Impact of AlphaFold on structure prediction of protein complexes: The CASP15-CAPRI experiment

We present the results for CAPRI Round 54, the 5th joint CASP-CAPRI protein assembly prediction challenge. The Round offered 37 targets, including 14 homodimers, 3 homo-trimers, 13 heterodimers including 3 antibody-antigen complexes, and 7 large assemblies. On average ~70 CASP and CAPRI predictor groups, including more than 20 automatics servers, submitted models for each target. A total of 21 941 models submitted by these groups and by 15 CAPRI scorer groups were evaluated using the CAPRI model quality measures and the DockQ score consolidating these measures. The prediction performance was quantified by a weighted score based on the number of models of acceptable quality or higher submitted by each group among their five best models. Results show substantial progress achieved across a significant fraction of the 60+ participating groups. High-quality models were produced for about 40% of the targets compared to 8% two years earlier. This remarkable improvement is due to the wide use of the AlphaFold2 and AlphaFold2-Multimer software and the confidence metrics they provide. Notably, expanded sampling of candidate solutions by manipulating these deep learning inference engines, enriching multiple sequence alignments, or integration of advanced modeling tools, enabled top performing groups to exceed the performance of a standard AlphaFold2-Multimer version used as a yard stick. This notwithstanding, performance remained poor for complexes with antibodies and nanobodies, where evolutionary relationships between the binding partners are lacking, and for complexes featuring conformational flexibility, clearly indicating that the prediction of protein complexes remains a challenging problem.

The H-subunit of the restriction endonuclease CglI contains a prototype DEAD-Z1 helicase-like motor

CglI is a restriction endonuclease from Corynebacterium glutamicum that forms a complex between: two R-subunits that have site specific-recognition and nuclease domains; and two H-subunits, with Superfamily 2 helicase-like DEAD domains, and uncharacterized Z1 and C-terminal domains. ATP hydrolysis by the H-subunits catalyses dsDNA translocation that is necessary for long-range movement along DNA that activates nuclease activity. Here, we provide biochemical and molecular modelling evidence that shows that Z1 has a fold distantly-related to RecA, and that the DEAD-Z1 domains together form an ATP binding interface and are the prototype of a previously undescribed monomeric helicase-like motor. The DEAD-Z1 motor has unusual Walker A and Motif VI sequences those nonetheless have their expected functions. Additionally, it contains DEAD-Z1-specific features: an H/H motif and a loop (aa 163–aa 172), that both play a role in the coupling of ATP hydrolysis to DNA cleavage. We also solved the crystal structure of the C-terminal domain which has a unique fold, and demonstrate that the Z1-C domains are the principal DNA binding interface of the H-subunit. Finally, we use small angle X-ray scattering to provide a model for how the H-subunit domains are arranged in a dimeric complex.

The PPI3D web server for searching, analyzing and modeling protein-protein interactions in the context of 3D structures

Summary

The PPI3D web server is focused on searching and analyzing the structural data on protein-protein interactions. Reducing the data redundancy by clustering and analyzing the properties of interaction interfaces using Voronoi tessellation makes this software a highly effective tool for addressing different questions related to protein interactions.

Availability and Implementation

The server is freely accessible at http://bioinformatics.lt/software/ppi3d/ .

Contact

ceslovas.venclovas@bti.vu.lt.

Supplementary information

Supplementary data are available at Bioinformatics online.

A structure-function analysis of the yeast Elg1 protein reveals the importance of PCNA unloading in genome stability maintenance

The sliding clamp, PCNA, plays a central role in DNA replication and repair. In the moving replication fork, PCNA is present at the leading strand and at each of the Okazaki fragments that are formed on the lagging strand. PCNA enhances the processivity of the replicative polymerases and provides a landing platform for other proteins and enzymes. The loading of the clamp onto DNA is performed by the Replication Factor C (RFC) complex, whereas its unloading can be carried out by an RFC-like complex containing Elg1. Mutations in ELG1 lead to DNA damage sensitivity and genome instability. To characterize the role of Elg1 in maintaining genomic integrity, we used homology modeling to generate a number of site-specific mutations in ELG1 that exhibit different PCNA unloading capabilities. We show that the sensitivity to DNA damaging agents and hyper-recombination of these alleles correlate with their ability to unload PCNA from the chromatin. Our results indicate that retention of modified and unmodified PCNA on the chromatin causes genomic instability. We also show, using purified proteins, that the Elg1 complex inhibits DNA synthesis by unloading SUMOylated PCNA from the DNA. Additionally, we find that mutations in ELG1 suppress the sensitivity of rad5Δ mutants to DNA damage by allowing trans-lesion synthesis to take place. Taken together, the data indicate that the Elg1–RLC complex plays an important role in the maintenance of genomic stability by unloading PCNA from the chromatin.

The CAD-score web server: contact area-based comparison of structures and interfaces of proteins, nucleic acids and their complexes

The Contact Area Difference score (CAD-score) web server provides a universal framework to compute and analyze discrepancies between different 3D structures of the same biological macromolecule or complex. The server accepts both single-subunit and multi-subunit structures and can handle all the major types of macromolecules (proteins, RNA, DNA and their complexes). It can perform numerical comparison of both structures and interfaces. In addition to entire structures and interfaces, the server can assess user-defined subsets. The CAD-score server performs both global and local numerical evaluations of structural differences between structures or interfaces. The results can be explored interactively using sortable tables of global scores, profiles of local errors, superimposed contact maps and 3D structure visualization. The web server could be used for tasks such as comparison of models with the native (reference) structure, comparison of X-ray structures of the same macromolecule obtained in different states (e.g. with and without a bound ligand), analysis of nuclear magnetic resonance (NMR) structural ensemble or structures obtained in the course of molecular dynamics simulation. The web server is freely accessible at: http://www.ibt.lt/bioinformatics/cad-score.

