We have implemented a new continuous domain identification method using the alignment of predicted secondary structures of target sequences against observed secondary structures of chains with known domain boundaries as assigned by Class Architecture Topology Homology (CATH). Taking top predictions only, the success rate of the method in correctly assigning domain number to the representative chain set is 73.3%. The top prediction for domain number and location of domain boundaries was correct for 24% of the multidomain set (±20 residues). These results have been put into context in relation to the results obtained from the other prediction methods assessed.

Structure, 2005

Debnath Pal and David Eisenberg* teractions between these functions form the basis for sustainable homeostasis. These multiple levels of func-UCLA-DOE Institute for Genomics and Proteomics tion are reflected in our procedure, described below, of Howard Hughes Medical Institute linking protein features to annotations at various levels. Box 951570 The repertoire of methods for in silico annotation of Los Angeles, California 90095 function has grown enormously over the past two decades. A protein with a high degree of sequence similarity to a family of well-characterized proteins can be Summary detected by BLAST (Altschul et al., 1990). With lower sequence similarity, more subtle methods such as "pro-Structural genomics has brought us three-dimensional files" (where patterns obvious from multiple sequence structures of proteins with unknown functions. To shed alignment are evident) (Altschul et al., 1997; Bork and light on such structures, we have developed ProKnow Gibson, 1996; Gribskov et al., 1987) or hidden Markov (http://www.doe-mbi.ucla.edu/Services/ProKnow/), which models (HMM) (Eddy et al., 1995) are required. These annotates proteins with Gene Ontology functional methods are based on the assumption that similar seterms. The method extracts features from the protein quences have descended from a common ancestor such as 3D fold, sequence, motif, and functional linkand share similar function. The assumption is, howages and relates them to function via the ProKnow ever, limited in validity, as demonstrated by numerous knowledgebase of features, which links features to studies (Devos and Valencia, 2000; Gerlt and Babbitt, annotated functions via annotation profiles. Bayes' 2000; Karp, 1998; Rost, 2002; Rost et al., 2003; Rost theorem is used to compute weights of the functions and Valencia, 1996; Tian and Skolnick, 2003; Whisstock assigned, using likelihoods based on the extracted and Lesk, 2003). To enhance accuracy of functional asfeatures. The description level of the assigned funcsignment, functional annotations can be inferred from tion is quantified by the ontology depth (from 1 = information on fold (Bowie et al., 1991; Holm and general to 9 = specific). Jackknife tests show 89% Sander, 1998; Jones et al., 1992), motif (Attwood et al., correct assignments at ontology depth 1 and 40% at 2003; Henikoff et al., 2000; Hulo et al., 2004), domain depth 9, with 93% coverage of 1507 distinct folded (Bateman et al., 2004), and orthology (Tatusov et al., proteins. Overall, about 70% of the assignments were 1997). Another class of annotation algorithms infers inferred correctly. This level of performance suggests protein function based on identification of functionally that ProKnow is a useful resource in functional assignificant residues. This class includes biodictionary sessments of novel proteins. "seqlets" mapping sequence patterns to their properties (Rigoutsos et al., 2002), evolutionary tracing (Land

Protein Science, 2000

We have developed a method for the prediction of an amino acid sequence that is compatible with a three-dimensional backbone structure. Using only a backbone structure of a protein as input, the algorithm is capable of designing sequences that closely resemble natural members of the protein family to which the template structure belongs. In general, the predicted sequences are shown to have multiple sequence profile scores that are dramatically higher than those of random sequences, and sometimes better than some of the natural sequences that make up the superfamily. As anticipated, highly conserved but poorly predicted residues are often those that contribute to the functional rather than structural properties of the protein. Overall, our analysis suggests that statistical profile scores of designed sequences are a novel and valuable figure of merit for assessing and improving protein design algorithms.

We describe the development of a scoring function based on the decomposition P(structure 0 sequence) ϰ P(sequence 0 structure) *P(structure), which outperforms previous scoring functions in correctly identifying native-like protein structures in large ensembles of compact decoys. The first term captures sequence-dependent features of protein structures, such as the burial of hydrophobic residues in the core, the second term, universal sequence-independent features, such as the assembly of ␤-strands into ␤-sheets. The efficacies of a wide variety of sequence-dependent and sequence-independent features of protein structures for recognizing native-like structures were systematically evaluated using ensembles of D30,000 compact conformations with fixed secondary structure for each of 17 small protein domains. The best results were obtained using a core scoring function with P(sequence 0 structure) parameterized similarly to our previous work (Simons et al., J Mol Biol 1997;268:209-225] and P(structure) focused on secondary structure packing preferences; while several additional features had some discriminatory power on their own, they did not provide any additional discriminatory power when combined with the core scoring function. Our results, on both the training set and the independent decoy set of Park and Levitt (J Mol Biol 1996;258:367-392), suggest that this scoring function should contribute to the prediction of tertiary structure from knowledge of sequence and secondary structure. Proteins 1999;34:82-95.