CASP9 target classification

Petabase-Scale Homology Search for Structure Prediction

The recent CASP15 competition highlighted the critical role of multiple sequence alignments (MSAs) in protein structure prediction, as demonstrated by the success of the top AlphaFold2-based prediction methods. To push the boundaries of MSA utilization, we conducted a petabase-scale search of the Sequence Read Archive (SRA), resulting in gigabytes of aligned homologs for CASP15 targets. These were merged with default MSAs produced by ColabFold-search and provided to ColabFold-predict. By using SRA data, we achieved highly accurate predictions (GDT_TS > 70) for 66% of the non-easy targets, whereas using ColabFold-search default MSAs scored highly in only 52%. Next, we tested the effect of deep homology search and ColabFold's advanced features, such as more recycles, on prediction accuracy. While SRA homologs were most significant for improving ColabFold's CASP15 ranking from 11th to 3rd place, other strategies contributed too. We analyze these in the context of existing strategies to improve prediction.

To split or not to split: CASP15 targets and their processing into tertiary structure evaluation units

Processing of CASP15 targets into evaluation units (EUs) and assigning them to evolutionary-based prediction classes is presented in this study. The targets were first split into structural domains based on compactness and similarity to other proteins. Models were then evaluated against these domains and their combinations. The domains were joined into larger EUs if predictors' performance on the combined units was similar to that on individual domains. Alternatively, if most predictors performed better on the individual domains, then they were retained as EUs. As a result, 112 evaluation units were created from 77 tertiary structure prediction targets. The EUs were assigned to four prediction classes roughly corresponding to target difficulty categories in previous CASPs: TBM (template-based modeling, easy or hard), FM (free modeling), and the TBM/FM overlap category. More than a third of CASP15 EUs were attributed to the historically most challenging FM class, where homology or structural analogy to proteins of known fold cannot be detected.

The transformative power of transformers in protein structure prediction

Transformer neural networks have revolutionized structural biology with the ability to predict protein structures at unprecedented high accuracy. Here, we report the predictive modeling performance of the state-of-the-art protein structure prediction methods built on transformers for 69 protein targets from the recently concluded 15th Critical Assessment of Structure Prediction (CASP15) challenge. Our study shows the power of transformers in protein structure modeling and highlights future areas of improvement.

Petascale Homology Search for Structure Prediction

The recent CASP15 competition highlighted the critical role of multiple sequence alignments (MSAs) in protein structure prediction, as demonstrated by the success of the top AlphaFold2-based prediction methods. To push the boundaries of MSA utilization, we conducted a petabase-scale search of the Sequence Read Archive (SRA), resulting in gigabytes of aligned homologs for CASP15 targets. These were merged with default MSAs produced by ColabFold-search and provided to ColabFold-predict. By using SRA data, we achieved highly accurate predictions (GDT_TS > 70) for 66% of the non-easy targets, whereas using ColabFold-search default MSAs scored highly in only 52%. Next, we tested the effect of deep homology search and ColabFold's advanced features, such as more recycles, on prediction accuracy. While SRA homologs were most significant for improving ColabFold's CASP15 ranking from 11th to 3rd place, other strategies contributed too. We analyze these in the context of existing strategies to improve prediction.

Key aspects of the past 30 years of protein design

Proteins are the workhorse of life. They are the building infrastructure of living systems; they are the most efficient molecular machines known, and their enzymatic activity is still unmatched in versatility by any artificial system. Perhaps proteins' most remarkable feature is their modularity. The large amount of information required to specify each protein's function is analogically encoded with an alphabet of just ∼20 letters. The protein folding problem is how to encode all such information in a sequence of 20 letters. In this review, we go through the last 30 years of research to summarize the state of the art and highlight some applications related to fundamental problems of protein evolution.

Assessment of domain interactions in the fourteenth round of the Critical Assessment of Structure Prediction (CASP14)

The high accuracy of some CASP14 models at the domain level prompted a more detailed evaluation of structure predictions on whole targets. For the first time in critical assessment of structure prediction (CASP), we evaluated accuracy of difficult domain assembly in models submitted for multidomain targets where the community predicted individual evaluation units (EUs) with greater accuracy than full-length targets. Ten proteins with domain interactions that did not show evidence of conformational change and were not involved in significant oligomeric contacts were chosen as targets for the domain interaction assessment. Groups were ranked using complementary interaction scores (F1, QS score, and Jaccard coefficient), and their predictions were evaluated for their ability to correctly model inter-domain interfaces and overall protein folds. Target performance was broadly grouped into two clusters. The first consisted primarily of targets containing two EUs wherein predictors more broadly predicted domain positioning and interfacial contacts correctly. The other consisted of complex two-EU and three-EU targets where few predictors performed well. The highest ranked predictor, AlphaFold2, produced high-accuracy models on eight out of 10 targets. Their interdomain scores on three of these targets were significantly higher than all other groups and were responsible for their overall outperformance in the category. We further highlight the performance of AlphaFold2 and the next best group, BAKER-experimental on several interesting targets.