Herpesviral helicase-primase subunit UL8 is inactivated B-family polymerase

MOTIVATION

Herpesviruses are large DNA viruses causing a variety of diseases in humans and animals. To develop effective treatment, it is important to understand the mechanisms of their replication. One of the components of the herpesviral DNA replication system is a helicase-primase complex, consisting of UL5 (helicase), UL52 (primase) and UL8. UL8 is an essential herpesviral protein involved in multiple protein-protein interactions. Intriguingly, so far no UL8 homologs outside of herpesviruses could be identified. Moreover, nothing is known about its structure or domain organization.

RESULTS

Here, combining sensitive homology detection methods and homology modeling, we found that the UL8 protein family is related to B-family polymerases. In the course of evolution, UL8 has lost the active site and has undergone a reduction of DNA-binding motifs. The loss of active site residues explains the failure to detect any catalytic activity of UL8. A structural model of human herpes virus 1 UL8 constructed as part of the study is consistent with the mutation data targeting its interaction with primase UL52. It also provides a platform for studying multiple interactions that UL8 is involved in. The two other components of helicase-primase complex show evolutionary links with a newly characterized human primase that also has DNA polymerase activity (PrimPol) and the Pif1 helicase, respectively. The role of these enzymes in recovering stalled replication forks suggests mechanistic and functional similarities with herpesviral proteins.

CONTACT

venclovas@ibt.lt

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

The use of interatomic contact areas to quantify discrepancies between RNA 3D models and reference structures

Growing interest in computational prediction of ribonucleic acid (RNA) three-dimensional structure has highlighted the need for reliable and meaningful methods to compare models and experimental structures. We present a structure superposition-free method to quantify both the local and global accuracy of RNA structural models with respect to the reference structure. The method, initially developed for proteins and here extended to RNA, closely reflects physical interactions, has a simple definition, a fixed range of values and no arbitrary parameters. It is based on the correspondence of respective contact areas between nucleotides or their components (base or backbone). The better is the agreement between respective contact areas in a model and the reference structure, the more accurate the model is considered to be. Since RNA bases account for the largest contact areas, we further distinguish stacking and non-stacking contacts. We have extensively tested the contact area-based evaluation method and found it effective in both revealing local discrepancies and ranking models by their overall quality. Compared to other reference-based RNA model evaluation methods, the new method shows a stronger emphasis on stereochemical quality of models. In addition, it takes into account model completeness, enabling a meaningful evaluation of full models and those missing some residues.

Voronota: A fast and reliable tool for computing the vertices of the Voronoi diagram of atomic balls

The Voronoi diagram of balls, corresponding to atoms of van der Waals radii, is particularly well-suited for the analysis of three-dimensional structures of biological macromolecules. However, due to the shortage of practical algorithms and the corresponding software, simpler approaches are often used instead. Here, we present a simple and robust algorithm for computing the vertices of the Voronoi diagram of balls. The vertices of Voronoi cells correspond to the centers of the empty tangent spheres defined by quadruples of balls. The algorithm is implemented as an open-source software tool, Voronota. Large-scale tests show that Voronota is a fast and reliable tool for processing both experimentally determined and computationally modeled macromolecular structures. Voronota can be easily deployed and may be used for the development of various other structure analysis tools that utilize the Voronoi diagram of balls. Voronota is available at: http://www.ibt.lt/bioinformatics/voronota.

A vitamin B₁₂ transporter in Mycobacterium tuberculosis

Vitamin B₁₂-dependent enzymes function in core biochemical pathways in Mycobacterium tuberculosis, an obligate pathogen whose metabolism in vivo is poorly understood. Although M. tuberculosis can access vitamin B₁₂in vitro, it is uncertain whether the organism is able to scavenge B₁₂ during host infection. This question is crucial to predictions of metabolic function, but its resolution is complicated by the absence in the M. tuberculosis genome of a direct homologue of BtuFCD, the only bacterial B₁₂ transport system described to date. We applied genome-wide transposon mutagenesis to identify M. tuberculosis mutants defective in their ability to use exogenous B₁₂. A small proportion of these mapped to Rv1314c, identifying the putative PduO-type ATP : co(I)rrinoid adenosyltransferase as essential for B₁₂ assimilation. Most notably, however, insertions in Rv1819c dominated the mutant pool, revealing an unexpected function in B₁₂ acquisition for an ATP-binding cassette (ABC)-type protein previously investigated as the mycobacterial BacA homologue. Moreover, targeted deletion of Rv1819c eliminated the ability of M. tuberculosis to transport B₁₂ and related corrinoids in vitro. Our results establish an alternative to the canonical BtuCD-type system for B₁₂ uptake in M. tuberculosis, and elucidate a role in B₁₂ metabolism for an ABC protein implicated in chronic mycobacterial infection.

Two distinct SSB protein families in nucleo-cytoplasmic large DNA viruses

MOTIVATION

Eukaryote-infecting nucleo-cytoplasmic large DNA viruses (NCLDVs) feature some of the largest genomes in the viral world. These viruses typically do not strongly depend on the host DNA replication systems. In line with this observation, a number of essential DNA replication proteins, such as DNA polymerases, primases, helicases and ligases, have been identified in the NCLDVs. One other ubiquitous component of DNA replisomes is the single-stranded DNA-binding (SSB) protein. Intriguingly, no NCLDV homologs of canonical OB-fold-containing SSB proteins had previously been detected. Only in poxviruses, one of seven NCLDV families, I3 was identified as the SSB protein. However, whether I3 is related to any known protein structure has not yet been established.

