Protein sequence comparison and fold recognition: progress and good-practice benchmarking

Söding, Johannes; Remmert, Michael

doi:10.1016/j.sbi.2011.03.005

Cited by 79 publications

(75 citation statements)

References 68 publications

(63 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Exceptions to this were members of Rossman-like folds (c.2-c.5, c.27, c.28, c.30 and c.31) and the four-to eight-bladed β-propellers (b.66-b.70), which are probably related and which we treated as 'unknown' . To prevent a few large folds from dominating the benchmark 4 , we weighed each hit with the value of one over the number of members in the query SCOP fold ('fold-weighted true positives and false positives'). All but the last search iteration were performed against the UniProt database.…”

Section: Adding Sequences From Significant Matches To the Query Hmmmentioning

confidence: 99%

“…Profile-profile and HMM-HMM alignment are the most sensitive classes of sequence-search methods. They are the methods of choice for identifying and aligning templates for threedimensional homology modeling 4 . Our HMM-HMM alignment method HHsearch 5 is used by many of the best protein structure prediction servers, among which is HHpred 6 , the top-ranked server for template-based protein structure prediction in last year's Critical Assessment of Techniques for Protein Structure Prediction exercise (http://predictioncenter.org/casp9/groups_ analysis.cgi?type=server&tbm=on/).…”

mentioning

confidence: 99%

See 1 more Smart Citation

HHblits: lightning-fast iterative protein sequence searching by HMM-HMM alignment

et al. 2011

Self Cite

View full text Add to dashboard Cite

are generally too slow for iteratively searching through large sequence databases such as UniProt or NCBI's nonredundant (nr) database. Here we present HMM-HMM-based lightningfast iterative sequence search (HHblits), which extends HHsearch to enable fast, iterative sequence searches. The profile-profile alignment prefilter of HHblits reduces the number of full HMM-HMM alignments from many millions to a few thousand, making it faster than PSI-BLAST but still as sensitive as HHsearch (Supplementary Fig. 1).For iterative searches, HHblits needs a database of HMMs that covers the entire sequence space. We devised a very fast method, kClust (M. Hauser, C.E. Mayer and J.S., unpublished data), for clustering large sequence databases down to 20-30% maximum pairwise sequence identity while requiring almost full-length alignability (>80% coverage of longer sequences). This strict coverage criterion enriches for orthologous sequences with the same domain architecture 7 : of the UniProt20 clusters containing more than two Swiss-Prot sequences with enzyme commission numbers, 98.4% had all four enzyme commission digits conserved ( Supplementary Fig. 2). kClust is sufficiently fast (~1,000 times faster than BLAST) to allow for regular reclustering of the updated UniProt and nr databases. UniProt20 (the version from July 2011) contained 15 million sequences in 2.6 million HMMs, with an average of 5.5 sequences per cluster.HHblits first converts the query sequence (or MSA) to an HMM. This is conventionally done by adding pseudocounts of amino acids that are physicochemically similar to the amino acid in the query. In contrast, HHblits calculates pseudocounts that depend on the local sequence context (that is, the 13 positions around each residue). This method had improved the sensitivity and alignment quality of the resulting profile considerably 8 . HHblits then searches the HMM database and adds the sequences from HMMs below a defined expected value (E value) threshold to the query MSA, from which the HMM for the next search iteration is built ( Fig. 1a and Supplementary Fig. 3). For speed and sensitivity, the prefilter is crucial. The key idea was to implement profile-profile comparison as a sequence-to-profile comparison by discretizing the vectors of 20 amino acid probabilities in each HMM column into an alphabet of 219 letters. Each letter represents a typical profile column ( Supplementary Fig. 4). We approximate the database HMMs by sequences over this extended alphabet, ignoring the insertion and deletion probabilities of the HMMs (Supplementary Fig. 5). Before prefiltering, we calculate the score of each query HMM column with each of the 219 letters, which results in a 219-row extended sequence profile. The prefiltering consists of two steps (Supplementary Fig. 3 Building protein multiple-sequence alignments (MSAs) by iterative sequence searches is of fundamental importance in computational biology, as MSAs are a key intermediate step in the sequence-based prediction of evolutionarily conserved properties, such as tert...