Continuous Automated Model EvaluatiOn (CAMEO)-Perspectives on the future of fully automated evaluation of structure prediction methods

The Continuous Automated Model EvaluatiOn (CAMEO) platform complements the biennial CASP experiment by conducting fully automated blind evaluations of three‐dimensional protein prediction servers based on the weekly prerelease of sequences of those structures, which are going to be published in the upcoming release of the Protein Data Bank. While in CASP14, significant success was observed in predicting the structures of individual protein chains with high accuracy, significant challenges remain in correctly predicting the structures of complexes. By implementing fully automated evaluation of predictions for protein–protein complexes, as well as for proteins in complex with ligands, peptides, nucleic acids, or proteins containing noncanonical amino acid residues, CAMEO will assist new developments in those challenging areas of active research.

Target classification in the 14th round of the critical assessment of protein structure prediction (CASP14)

An evolutionary-based definition and classification of target evaluation units (EUs) is presented for the 14th round of the critical assessment of structure prediction (CASP14). CASP14 targets included 84 experimental models submitted by various structural groups (designated T1024-T1101). Targets were split into EUs based on the domain organization of available templates and performance of server groups. Several targets required splitting (19 out of 25 multidomain targets) due in part to observed conformation changes. All in all, 96 CASP14 EUs were defined and assigned to tertiary structure assessment categories (Topology-based FM or High Accuracy-based TBM-easy and TBM-hard) considering their evolutionary relationship to existing ECOD fold space: 24 family level, 50 distant homologs (H-group), 12 analogs (X-group), and 10 new folds. Principal component analysis and heatmap visualization of sequence and structure similarity to known templates as well as performance of servers highlighted trends in CASP14 target difficulty. The assigned evolutionary levels (i.e., H-groups) and assessment classes (i.e., FM) displayed overlapping clusters of EUs. Many viral targets diverged considerably from their template homologs and thus were more difficult for prediction than other homology-related targets. On the other hand, some targets did not have sequence-identifiable templates, but were predicted better than expected due to relatively simple arrangements of secondary structural elements. An apparent improvement in overall server performance in CASP14 further complicated traditional classification, which ultimately assigned EUs into high-accuracy modeling (27 TBM-easy and 31 TBM-hard), topology (23 FM), or both (15 FM/TBM).

Probabilistic divergence of a template-based modelling methodology from the ideal protocol

Protein structural information is essential for the detailed mapping of a functional protein network. For a higher modelling accuracy and quicker implementation, template-based algorithms have been extensively deployed and redefined. The methods only assess the predicted structure against its native state/template and do not estimate the accuracy for each modelling step. A divergence measure is therefore postulated to estimate the modelling accuracy against its theoretical optimal benchmark. By freezing the domain boundaries, the divergence measures are predicted for the most crucial steps of a modelling algorithm. To precisely refine the score using weighting constants, big data analysis could further be deployed.

Comparative analysis of methods for evaluation of protein models against native structures

MOTIVATION

Measuring discrepancies between protein models and native structures is at the heart of development of protein structure prediction methods and comparison of their performance. A number of different evaluation methods have been developed; however, their comprehensive and unbiased comparison has not been performed.

RESULTS

We carried out a comparative analysis of several popular model assessment methods (RMSD, TM-score, GDT, QCS, CAD-score, LDDT, SphereGrinder and RPF) to reveal their relative strengths and weaknesses. The analysis, performed on a large and diverse model set derived in the course of three latest community-wide CASP experiments (CASP10-12), had two major directions. First, we looked at general differences between the scores by analyzing distribution, correspondence and correlation of their values as well as differences in selecting best models. Second, we examined the score differences taking into account various structural properties of models (stereochemistry, hydrogen bonds, packing of domains and chain fragments, missing residues, protein length and secondary structure). Our results provide a solid basis for an informed selection of the most appropriate score or combination of scores depending on the task at hand.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

CASP13 target classification into tertiary structure prediction categories

Protein target structures for the Critical Assessment of Structure Prediction round 13 (CASP13) were split into evaluation units (EUs) based on their structural domains, the domain organization of available templates, and the performance of servers on whole targets compared to split target domains. Eighty targets were split into 112 EUs. The EUs were classified into categories suitable for assessment of high accuracy modeling (or template-based modeling [TBM]) and topology (or free modeling [FM]) based on target difficulty. Assignment into assessment categories considered the following criteria: (a) the evolutionary relationship of target domains to existing fold space as defined by the Evolutionary Classification of Protein Domains (ECOD) database; (b) the clustering of target domains using eight objective sequence, structure, and performance measures; and (c) the placement of target domains in a scatter plot of target difficulty against server performance used in the previous CASP. Generally, target domains with good server predictions had close template homologs and were classified as TBM. Alternately, targets with poor server predictions represent a mixture of fast evolving homologs, structure analogs, and new folds, and were classified as FM or FM/TBM overlap.

DeepSF: deep convolutional neural network for mapping protein sequences to folds

Motivation

Protein fold recognition is an important problem in structural bioinformatics. Almost all traditional fold recognition methods use sequence (homology) comparison to indirectly predict the fold of a target protein based on the fold of a template protein with known structure, which cannot explain the relationship between sequence and fold. Only a few methods had been developed to classify protein sequences into a small number of folds due to methodological limitations, which are not generally useful in practice.