RESULTS

Here, we addressed the case of 'missing' canonical SSB proteins in the NCLDVs and also probed evolutionary origins of the I3 family. Using advanced computational methods, in four NCLDV families, we detected homologs of the bacteriophage T7 SSB protein (gp2.5). We found the properties of these homologs to be consistent with the SSB function. Moreover, we implicated specific residues in single-stranded DNA binding. At the same time, we found no evolutionary link between the T7 gp2.5-like NCLDV SSB homologs and the poxviral SSB protein (I3). Instead, we identified a distant relationship between I3 and small protein B (SmpB), a bacterial RNA-binding protein. Thus, apparently, the NCLDVs have the two major distinct sets of SSB proteins having bacteriophage and bacterial origins, respectively.

CAD-score: a new contact area difference-based function for evaluation of protein structural models

Evaluation of protein models against the native structure is essential for the development and benchmarking of protein structure prediction methods. Although a number of evaluation scores have been proposed to date, many aspects of model assessment still lack desired robustness. In this study we present CAD-score, a new evaluation function quantifying differences between physical contacts in a model and the reference structure. The new score uses the concept of residue-residue contact area difference (CAD) introduced by Abagyan and Totrov (J Mol Biol 1997; 268:678-685). Contact areas, the underlying basis of the score, are derived using the Voronoi tessellation of protein structure. The newly introduced CAD-score is a continuous function, confined within fixed limits, free of any arbitrary thresholds or parameters. The built-in logic for treatment of missing residues allows consistent ranking of models of any degree of completeness. We tested CAD-score on a large set of diverse models and compared it to GDT-TS, a widely accepted measure of model accuracy. Similarly to GDT-TS, CAD-score showed a robust performance on single-domain proteins, but displayed a stronger preference for physically more realistic models. Unlike GDT-TS, the new score revealed a balanced assessment of domain rearrangement, removing the necessity for different treatment of single-domain, multi-domain, and multi-subunit structures. Moreover, CAD-score makes it possible to assess the accuracy of inter-domain or inter-subunit interfaces directly. In addition, the approach offers an alternative to the superposition-based model clustering. The CAD-score implementation is available both as a web server and a standalone software package at http://www.ibt.lt/bioinformatics/cad-score/.

Computational analysis of DNA replicases in double-stranded DNA viruses: relationship with the genome size

Genome duplication in free-living cellular organisms is performed by DNA replicases that always include a DNA polymerase, a DNA sliding clamp and a clamp loader. What are the evolutionary solutions for DNA replicases associated with smaller genomes? Are there some general principles? To address these questions we analyzed DNA replicases of double-stranded (ds) DNA viruses. In the process we discovered highly divergent B-family DNA polymerases in phiKZ-like phages and remote sliding clamp homologs in Ascoviridae family and Ma-LMM01 phage. The analysis revealed a clear dependency between DNA replicase components and the viral genome size. As the genome size increases, viruses universally encode their own DNA polymerases and frequently have homologs of DNA sliding clamps, which sometimes are accompanied by clamp loader subunits. This pattern is highly non-random. The absence of sliding clamps in large viral genomes usually coincides with the presence of atypical polymerases. Meanwhile, sliding clamp homologs, not accompanied by clamp loaders, have an elevated positive electrostatic potential, characteristic of non-ring viral processivity factors that bind the DNA directly. Unexpectedly, we found that similar electrostatic properties are shared by the eukaryotic 9-1-1 clamp subunits, Hus1 and, to a lesser extent, Rad9, also suggesting the possibility of direct DNA binding.

Methods for sequence-structure alignment

Homology modeling is based on the observation that related protein sequences adopt similar three-dimensional structures. Hence, a homology model of a protein can be derived using related protein structure(s) as modeling template(s). A key step in this approach is the establishment of correspondence between residues of the protein to be modeled and those of modeling template(s). This step, often referred to as sequence-structure alignment, is one of the major determinants of the accuracy of a homology model. This chapter gives an overview of methods for deriving sequence-structure alignments and discusses recent methodological developments leading to improved performance. However, no method is perfect. How to find alignment regions that may have errors and how to make improvements? This is another focus of this chapter. Finally, the chapter provides a practical guidance of how to get the most of the available tools in maximizing the accuracy of sequence-structure alignments.

Identification of new homologs of PD-(D/E)XK nucleases by support vector machines trained on data derived from profile-profile alignments

PD-(D/E)XK nucleases, initially represented by only Type II restriction enzymes, now comprise a large and extremely diverse superfamily of proteins. They participate in many different nucleic acids transactions including DNA degradation, recombination, repair and RNA processing. Different PD-(D/E)XK families, although sharing a structurally conserved core, typically display little or no detectable sequence similarity except for the active site motifs. This makes the identification of new superfamily members using standard homology search techniques challenging. To tackle this problem, we developed a method for the detection of PD-(D/E)XK families based on the binary classification of profile–profile alignments using support vector machines (SVMs). Using a number of both superfamily-specific and general features, SVMs were trained to identify true positive alignments of PD-(D/E)XK representatives. With this method we identified several PFAM families of uncharacterized proteins as putative new members of the PD-(D/E)XK superfamily. In addition, we assigned several unclassified restriction enzymes to the PD-(D/E)XK type. Results show that the new method is able to make confident assignments even for alignments that have statistically insignificant scores. We also implemented the method as a freely accessible web server at http://www.ibt.lt/bioinformatics/software/pdexk/.