show abstract

Section: Adding Sequences From Significant Matches To the Query Hmmmentioning

confidence: 99%

mentioning

confidence: 99%

HHblits: lightning-fast iterative protein sequence searching by HMM-HMM alignment

et al. 2011

Self Cite

View full text Add to dashboard Cite

show abstract

“…Many modeling web servers and programs simply present a list of templates with their PDB codes and perhaps the protein names but the user has no way of knowing whether the hits all come from the same protein family and whether they would produce the same protein folds or entirely different ones. Because BAM finds the Pfams of the target and the Pfams of the PDB proteins in separate steps, the procedure is effectively an intermediate sequence profile search [23], where the intermediate profile is the Pfam HMM. This is an effective way of finding remote homologues in a computationally efficient manner.…”

Section: Resultsmentioning

confidence: 99%

BioAssemblyModeler (BAM): User-Friendly Homology Modeling of Protein Homo- and Heterooligomers

Shapovalov

Wang

et al. 2014

PLoS ONE

View full text Add to dashboard Cite

Many if not most proteins function in oligomeric assemblies of one or more protein sequences. The Protein Data Bank provides coordinates for biological assemblies for each entry, at least 60% of which are dimers or larger assemblies. BioAssemblyModeler (BAM) is a graphical user interface to the basic steps in homology modeling of protein homooligomers and heterooligomers from the biological assemblies provided in the PDB. BAM takes as input up to six different protein sequences and begins by assigning Pfam domains to the target sequences. The program utilizes a complete assignment of Pfam domains to sequences in the PDB, PDBfam (http://dunbrack2.fccc.edu/protcid/pdbfam), to obtain templates that contain any or all of the domains assigned to the target sequence(s). The contents of the biological assemblies of potential templates are provided, and alignments of the target sequences to the templates are produced with a profile-profile alignment algorithm. BAM provides for visual examination and mouse-editing of the alignments supported by target and template secondary structure information and a 3D viewer of the template biological assembly. Side-chain coordinates for a model of the biological assembly are built with the program SCWRL4. A built-in protocol navigation system guides the user through all stages of homology modeling from input sequences to a three-dimensional model of the target complex. Availability: http://dunbrack.fccc.edu/BAM.

show abstract

“…As it is highlighted in [49], the most sensitive methods for fold recognition use sequence profiles to represent both the query and the data base proteins. The robustness and sensitivity of PSSM and SPINE-X for feature extraction have been addressed in [23], [46], [49]. In continuation, the global and local features extracted in this study will be explained in detail.…”

Section: Feature Extraction Methodsmentioning

confidence: 99%

A Segmentation-Based Method to Extract Structural and Evolutionary Features for Protein Fold Recognition

Dehzangi

Paliwal

Lyons

et al. 2014

IEEE/ACM Trans. Comput. Biol. and Bioinf.

View full text Add to dashboard Cite

Abstract-Protein fold recognition (PFR) is considered as an important step towards the protein structure prediction problem. Despite all the efforts that have been made so far, finding an accurate and fast computational approach to solve the PFR still remains a challenging problem for bioinformatics and computational biology. In this study, we propose the concept of segmented-based feature extraction technique to provide local evolutionary information embedded in position specific scoring matrix (PSSM) and structural information embedded in the predicted secondary structure of proteins using SPINE-X. We also employ the concept of occurrence feature to extract global discriminatory information from PSSM and SPINE-X. By applying a support vector machine (SVM) to our extracted features, we enhance the protein fold prediction accuracy for 7.4 percent over the best results reported in the literature. We also report 73.8 percent prediction accuracy for a data set consisting of proteins with less than 25 percent sequence similarity rates and 80.7 percent prediction accuracy for a data set with proteins belonging to 110 folds with less than 40 percent sequence similarity rates. We also investigate the relation between the number of folds and the number of features being used and show that the number of features should be increased to get better protein fold prediction results when the number of folds is relatively large.

show abstract

Protein sequence comparison and fold recognition: progress and good-practice benchmarking

Cited by 79 publications

References 68 publications

HHblits: lightning-fast iterative protein sequence searching by HMM-HMM alignment

HHblits: lightning-fast iterative protein sequence searching by HMM-HMM alignment

BioAssemblyModeler (BAM): User-Friendly Homology Modeling of Protein Homo- and Heterooligomers

A Segmentation-Based Method to Extract Structural and Evolutionary Features for Protein Fold Recognition

Contact Info

Product

Resources

About