Results

We develop a deep 1D-convolution neural network (DeepSF) to directly classify any protein sequence into one of 1195 known folds, which is useful for both fold recognition and the study of sequence–structure relationship. Different from traditional sequence alignment (comparison) based methods, our method automatically extracts fold-related features from a protein sequence of any length and maps it to the fold space. We train and test our method on the datasets curated from SCOP1.75, yielding an average classification accuracy of 75.3%. On the independent testing dataset curated from SCOP2.06, the classification accuracy is 73.0%. We compare our method with a top profile–profile alignment method—HHSearch on hard template-based and template-free modeling targets of CASP9-12 in terms of fold recognition accuracy. The accuracy of our method is 12.63–26.32% higher than HHSearch on template-free modeling targets and 3.39–17.09% higher on hard template-based modeling targets for top 1, 5 and 10 predicted folds. The hidden features extracted from sequence by our method is robust against sequence mutation, insertion, deletion and truncation, and can be used for other protein pattern recognition problems such as protein clustering, comparison and ranking.

Availability and implementation

The DeepSF server is publicly available at: http://iris.rnet.missouri.edu/DeepSF/.

Supplementary information

Supplementary data are available at Bioinformatics online.

Scalable remote homology detection and fold recognition in massive protein networks

The global connectivities in very large protein similarity networks contain traces of evolution among the proteins for detecting protein remote evolutionary relations or structural similarities. To investigate how well a protein network captures the evolutionary information, a key limitation is the intensive computation of pairwise sequence similarities needed to construct very large protein networks. In this article, we introduce label propagation on low-rank kernel approximation (LP-LOKA) for searching massively large protein networks. LP-LOKA propagates initial protein similarities in a low-rank graph by Nyström approximation without computing all pairwise similarities. With scalable parallel implementations based on distributed-memory using message-passing interface and Apache-Hadoop/Spark on cloud, LP-LOKA can search protein networks with one million proteins or more. In the experiments on Swiss-Prot/ADDA/CASP data, LP-LOKA significantly improved protein ranking over the widely used HMM-HMM or profile-sequence alignment methods utilizing large protein networks. It was observed that the larger the protein similarity network, the better the performance, especially on relatively small protein superfamilies and folds. The results suggest that computing massively large protein network is necessary to meet the growing need of annotating proteins from newly sequenced species and LP-LOKA is both scalable and accurate for searching massively large protein networks.

Evaluation of the template-based modeling in CASP12

The article describes results of numerical evaluation of CASP12 models submitted on targets for which structural templates could be identified and for which servers produced models of relatively high accuracy. The emphasis is on analysis of details of models, and how well the models compete with experimental structures. Performance of contributing research groups is measured in terms of backbone accuracy, all-atom local geometry, and the ability to estimate local errors in models. Separate analyses for all participating groups and automatic servers were carried out. Compared with the last CASP, two years ago, there have been significant improvements in a number of areas, particularly the accuracy of protein backbone atoms, accuracy of sequence alignment between models and available structures, increased accuracy over that which can be obtained from simple copying of a closest template, and accuracy of modeling of sub-structures not present in the closest template. These advancements are likely associated with more effective strategies to build non-template regions of the targets ab initio, better algorithms to combine information from multiple templates, enhanced refinement methods, and better methods for estimating model accuracy.

Lipopolysaccharides-Induced Suppression of Innate-Like B Cell Apoptosis Is Enhanced by CpG Oligodeoxynucleotide and Requires Toll-Like Receptors 2 and 4

Innate-like B lymphocytes play an important role in innate immunity in periodontal disease through Toll-like receptor (TLR) signaling. However, it is unknown how innate-like B cell apoptosis is affected by the periodontal infection-associated innate signals. This study is to determine the effects of two major TLR ligands, lipopolysaccharide (LPS) and CpG-oligodeoxynucleotides (CpG-ODN), on innate-like B cell apoptosis. Spleen B cells were isolated from wild type (WT), TLR2 knockout (KO) and TLR4 KO mice and cultured with E. coli LPS alone, P. gingivalis LPS alone, or combined with CpG-ODN for 2 days. B cell apoptosis and expressions of specific apoptosis-related genes were analyzed by flow cytometry and real-time PCR respectively. P. gingivalis LPS, but not E. coli LPS, reduced the percentage of AnnexinV⁺/7-AAD^- cells within IgM^highCD23^lowCD43^-CD93^- marginal zone (MZ) B cell sub-population and IgM^highCD23^lowCD43⁺CD93⁺ innate response activator (IRA) B cell sub-population in WT but not TLR2KO or TLR4KO mice. CpG-ODN combined with P. gingivalis LPS further reduced the percentage of AnnexinV⁺/7-AAD^- cells within MZ B cells and IRA B cells in WT but not TLR2 KO or TLR4 KO mice. Pro-apoptotic CASP4, CASP9 and Dapk1 were significantly down-regulated in P. gingivalis LPS- and CpG-ODN-treated B cells from WT but not TLR2 KO or TLR4 KO mice. Anti-apoptotic IL-10 was significantly up-regulated in P. gingivalis LPS- and CpG-ODN-treated B cells from WT and TLR2 KO but not TLR4 KO mice. These results suggested that both TLR2 and TLR4 signaling are required for P. gingivalis LPS-induced, CpG-ODN-enhanced suppression of innate-like B cell apoptosis.