Thermodynamics of radicicol binding to human Hsp90 alpha and beta isoforms

Radicicol is a natural antibiotic that specifically inhibits chaperone Hsp90 activity and binds to its active site with nanomolar affinity. Radicicol has been widely used as a lead compound to generate synthetic analogs with reduced toxicity and increased stability that could be employed clinically. Here we present a detailed thermodynamic description of radicicol binding to human Hsp90 and yeast Hsc82 studied by isothermal titration calorimetry and thermal shift assay. Titrations as a function of pH showed a linked protonation event upon radicicol binding. The intrinsic binding constant and the thermodynamic parameters (including the enthalpy, entropy, and heat capacity) were determined for yeast Hsc82, and human alpha and beta Hsp90. Recent experimental evidence in literature shows that yeast Hsc82 has significant differences from human Hsp90 isozymes. Here we support this by demonstrating differences in radicicol binding thermodynamics to these proteins. The intrinsic enthalpy of radicicol binding to Hsc82 was -46.7 kJ/mol, to Hsp90alpha -70.7 kJ/mol, and to Hsp90beta was -66.8 kJ/mol. The enthalpies of binding were significantly different, while the intrinsic dissociation constants were quite similar, equal to 0.25, 0.04, and 0.15 nM, respectively. The structural features responsible for such large difference in binding enthalpy but small difference in the intrinsic binding Gibbs free energy are discussed.

COMA server for protein distant homology search

SUMMARY

Detection of distant homology is a widely used computational approach for studying protein evolution, structure and function. Here, we report a homology search web server based on sequence profile-profile comparison. The user may perform searches in one of several regularly updated profile databases using either a single sequence or a multiple sequence alignment as an input. The same profile databases can also be downloaded for local use. The capabilities of the server are illustrated with the identification of new members of the highly diverse PD-(D/E)XK nuclease superfamily.

AVAILABILITY

http://www.ibt.lt/bioinformatics/coma/

Detection of distant evolutionary relationships between protein families using theory of sequence profile-profile comparison

Background

Detection of common evolutionary origin (homology) is a primary means of inferring protein structure and function. At present, comparison of protein families represented as sequence profiles is arguably the most effective homology detection strategy. However, finding the best way to represent evolutionary information of a protein sequence family in the profile, to compare profiles and to estimate the biological significance of such comparisons, remains an active area of research.

Results

Here, we present a new homology detection method based on sequence profile-profile comparison. The method has a number of new features including position-dependent gap penalties and a global score system. Position-dependent gap penalties provide a more biologically relevant way to represent and align protein families as sequence profiles. The global score system enables an analytical solution of the statistical parameters needed to estimate the statistical significance of profile-profile similarities. The new method, together with other state-of-the-art profile-based methods (HHsearch, COMPASS and PSI-BLAST), is benchmarked in all-against-all comparison of a challenging set of SCOP domains that share at most 20% sequence identity. For benchmarking, we use a reference ("gold standard") free model-based evaluation framework. Evaluation results show that at the level of protein domains our method compares favorably to all other tested methods. We also provide examples of the new method outperforming structure-based similarity detection and alignment. The implementation of the new method both as a standalone software package and as a web server is available at http://www.ibt.lt/bioinformatics/coma.

Conclusion

Due to a number of developments, the new profile-profile comparison method shows an improved ability to match distantly related protein domains. Therefore, the method should be useful for annotation and homology modeling of uncharacterized proteins.

The use of automatic tools and human expertise in template-based modeling of CASP8 target proteins

Here, we describe our template-based protein modeling approach and its performance during the eighth community-wide experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP8, http://predictioncenter.org/casp8). In CASP8, our modeling approach was supplemented by the newly developed distant homology detection method based on sequence profile-profile comparison. Detection of structural homologs that could be used as modeling templates was largely achieved by automated profile-based searches. However, the other two major steps in template-based modeling (TBM) (selection of the best template(s) and construction of the optimal sequence-structure alignment) to a large degree relied on the combination of automatic tools and manual input. The analysis of 64 domains categorized by CASP8 assessors as TBM domains revealed that we missed correct structural templates for only four of them. The use of multiple templates or their fragments enabled us to improve over the structure of the single best PDB template in about 1/3 of our models for TBM domains. Our results for sequence-structure alignments are mixed. Although many models have optimal or near optimal sequence mapping, a large fraction contains one or more misaligned regions. Strikingly, in spite of this, our TBM models have the best overall alignment accuracy scores. This clearly suggests that the correct mapping of protein sequence onto three-dimensional structure remains one of the big challenges in protein structure prediction.

Re-searcher: a system for recurrent detection of homologous protein sequences

Background

Sequence searches are routinely employed to detect and annotate related proteins. However, a rapid growth of databases necessitates a frequent repetition of sequence searches and subsequent analysis of obtained results. Although there are several automatic systems available for executing periodical sequence searches and reporting results, they all suffer either from a lack of sensitivity, restrictive database choice or limited flexibility in setting up search strategies. Here, a new sequence search and reporting software package designed to address these shortcomings is described.

Results

Re-searcher is an open-source highly configurable system for recurrent detection and reporting of new homologs for the sequence of interest in specified protein sequence databases. Searches are performed using PSI-BLAST at desired time intervals either within NCBI or local databases. In addition to searches against individual databases, the system can perform "PDB-BLAST"-like combined searches, when PSI-BLAST profile generated during search against the first database is used to search the second database. The system supports multiple users enabling each to separately keep track of multiple queries and query-specific results.

Conclusions

Re-searcher features a large number of options enabling automatic periodic detection of both close and distant homologs. At the same time it has a simple and intuitive interface, making the analysis of results even for a large number of queries a straightforward task.

Comparative modeling in CASP6 using consensus approach to template selection, sequence-structure alignment, and structure assessment

Along with over 150 other groups we have tested our template-based protein structure prediction approach by submitting models for 30 target proteins to the sixth round of the Critical Assessment of Protein Structure Prediction Methods (CASP6, http://predictioncenter.org). Most of our modeled proteins fall into the comparative or homology modeling (CM) category, and some are fold recognition (FR) targets. The key feature of our structure prediction strategy in CASP6 was an attempt to optimally select structural templates and to make accurate sequence-structure alignments. Template selection was based mainly on consensus results of multiple sequence searches. Likewise, the consensus of multiple alignment variants (or lack of it) was used to initially delineate reliable and unreliable alignment regions. Structure evaluation approaches were then used to identify the correct sequence-structure mapping. Our results suggest that in many cases use of multiple templates is advantageous. Selecting correct alignments even within the context of a three-dimensional structure remains a challenge. Together with more effective energy evaluation methods the simultaneous relaxation/refinement of a "frozen" backbone inherited from the template is likely needed to see a clear progress in tackling this problem. Our analysis also suggests that human input has little to contribute to automatic methods in modeling high homology targets. On the other hand, human expertise can be very valuable in modeling distantly related proteins and critical in cases of unexpected evolutionary changes in protein structure.

Progress over the first decade of CASP experiments

CASP has now completed a decade of monitoring the state of the art in protein structure prediction. The quality of structure models produced in the latest experiment, CASP6, has been compared with that in earlier CASPs. Significant although modest progress has again been made in the fold recognition regime, and cumulatively, progress in this area is impressive. Models of previously unknown folds again appear to have modestly improved, and several mixed alpha/beta structures have been modeled in a topologically correct manner. Progress remains hard to detect in high sequence identity comparative modeling, but server performance in this area has moved forward.

DNA Sliding Clamps: Just the Right Twist to Load onto DNA

Two recent papers illuminate a key step in DNA sliding clamp loading: one reveals the structure of the PCNA clamp wrapped around DNA--still open from being loaded--while the other finds that the clamp may assist this process by forming a right-handed helix upon opening.

HhaI DNA methyltransferase uses the protruding Gln237 for active flipping of its target cytosine

Access to a nucleotide by its rotation out of the DNA helix (base flipping) is used by numerous DNA modification and repair enzymes. Despite extensive studies of the paradigm HhaI methyltransferase, initial events leading to base flipping remained elusive. Here we demonstrate that the replacement of the target C:G pair with the 2-aminopurine:T pair in the DNA or shortening of the side chain of Gln237 in the protein severely perturb base flipping, but retain specific DNA binding. Kinetic analyses and molecular modeling suggest that a steric interaction between the protruding side chain of Gln237 and the target cytosine in B-DNA reduces the energy barrier for flipping by 3 kcal/mol. Subsequent stabilization of an open state by further 4 kcal/mol is achieved through specific hydrogen bonding of the side chain to the orphan guanine. Gln237 thus plays a key role in actively opening the target C:G pair by a "push-and-bind" mechanism.

Sequence-structure mapping errors in the PDB: OB-fold domains

The Protein Data Bank (PDB) is the single most important repository of structural data for proteins and other biologically relevant molecules. Therefore, it is critically important to keep the PDB data, as much as possible, error-free. In this study, we have analyzed PDB crystal structures possessing oligonucleotide/oligosaccharide binding (OB)-fold, one of the highly populated folds, for the presence of sequence-structure mapping errors. Using energy-based structure quality assessment coupled with sequence analyses, we have found that there are at least five OB-structures in the PDB that have regions where sequences have been incorrectly mapped onto the structure. We have demonstrated that the combination of these computation techniques is effective not only in detecting sequence-structure mapping errors, but also in providing guidance to correct them. Namely, we have used results of computational analysis to direct a revision of X-ray data for one of the PDB entries containing a fairly inconspicuous sequence-structure mapping error. The revised structure has been deposited with the PDB. We suggest use of computational energy assessment and sequence analysis techniques to facilitate structure determination when homologs having known structure are available to use as a reference. Such computational analysis may be useful in either guiding the sequence-structure assignment process or verifying the sequence mapping within poorly defined regions.

Assessment of progress over the CASP experiments

The quality of structure models produced in the CASP5 experiment has been compared with that in earlier CASPs. The most significant progress is in the fold recognition regime, where the development of meta-servers has allowed more accurate consensus models to be generated. In contrast to this, there is little evidence of progress in producing more accurate comparative models, particularly those based on sequence identities > 30%. For comparative models based on low-sequence identity and for fold recognition models, accuracy depends primarily on the fraction of the target structure that is similar to an available template, and the quality of the alignment. Overall, these results indicate that there are still no effective methods of improving model quality beyond that obtained by successfully copying a template structure. For models of proteins with previously unknown folds, there appears to be a pause in the previous consistent improvement. There is some evidence that more groups are producing top-quality models, however. Although specific progress between successive experiments is sometimes difficulty to identify, over the history of all the CASPs there has been steady, if sometimes slow, progress in all modeling regimes.

Comparative modeling in CASP5: Progress is evident, but alignment errors remain a significant hindrance

Models for 20 comparative modeling targets were submitted for the fifth round of the "blind" test of protein structure prediction methods (CASP5; http://predictioncenter.llnl.gov/casp5). The modeling approach used in CASP5 was similar to that used 2 years ago in CASP4 (Venclovas, Proteins 2001; Suppl 5:47-54). The main features of this approach include use of multiple templates, initial assessment of alignment reliability in a region-specific manner, and structure-based selection of alignment variants in unreliable regions. The CASP5 modeling results presented here show significant improvement in comparison to CASP4, especially in the area of distant homology. The improvements include more effective use of multiple templates and better alignments. However, a number of structurally conserved regions in submitted distant homology models were misaligned. Analysis of these errors indicates that the absolute majority of them occurred in regions deemed unreliable in the course of model building. Most of these error-prone regions can be characterized by their peripheral location and a lack of conserved sequence patterns. For a few of the error-prone regions, all methods evaluated during CASP5 proved ineffective, pointing to the need for more sensitive energy-based methods. Despite these remaining issues, the applicability of comparative modeling continues to expand into more distant evolutionary relationships, providing a means to structurally characterize a significant number of currently available protein sequences.