CASP 11 target classification

Protein target structures for the Critical Assessment of Structure Prediction round 11 (CASP11) and CASP ROLL were split into domains and classified into categories suitable for assessment of template-based modeling (TBM) and free modeling (FM) based on their evolutionary relatedness to existing structures classified by the Evolutionary Classification of Protein Domains (ECOD) database. First, target structures were divided into domain-based evaluation units. Target splits were based on the domain organization of available templates as well as the performance of servers on whole targets compared to split target domains. Second, evaluation units were classified into TBM and FM categories using a combination of measures that evaluate prediction quality and template detectability. Generally, target domains with sequence-related templates and good server prediction performance were classified as TBM, whereas targets without sequence-identifiable templates and low server performance were classified as FM. As in previous CASP experiments, the boundaries for classification were blurred due to the presence of significant insertions and deteriorations in the targets with respect to homologous templates, as well as the presence of templates with partial coverage of new folds. The FM category included 45 target domains, which represents an unprecedented number of difficult CASP targets provided for modeling. Proteins 2016; 84(Suppl 1):20-33. © 2016 Wiley Periodicals, Inc.

Evaluation of free modeling targets in CASP11 and ROLL

We present an assessment of 'template-free modeling' (FM) in CASP11and ROLL. Community-wide server performance suggested the use of automated scores similar to previous CASPs would provide a good system of evaluating performance, even in the absence of comprehensive manual assessment. The CASP11 FM category included several outstanding examples, including successful prediction by the Baker group of a 256-residue target (T0806-D1) that lacked sequence similarity to any existing template. The top server model prediction by Zhang's Quark, which was apparently selected and refined by several manual groups, encompassed the entire fold of target T0837-D1. Methods from the same two groups tended to dominate overall CASP11 FM and ROLL rankings. Comparison of top FM predictions with those from the previous CASP experiment revealed progress in the category, particularly reflected in high prediction accuracy for larger protein domains. FM prediction models for two cases were sufficient to provide functional insights that were otherwise not obtainable by traditional sequence analysis methods. Importantly, CASP11 abstracts revealed that alignment-based contact prediction methods brought about much of the CASP11 progress, producing both of the functionally relevant models as well as several of the other outstanding structure predictions. These methodological advances enabled de novo modeling of much larger domain structures than was previously possible and allowed prediction of functional sites. Proteins 2016; 84(Suppl 1):51-66. © 2015 Wiley Periodicals, Inc.

Modularity of Protein Folds as a Tool for Template-Free Modeling of Structures

Predicting the three-dimensional structure of proteins from their amino acid sequences remains a challenging problem in molecular biology. While the current structural coverage of proteins is almost exclusively provided by template-based techniques, the modeling of the rest of the protein sequences increasingly require template-free methods. However, template-free modeling methods are much less reliable and are usually applicable for smaller proteins, leaving much space for improvement. We present here a novel computational method that uses a library of supersecondary structure fragments, known as Smotifs, to model protein structures. The library of Smotifs has saturated over time, providing a theoretical foundation for efficient modeling. The method relies on weak sequence signals from remotely related protein structures to create a library of Smotif fragments specific to the target protein sequence. This Smotif library is exploited in a fragment assembly protocol to sample decoys, which are assessed by a composite scoring function. Since the Smotif fragments are larger in size compared to the ones used in other fragment-based methods, the proposed modeling algorithm, SmotifTF, can employ an exhaustive sampling during decoy assembly. SmotifTF successfully predicts the overall fold of the target proteins in about 50% of the test cases and performs competitively when compared to other state of the art prediction methods, especially when sequence signal to remote homologs is diminishing. Smotif-based modeling is complementary to current prediction methods and provides a promising direction in addressing the structure prediction problem, especially when targeting larger proteins for modeling.

Each protein folds into a unique three-dimensional structure that enables it to carry out its biological function. Knowledge of the atomic details of protein structures is therefore a key to understanding their function. Advances in high throughput experimental technologies have lead to an exponential increase in the availability of known protein sequences. Although strong progress has been made in experimental protein structure determination, it remains a fact that more than 99% of structural information is provided by computational modeling methods. We describe here a novel structure prediction method, SmotifTF, which uses a unique library of known protein fragments to assemble the three-dimensional structure of a sequence. The fragment library has saturated over time and therefore provides a complete set of building blocks required for model building. The method performs competitively compared to existing methods of structure prediction.

LoopIng: a template-based tool for predicting the structure of protein loops

Motivation: Predicting the structure of protein loops is very challenging, mainly because they are not necessarily subject to strong evolutionary pressure. This implies that, unlike the rest of the protein, standard homology modeling techniques are not very effective in modeling their structure. However, loops are often involved in protein function, hence inferring their structure is important for predicting protein structure as well as function.

Results: We describe a method, LoopIng, based on the Random Forest automated learning technique, which, given a target loop, selects a structural template for it from a database of loop candidates. Compared to the most recently available methods, LoopIng is able to achieve similar accuracy for short loops (4–10 residues) and significant enhancements for long loops (11–20 residues). The quality of the predictions is robust to errors that unavoidably affect the stem regions when these are modeled. The method returns a confidence score for the predicted template loops and has the advantage of being very fast (on average: 1 min/loop).

Availability and implementation:www.biocomputing.it/looping

Contact:anna.tramontano@uniroma1.it

Supplementary information:Supplementary data are available at Bioinformatics online.