Selection and characterization of anti-MUC-1 scFvs intended for targeted therapy

PURPOSE

The selection and characterization of anti-MUC-1 single-chain antibody fragments (scFv) is a first step toward the construction of new anticancer molecules designed for optimal blood clearance and tumor penetration. The mucin MUC-1 was chosen as an antigen because it is abundantly expressed on epithelial cancers in an aberrantly glycosylated form, making it structurally and antigenically distinct from MUC-1 expressed on normal cells.

EXPERIMENTAL DESIGN

A previously constructed anti-MUC-1 phage display library from hyperimmunized mice, with 5 x 10(5) calculated variants, was screened for the selection of anti-MUC-1 scFvs. Selection criteria were high binding to a MUC-1 peptide containing 4 tandem repeats of 20 amino acids and to MUC-1-positive MCF-7 (human breast cancer) cell lysates in ELISA.

RESULTS

Six anti-MUC-1 scFv clones were selected and characterized. Nucleotide sequencing showed that four of them were full length scFv genes (variable heavy chain + variable light chain), whereas the remaining two contained either a variable heavy chain or a variable light chain alone. Their binding affinities (K(a)) range between 8 x 10(7) and 10(9) M(-1). Immunohistopathology demonstrated reactivity with breast cancer cells (MCF-7 and BT20) and human breast biopsy tissue. Molecular modeling revealed high structural similarity of the anti-MUC-1 scFvs with the X-ray-determined structure of the anti-CEA scFv (MFE-23).

CONCLUSIONS

In vitro antigen binding was demonstrated for the selected anti-MUC-1 scFvs. The binding affinities of these scFvs are in a promising range for efficient in vivo antigen binding. These anti-MUC-1 scFvs will be evaluated as antigen-binding modules in new multifunctional agents for the detection and therapy of cancer.

Molecular modeling-based analysis of interactions in the RFC-dependent clamp-loading process

Replication and related processes in eukaryotic cells require replication factor C (RFC) to load a molecular clamp for DNA polymerase in an ATP-driven process, involving multiple molecular interactions. The detailed understanding of this mechanism is hindered by the lack of data regarding structure, mutual arrangement, and dynamics of the players involved. In this study, we analyzed interactions that take place during loading onto DNA of either the PCNA clamp or the Rad9-Rad1-Hus1 checkpoint complex, using computationally derived molecular models. Combining the modeled structures for each RFC subunit with known structural, biochemical, and genetic data, we propose detailed models of how two of the RFC subunits, RFC1 and RFC3, interact with the C-terminal regions of PCNA. RFC1 is predicted to bind PCNA similarly to the p21-PCNA interaction, while the RFC3-PCNA binding is proposed to be similar to the E. coli delta-beta interaction. Additional sequence and structure analysis, supported by experimental data, suggests that RFC5 might be the third clamp loader subunit to bind the equivalent PCNA region. We discuss functional implications stemming from the proposed model of the RFC1-PCNA interaction and compare putative clamp-interacting regions in RFC1 and its paralogs, Rad17 and Ctf18. Based on the individual intermolecular interactions, we propose RFC and PCNA arrangement that places three RFC subunits in association with each of the three C-terminal regions in PCNA. The two other RFC subunits are positioned at the two PCNA interfaces, with the third PCNA interface left unobstructed. In addition, we map interactions at the level of individual subunits between the alternative clamp loader/clamp system, Rad17-RFC(2-5)/Rad9-Rad1-Hus1. The proposed models of interaction between two clamp/clamp loader pairs provide both structural framework for interpretation of existing experimental data and a number of specific findings that can be subjected to direct experimental testing.

Comparative modeling of CASP4 target proteins: combining results of sequence search with three-dimensional structure assessment

Comparative modeling aims at constructing molecular models for proteins of unknown structure, by using known structures of related proteins as templates. To test the comparative modeling approach reported here, predictions for 13 target proteins were submitted during the fourth round of "blind" protein structure prediction experiment (CASP4; http://PredictionCenter.llnl.gov/casp4). Sequence identity between these target proteins and the closest known structures ranged from 13 to 58%, indicating a broad spectrum of prediction difficulty. Although this broad difficulty range required addressing a variety of issues, the most important proved to be sequence-structure alignment for distant homology targets. The alignment step was based on structure-based evaluation of alignment variants produced mainly with PSI-BLAST intermediate sequence search procedure (PSI-BLAST-ISS). Although a fraction of correctly aligned residues in resulting models was markedly better than the average in all cases, for distant homology targets it was still considerably below the estimated achievable level. Results with CASP4 targets show that, along with the correctness of sequence-structure alignments, effective use of multiple template structures may significantly increase accuracy of the model structure. Improvement in this area should also result in more accurate loop modeling and side-chain prediction.

Comparison of performance in successive CASP experiments

As the number of completed CASP (Critical Assessment of Protein Structure Prediction) experiments grows, so does the need for stable, standard methods for comparing performance in successive experiments. It is critical to develop methods for determining the areas in which there is progress and in which areas are static. We have added an analysis of the CASP4 results to that previously published for CASPs 1, 2, and 3. We again use a unified difficulty scale to permit comparison of performance as a function of target difficulty in the different CASPs. The scale is used to compare performance in aligning target sequences to a structural template. There was a clear improvement in alignment quality between CASP1 (1994) and CASP2 (1996). No change is apparent between CASP2 and CASP3 (1998). There is a small barely detectable improvement between CASP3 and the latest experiment (CASP4, 2000). Alignment remains the major source of error in all models based on less than about 30% sequence identity. Comparison of performance in the new fold modeling regime is complicated by issues in devising an objective target difficulty scale. We have found limited numerical support for significant progress between CASP3 and CASP4 in this area. More subjectively, most observers are convinced that there has been substantial progress. Progress is dominated by a single group.

Structure-based sequence alignment for the beta-trefoil subdomain of the clostridial neurotoxin family provides residue level information about the putative ganglioside binding site

Clostridial neurotoxins embrace a family of extremely potent toxins comprised of tetanus toxin (TeNT) and seven different serotypes of botulinum toxin (BoNT/A-G). The beta-trefoil subdomain of the C-terminal part of the heavy chain (H(C)), responsible for ganglioside binding, is the most divergent region in clostridial neurotoxins with sequence identity as low as 15%. We re-examined the alignment between family sequences within this subdomain, since in this region all alignments published to date show obvious inconsistencies with the beta-trefoil fold. The final alignment was obtained by considering the general constraints imposed by this fold, and homology modeling studies based on the TeNT structure. Recently solved structures of BoNT/A confirm the validity of this structure-based approach. Taking into account biochemical data and crystal structures of TeNT and BoNT/A, we also re-examined the location of the putative ganglioside binding site and, using the new alignment, characterized this site in other BoNT serotypes.

Structure-based predictions of Rad1, Rad9, Hus1 and Rad17 participation in sliding clamp and clamp-loading complexes

The repair of damaged DNA is coupled to the completion of DNA replication by several cell cycle checkpoint proteins, including, for example, in fission yeast Rad1(Sp), Hus1(Sp), Rad9(Sp) and Rad17(Sp). We have found that these four proteins are conserved with protein sequences throughout eukaryotic evolution. Using computational techniques, including fold recognition, comparative modeling and generalized sequence profiles, we have made high confidence structure predictions for the each of the Rad1, Hus1 and Rad9 protein families (Rad17(Sc), Mec3(Sc) and Ddc1(Sc) in budding yeast, respectively). Each of these families was found to share a common protein fold with that of PCNA, the sliding clamp protein that tethers DNA polymerase to its template. We used previously reported genetic and biochemical data for these proteins from yeast and human cells to predict a heterotrimeric PCNA-like ring structure for the functional Rad1/Rad9/Hus1 complex and to determine their exact order within it. In addition, for each individual protein family, contact regions with neighbors within the PCNA-like ring were identified. Based on a molecular model for Rad17(Sp), we concluded that members of this family, similar to the subunits of the RFC clamp-loading complex, are capable of coupling ATP binding with conformational changes required to load a sliding clamp onto DNA. This model substantiates previous findings regarding the behavior of Rad17 family proteins upon DNA damage and within the RFC complex of clamp-loading proteins.

Addressing the issue of sequence-to-structure alignments in comparative modeling of CASP3 target proteins

During a blind protein structure prediction experiment (the third round of the Critical Assessment of Techniques for Protein Structure Prediction; URL http://PredictionCenter.llnl.gov/casp3/) , four target proteins, T0047, T0048, T0055, and T0070, were modeled by comparison. These proteins display 62%, 29%, 24%, and 19% sequence identity, respectively, to the structurally homologous proteins most similar in sequence. The issue of sequence-to-structure alignment in cases of low sequence homology was the main emphasis. Selection of alignments was made by constructing and evaluating three-dimensional models based on series of samples produced mainly by automatic multiple sequence alignments. Sequence-to-structure alignments were correct in all but two regions, in which significant changes in target structures compared with related proteins were the source of errors. Template choice is an important determinant of model quality, and a correct selection was made of a lower homology template for modeling of T0070; however, in the case of T0055, a template with 8% greater sequence homology proved deceptive. Loops and some ungapped template regions were assigned conformations taken from other proteins. Using fragments from homologous structures led to improvement over template backbone more often than cases in which nonhomologous structures were the source. The results also indicate that side-chain prediction accuracy depends not only on sequence similarity but also on accuracy of the backbone.

Some measures of comparative performance in the three CASPs

Performance in the three Critical Assessment of protein Structure Prediction (CASP) experiments has been compared in the areas of alignment accuracy for models based on homology and three-dimensional accuracy for models produced by using ab initio prediction methods. The homologous models span the comparative modeling and fold-recognition regimes. Each CASP target is assigned a relative difficulty based on the extent of sequence identity and the degree of structural overlap with the best available template. There is a clear improvement in alignment accuracy between CASP1 and CASPs 2 and 3 over much of the difficulty scale but no apparent improvement between CASP2 and CASP3. Encouragingly, the best ab initio models of small targets are clearly more accurate in CASP3 than in CASPs 1 and 2.

Processing and analysis of CASP3 protein structure predictions

Livermore Prediction Center provides basic infrastructure for the CASP (Critical Assessment of Structure Prediction) experiments, including prediction processing and verification servers, a system of prediction evaluation tools, and interactive numerical and graphical displays. Here we outline the essentials of our approach, with discussion of the superposition procedures, definitions of basic measures, and descriptions of new methods developed to analyze predictions. Our primary focus is on the evaluation of three-dimensional models and secondary structure predictions. To put the results of the three prediction experiments held to date on the same footing, the latest CASP3 evaluation criteria were retrospectively applied to both CASP1 and CASP2 predictions. Finally, we give an overview of our website (http:/(/)PredictionCenter.llnl.gov), which makes the target structures, predictions, and the evaluation system accessible to the community.