Transferable coarse-grained potential for de novo protein folding and design

Protein folding and design are major biophysical problems, the solution of which would lead to important applications especially in medicine. Here we provide evidence of how a novel parametrization of the Caterpillar model may be used for both quantitative protein design and folding. With computer simulations it is shown that, for a large set of real protein structures, the model produces designed sequences with similar physical properties to the corresponding natural occurring sequences. The designed sequences require further experimental testing. For an independent set of proteins, previously used as benchmark, the correct folded structure of both the designed and the natural sequences is also demonstrated. The equilibrium folding properties are characterized by free energy calculations. The resulting free energy profiles not only are consistent among natural and designed proteins, but also show a remarkable precision when the folded structures are compared to the experimentally determined ones. Ultimately, the updated Caterpillar model is unique in the combination of its fundamental three features: its simplicity, its ability to produce natural foldable designed sequences, and its structure prediction precision. It is also remarkable that low frustration sequences can be obtained with such a simple and universal design procedure, and that the folding of natural proteins shows funnelled free energy landscapes without the need of any potentials based on the native structure.

The expanded FindCore method for identification of a core atom set for assessment of protein structure prediction

Maximizing the scientific impact of NMR-based structure determination requires robust and statistically sound methods for assessing the precision of NMR-derived structures. In particular, a method to define a core atom set for calculating superimpositions and validating structure predictions is critical to the use of NMR-derived structures as targets in the CASP competition. FindCore (Snyder and Montelione, Proteins 2005;59:673-686) is a superimposition independent method for identifying a core atom set and partitioning that set into domains. However, as FindCore optimizes superimposition by sensitively excluding not-well-defined atoms, the FindCore core may not comprise all atoms suitable for use in certain applications of NMR structures, including the CASP assessment process. Adapting the FindCore approach to assess predicted models against experimental NMR structures in CASP10 required modification of the FindCore method. This paper describes conventions and a standard protocol to calculate an "Expanded FindCore" atom set suitable for validation and application in biological and biophysical contexts. A key application of the Expanded FindCore method is to identify a core set of atoms in the experimental NMR structure for which it makes sense to validate predicted protein structure models. We demonstrate the application of this Expanded FindCore method in characterizing well-defined regions of 18 NMR-derived CASP10 target structures. The Expanded FindCore protocol defines "expanded core atom sets" that match an expert's intuition of which parts of the structure are sufficiently well defined to use in assessing CASP model predictions. We also illustrate the impact of this analysis on the CASP GDT assessment scores.

Assessment of template-based protein structure predictions in CASP10

Template-based modeling (TBM) is a major component of the critical assessment of protein structure prediction (CASP). In CASP10, some 41,740 predicted models submitted by 150 predictor groups were assessed as TBM predictions. The accuracy of protein structure prediction was assessed by geometric comparison with experimental X-ray crystal and NMR structures using a composite score that included both global alignment metrics and distance-matrix-based metrics. These included GDT-HA and GDC-all global alignment scores, and the superimposition-independent LDDT distance-matrix-based score. In addition, a superimposition-independent RPF metric, similar to that described previously for comparing protein models against experimental NMR data, was used for comparing predicted protein structure models against experimental protein structures. To score well on all four of these metrics, models must feature accurate predictions of both backbone and side-chain conformations. Performance rankings were determined independently for server and the combined server plus human-curated predictor groups. Final rankings were made using paired head-to-head Student's t-test analysis of raw metric scores among the top 25 performing groups in each category.

Template-based C8-SCORPION: a protein 8-state secondary structure prediction method using structural information and context-based features

Background

Secondary structures prediction of proteins is important to many protein structure modeling applications. Correct prediction of secondary structures can significantly reduce the degrees of freedom in protein tertiary structure modeling and therefore reduces the difficulty of obtaining high resolution 3D models.

Methods

In this work, we investigate a template-based approach to enhance 8-state secondary structure prediction accuracy. We construct structural templates from known protein structures with certain sequence similarity. The structural templates are then incorporated as features with sequence and evolutionary information to train two-stage neural networks. In case of structural templates absence, heuristic structural information is incorporated instead.

Results

After applying the template-based 8-state secondary structure prediction method, the 7-fold cross-validated Q8 accuracy is 78.85%. Even templates from structures with only 20%~30% sequence similarity can help improve the 8-state prediction accuracy. More importantly, when good templates are available, the prediction accuracy of less frequent secondary structures, such as 3-10 helices, turns, and bends, are highly improved, which are useful for practical applications.

Conclusions

Our computational results show that the templates containing structural information are effective features to enhance 8-state secondary structure predictions. Our prediction algorithm is implemented on a web server named "C8-SCORPION" available at: http://hpcr.cs.odu.edu/c8scorpion.

A comparative assessment and analysis of 20 representative sequence alignment methods for protein structure prediction

Protein sequence alignment is essential for template-based protein structure prediction and function annotation. We collect 20 sequence alignment algorithms, 10 published and 10 newly developed, which cover all representative sequence- and profile-based alignment approaches. These algorithms are benchmarked on 538 non-redundant proteins for protein fold-recognition on a uniform template library. Results demonstrate dominant advantage of profile-profile based methods, which generate models with average TM-score 26.5% higher than sequence-profile methods and 49.8% higher than sequence-sequence alignment methods. There is no obvious difference in results between methods with profiles generated from PSI-BLAST PSSM matrix and hidden Markov models. Accuracy of profile-profile alignments can be further improved by 9.6% or 21.4% when predicted or native structure features are incorporated. Nevertheless, TM-scores from profile-profile methods including experimental structural features are still 37.1% lower than that from TM-align, demonstrating that the fold-recognition problem cannot be solved solely by improving accuracy of structure feature predictions.