A modified definition of Sov, a segment-based measure for protein secondary structure prediction assessment

We present a measure for the evaluation of secondary structure prediction methods that is based on secondary structure segments rather than individual residues. The algorithm is an extension of the segment overlap measure Sov, originally defined by Rost et al. (J Mol Biol 1994;235:13-26). The new definition of Sov corrects the normalization procedure and improves Sov's ability to discriminate between similar and dissimilar segment distributions. The method has been comprehensively tested during the second Critical Assessment of Techniques for Protein Structure Prediction (CASP2). Here, we describe the underlying concepts, modifications to the original definition, and their significance.

Criteria for evaluating protein structures derived from comparative modeling

Following the first experiment for the Critical Assessment of methods for protein Structure Prediction (CASP1), numerical criteria were devised to analyze the performance of prediction methods. We report here the criteria for comparative modeling, and how effective they were in CASP2. These criteria are intended to evolve into a set of numerical measures that provide a comprehensive assessment of the quality of a structure produced by comparative modeling, and provide a means of investigating which modeling methods are most effective, so as to establish where future effort may be most productively applied.

Numerical criteria for the evaluation of ab initio predictions of protein structure

As part of the CASP2 protein structure prediction experiment, a set of numerical criteria were defined for the evaluation of "ab initio" predictions. The evaluation package comprises a series of electronic submission formats, a submission validator, evaluation software, and a series of scripts to summarize the results for the CASP2 meeting and for presentation via the World Wide Web (WWW). The evaluation package is accessible for use on new predictions via WWW so that results can be compared to those submitted to CASP2. With further input from the community, the evaluation criteria are expected to evolve into a comprehensive set of measures capturing the overall quality of a prediction as well as critical detail essential for further development of prediction methods. We discuss present measures, limitations of the current criteria, and possible improvements.

Different enzymes with similar structures involved in Mg(2+)-mediated polynucleotidyl transfer

Comparison of X-ray structures of restriction endonucleases and polynucleotidyl transferase superfamily enzymes reveals a structural resemblance.

Five-stranded beta-sheet sandwiched with two alpha-helices: a structural link between restriction endonucleases EcoRI and EcoRV

Examination of crystal structures of restriction endonucleases EcoRI and EcoRV complexes with their cognate DNA revealed a common structural element, which forms the core of both proteins. This element consists of a five-stranded beta-sheet and two alpha-helices packed against it and could be described as alpha-beta sandwich in which helices and beta-strands lie in two stacked layers. While the spatial structure of this alpha-beta sandwich is conserved in both enzymes, there are not detectable similarities between amino acid sequences except of a few residues involved in active site formation. Probably, other restriction endonucleases which have similar organization of the active site might possess similar structural element regardless of DNA sequence recognized and recognition elements in the enzyme used.

Codon-anticodon pairing. A model for interacting codon-anticodon duplexes located at the ribosomal A- and P-sites

The interaction between two codon-anticodon duplexes of the ribosomal A- and P-site-bound tRNAs is the key feature of the proposed model. This interaction prohibits non-canonical base pairing at the first and second positions of the codon and controls base pairing at the third position (wobbling rules ensuing from the model are in good accord with those generated from experiments). The model is capable of predicting codon context effects. It follows from the model that modifications of the first anticodon residue of the P-site tRNA can affect the stability of the A-site duplex, and that the translation of a DNA single chain analogue of mRNA should be accompanied by non-canonical base pairing at all three positions of the codon. These predictions of the model can be subjected to experimental tests.

How are tRNAs and mRNA arranged in the ribosome? An attempt to correlate the stereochemistry of the tRNA-mRNA interaction with constraints imposed by the ribosomal topography

Two tRNA molecules at the ribosomal A- and P-sites, with a relatively small angle between the planes of the L-shaped molecules, can be arranged in two mutually exclusive orientations. In one (the 'R'-configuration), the T-loop of the A-site tRNA faces the D-loop of the P-site tRNA, whereas in the other (the 'S'-configuration) the D-loop of the A-site tRNA faces the T-loop of the P-site tRNA. A number of stereochemical arguments, based on the crystal structure of 'free' tRNA, favour the R-configuration. In the ribosome, the CCA-ends of the tRNA molecules are 'fixed' at the base of the central protuberance (the peptidyl transferase centre) of the 50S subunit, and the anticodon loops lie in the neck region (the decoding site) of the 30S subunit. The translocation step is essentially a rotational movement of the tRNA from the A- to the P-site, and there is convincing evidence that the A-site must be located nearest to the L7/L12 protuberance of the 50S subunit. The mRNA in the two codon-anticodon duplexes lies on the 'inside' of the 'elbows' of the tRNA molecules (in both the S-type and R-type configurations), and runs up between the two molecules from the A- to the P-site in the 3' to 5'-direction. These considerations have the consequence that in the S-configuration the mRNA in the codon-anticodon duplexes is directed towards the 50S subunit, whereas in the R-configuration it is directed towards the 30S subunit. The results of site-directed cross-linking experiments, in particular cross-links to mRNA at positions within or very close to the codons interacting with A- or P-site tRNA, favour the latter situation. This conclusion is in direct contradiction to other current models for the arrangement of mRNA and tRNA on the ribosome.

The path of a protein chain can be approximated by the conformation dictated by interpeptide ionic bridges

A stereochemical simulation of the formation of ionic bridges between adjacent peptide groups along the polypeptide chain has been made. Such ionic bridges constrain the amino-acid residues into eight conformations. It is shown that the path of any protein-chain fragment 10-15 residues long can be approximated well by these conformations. This suggests that the conformations dictated by the ionic bridges can be used as blocks in the formation of the spatial protein structure.