Context-based features enhance protein secondary structure prediction accuracy

We report a new approach of using statistical context-based scores as encoded features to train neural networks to achieve secondary structure prediction accuracy improvement. The context-based scores are pseudo-potentials derived by evaluating statistical, high-order inter-residue interactions, which estimate the favorability of a residue adopting certain secondary structure conformation within its amino acid environment. Encoding these context-based scores as important training and prediction features provides a way to address a long-standing difficulty in neural network-based secondary structure predictions of taking interdependency among secondary structures of neighboring residues into account. Our computational results have shown that the context-based scores are effective features to enhance the prediction accuracy of secondary structure predictions. An overall 7-fold cross-validated Q3 accuracy of 82.74% and Segment Overlap Accuracy (SOV) accuracy of 86.25% are achieved on a set of more than 7987 protein chains with, at most, 25% sequence identity. The Q3 prediction accuracy on benchmarks of CB513, Manesh215, Carugo338, as well as CASP9 protein chains is higher than popularly used secondary structure prediction servers, including Psipred, Profphd, Jpred, Porter (ab initio), and Netsurf. More significant improvement is observed in the SOV accuracy, where more than 4% enhancement is observed, compared to the server with the best SOV accuracy. A Q8 accuracy of >70% (71.5%) is also found in eight-state secondary structure prediction. The majority of the Q3 accuracy improvement is contributed from correctly identifying β-sheets and α-helices. When the context-based scores are incorporated, there are 15.5% more residues predicted with >90% confidence. These high-confidence predictions usually have a rather high accuracy (averagely ~95%). The three- and eight-state prediction servers (SCORPION) implementing our methods are available online.

Protein structure refinement of CASP target proteins using GNEIMO torsional dynamics method

A longstanding challenge in using computational methods for protein structure prediction is the refinement of low-resolution structural models derived from comparative modeling methods into highly accurate atomistic models useful for detailed structural studies. Previously, we have developed and demonstrated the utility of the internal coordinate molecular dynamics (MD) technique, generalized Newton-Euler inverse mass operator (GNEIMO), for refinement of small proteins. Using GNEIMO, the high-frequency degrees of freedom are frozen and the protein is modeled as a collection of rigid clusters connected by torsional hinges. This physical model allows larger integration time steps and focuses the conformational search in the low frequency torsional degrees of freedom. Here, we have applied GNEIMO with temperature replica exchange to refine low-resolution protein models of 30 proteins taken from the continuous assessment of structure prediction (CASP) competition. We have shown that GNEIMO torsional MD method leads to refinement of up to 1.3 Å in the root-mean-square deviation in coordinates for 30 CASP target proteins without using any experimental data as restraints in performing the GNEIMO simulations. This is in contrast with the unconstrained all-atom Cartesian MD method performed under the same conditions, where refinement requires the use of restraints during the simulations.

Critical assessment of methods of protein structure prediction (CASP)--round x

This article is an introduction to the special issue of the journal PROTEINS, dedicated to the tenth Critical Assessment of Structure Prediction (CASP) experiment to assess the state of the art in protein structure modeling. The article describes the conduct of the experiment, the categories of prediction included, and outlines the evaluation and assessment procedures. The 10 CASP experiments span almost 20 years of progress in the field of protein structure modeling, and there have been enormous advances in methods and model accuracy in that period. Notable in this round is the first sustained improvement of models with refinement methods, using molecular dynamics. For the first time, we tested the ability of modeling methods to make use of sparse experimental three-dimensional contact information, such as may be obtained from new experimental techniques, with encouraging results. On the other hand, new contact prediction methods, though holding considerable promise, have yet to make an impact in CASP testing. The nature of CASP targets has been changing in recent CASPs, reflecting shifts in experimental structural biology, with more irregular structures, more multi-domain and multi-subunit structures, and less standard versions of known folds. When allowance is made for these factors, we continue to see steady progress in the overall accuracy of models, particularly resulting from improvement of non-template regions.

Critical assessment of methods of protein structure prediction (CASP)--round x

This article is an introduction to the special issue of the journal PROTEINS, dedicated to the tenth Critical Assessment of Structure Prediction (CASP) experiment to assess the state of the art in protein structure modeling. The article describes the conduct of the experiment, the categories of prediction included, and outlines the evaluation and assessment procedures. The 10 CASP experiments span almost 20 years of progress in the field of protein structure modeling, and there have been enormous advances in methods and model accuracy in that period. Notable in this round is the first sustained improvement of models with refinement methods, using molecular dynamics. For the first time, we tested the ability of modeling methods to make use of sparse experimental three-dimensional contact information, such as may be obtained from new experimental techniques, with encouraging results. On the other hand, new contact prediction methods, though holding considerable promise, have yet to make an impact in CASP testing. The nature of CASP targets has been changing in recent CASPs, reflecting shifts in experimental structural biology, with more irregular structures, more multi-domain and multi-subunit structures, and less standard versions of known folds. When allowance is made for these factors, we continue to see steady progress in the overall accuracy of models, particularly resulting from improvement of non-template regions.

Critical assessment of methods of protein structure prediction (CASP)--round x

This article is an introduction to the special issue of the journal PROTEINS, dedicated to the tenth Critical Assessment of Structure Prediction (CASP) experiment to assess the state of the art in protein structure modeling. The article describes the conduct of the experiment, the categories of prediction included, and outlines the evaluation and assessment procedures. The 10 CASP experiments span almost 20 years of progress in the field of protein structure modeling, and there have been enormous advances in methods and model accuracy in that period. Notable in this round is the first sustained improvement of models with refinement methods, using molecular dynamics. For the first time, we tested the ability of modeling methods to make use of sparse experimental three-dimensional contact information, such as may be obtained from new experimental techniques, with encouraging results. On the other hand, new contact prediction methods, though holding considerable promise, have yet to make an impact in CASP testing. The nature of CASP targets has been changing in recent CASPs, reflecting shifts in experimental structural biology, with more irregular structures, more multi-domain and multi-subunit structures, and less standard versions of known folds. When allowance is made for these factors, we continue to see steady progress in the overall accuracy of models, particularly resulting from improvement of non-template regions.

Assessment of template-free modeling in CASP10 and ROLL

We present the assessment of predictions for Template-Free Modeling in CASP10 and a report on the first ROLL experiment wherein predictions are collected year round for review at the regular CASP season. Models were first clustered so that duplicated or very similar ones were grouped together and represented by one model in the cluster. The representatives were then compared with targets using GDT_TS, QCS, and three additional superposition-independent score functions newly developed for CASP10. For each target, the top 15 representatives by each score were pooled to form the Top15Union set. All models in this set were visually inspected by four of us independently using the new plugin, EvalScore, which we developed with the UCSF Chimera group. The best models were selected for each target after extensive debate among the four examiners. Groups were ranked by the number of targets (hits) for which a group's model was selected as one of the best models. The Keasar group had most hits in both categories, with four of 19 FM and eight of 36 ROLL targets. The most successful prediction servers were QUARK from Zhang's group for FM category with three hits and Zhang-server for the ROLL category with seven hits. As observed in CASP9, many successful groups were not true "template-free" modelers but used remote templates and/or server models to obtain their winning models. The results of the first ROLL experiment were broadly similar to those of the CASP10 FM exercise.

CASP10 results compared to those of previous CASP experiments

We compare results of the community efforts in modeling protein structures in the tenth CASP experiment, with those in earlier CASPs particularly in CASP5, a decade ago. There is a substantial improvement in template based model accuracy as reflected in more successful modeling of regions of structure not easily derived from a single experimental structure template, most likely reflecting intensive work within the modeling community in developing methods that make use of multiple templates, as well as the increased number of experimental structures available. Deriving structural information not obvious from a template is the most demanding as well as one of the most useful tasks that modeling can perform. Thus this is gratifying progress. By contrast, overall backbone accuracy of models appears little changed in the last decade. This puzzling result is explained by two factors--increased database size in some ways makes it harder to choose the best available templates, and the increased intrinsic difficulty of CASP targets as experimental work has progressed to larger and more unusual structures. There is no detectable recent improvement in template-free modeling, but again, this may reflect the changing nature of CASP targets.

Definition and classification of evaluation units for CASP10

For the 10th experiment on Critical Assessment of the techniques of protein Structure Prediction (CASP), the prediction target proteins were broken into independent evaluation units (EUs), which were then classified into template-based modeling (TBM) or free modeling (FM) categories. We describe here how the EUs were defined and classified, what issues arose in the process, and how we resolved them. EUs are frequently not the whole target proteins but the constituting structural domains. However, the assessors from CASP7 on combined more than one domain into 1 EU for some targets, which implied that the assessment also included evaluation of the prediction of the relative position and orientation of these domains. In CASP10, we followed and expanded this notion by defining multidomain EUs for a number of targets. These included 3 EUs, each made of two domains of familiar fold but arranged in a novel manner and for which the focus of evaluation was the interdomain arrangement. An EU was classified to the TBM category if a template could be found by sequence similarity searches and to FM if a structural template could not be found by structural similarity searches. The EUs that did not fall cleanly in either of these cases were classified case-by-case, often including consideration of the overall quality and characteristics of the predictions.

Dinosolve: a protein disulfide bonding prediction server using context-based features to enhance prediction accuracy

Background

Disulfide bonds play an important role in protein folding and structure stability. Accurately predicting disulfide bonds from protein sequences is important for modeling the structural and functional characteristics of many proteins.

Methods

In this work, we introduce an approach of enhancing disulfide bonding prediction accuracy by taking advantage of context-based features. We firstly derive the first-order and second-order mean-force potentials according to the amino acid environment around the cysteine residues from large number of cysteine samples. The mean-force potentials are integrated as context-based scores to estimate the favorability of a cysteine residue in disulfide bonding state as well as a cysteine pair in disulfide bond connectivity. These context-based scores are then incorporated as features together with other sequence and evolutionary information to train neural networks for disulfide bonding state prediction and connectivity prediction.

Results

The 10-fold cross validated accuracy is 90.8% at residue-level and 85.6% at protein-level in classifying an individual cysteine residue as bonded or free, which is around 2% accuracy improvement. The average accuracy for disulfide bonding connectivity prediction is also improved, which yields overall sensitivity of 73.42% and specificity of 91.61%.

Conclusions

Our computational results have shown that the context-based scores are effective features to enhance the prediction accuracies of both disulfide bonding state prediction and connectivity prediction. Our disulfide prediction algorithm is implemented on a web server named "Dinosolve" available at: http://hpcr.cs.odu.edu/dinosolve.

lDDT: a local superposition-free score for comparing protein structures and models using distance difference tests

Motivation: The assessment of protein structure prediction techniques requires objective criteria to measure the similarity between a computational model and the experimentally determined reference structure. Conventional similarity measures based on a global superposition of carbon α atoms are strongly influenced by domain motions and do not assess the accuracy of local atomic details in the model.

Results: The Local Distance Difference Test (lDDT) is a superposition-free score that evaluates local distance differences of all atoms in a model, including validation of stereochemical plausibility. The reference can be a single structure, or an ensemble of equivalent structures. We demonstrate that lDDT is well suited to assess local model quality, even in the presence of domain movements, while maintaining good correlation with global measures. These properties make lDDT a robust tool for the automated assessment of structure prediction servers without manual intervention.

Availability and implementation: Source code, binaries for Linux and MacOSX, and an interactive web server are available at http://swissmodel.expasy.org/lddt

Contact:torsten.schwede@unibas.ch

Supplementary information: Supplementary data are available at Bioinformatics online.

A homology/ab initio hybrid algorithm for sampling near-native protein conformations

One of the major challenges for protein tertiary structure prediction strategies is the quality of conformational sampling algorithms, which can effectively and readily search the protein fold space to generate near-native conformations. In an effort to advance the field by making the best use of available homology as well as fold recognition approaches along with ab initio folding methods, we have developed Bhageerath-H Strgen, a homology/ab initio hybrid algorithm for protein conformational sampling. The methodology is tested on the benchmark CASP9 dataset of 116 targets. In 93% of the cases, a structure with TM-score ≥ 0.5 is generated in the pool of decoys. Further, the performance of Bhageerath-H Strgen was seen to be efficient in comparison with different decoy generation methods. The algorithm is web enabled as Bhageerath-H Strgen web tool which is made freely accessible for protein decoy generation (http://www.scfbio-iitd.res.in/software/Bhageerath-HStrgen1.jsp).

Modeling proteins using a super-secondary structure library and NMR chemical shift information

A remaining challenge in protein modeling is to predict structures for sequences with no sequence similarity to any experimentally solved structure. Based on earlier observations, the library of protein backbone supersecondary structure motifs (Smotifs) saturated about a decade ago. Therefore, it should be possible to build any structure from a combination of existing Smotifs with the help of limited experimental data that are sufficient to relate the backbone conformations of Smotifs between target proteins and known structures. Here, we present a hybrid modeling algorithm that relies on an exhaustive Smotif library and on nuclear magnetic resonance chemical shift patterns without any input of primary sequence information. In a test of 102 proteins, the algorithm delivered 90 homology-model-quality models, among them 24 high-quality ones, and a topologically correct solution for almost all cases. The current approach opens a venue to address the modeling of larger protein structures for which chemical shifts are available.

DNdisorder: predicting protein disorder using boosting and deep networks

Background

A number of proteins contain regions which do not adopt a stable tertiary structure in their native state. Such regions known as disordered regions have been shown to participate in many vital cell functions and are increasingly being examined as drug targets.

Results

This work presents a new sequence based approach for the prediction of protein disorder. The method uses boosted ensembles of deep networks to make predictions and participated in the CASP10 experiment. In a 10 fold cross validation procedure on a dataset of 723 proteins, the method achieved an average balanced accuracy of 0.82 and an area under the ROC curve of 0.90. These results are achieved in part by a boosting procedure which is able to steadily increase balanced accuracy and the area under the ROC curve over several rounds. The method also compared competitively when evaluated against a number of state-of-the-art disorder predictors on CASP9 and CASP10 benchmark datasets.

Conclusions

DNdisorder is available as a web service at http://iris.rnet.missouri.edu/dndisorder/.

Designing and benchmarking the MULTICOM protein structure prediction system

Background

Predicting protein structure from sequence is one of the most significant and challenging problems in bioinformatics. Numerous bioinformatics techniques and tools have been developed to tackle almost every aspect of protein structure prediction ranging from structural feature prediction, template identification and query-template alignment to structure sampling, model quality assessment, and model refinement. How to synergistically select, integrate and improve the strengths of the complementary techniques at each prediction stage and build a high-performance system is becoming a critical issue for constructing a successful, competitive protein structure predictor.

Results

Over the past several years, we have constructed a standalone protein structure prediction system MULTICOM that combines multiple sources of information and complementary methods at all five stages of the protein structure prediction process including template identification, template combination, model generation, model assessment, and model refinement. The system was blindly tested during the ninth Critical Assessment of Techniques for Protein Structure Prediction (CASP9) in 2010 and yielded very good performance. In addition to studying the overall performance on the CASP9 benchmark, we thoroughly investigated the performance and contributions of each component at each stage of prediction.

Conclusions

Our comprehensive and comparative study not only provides useful and practical insights about how to select, improve, and integrate complementary methods to build a cutting-edge protein structure prediction system but also identifies a few new sources of information that may help improve the design of a protein structure prediction system. Several components used in the MULTICOM system are available at: http://sysbio.rnet.missouri.edu/multicom_toolbox/.

Building a knowledge-based statistical potential by capturing high-order inter-residue interactions and its applications in protein secondary structure assessment

The rapidly increasing number of protein crystal structures available in the Protein Data Bank (PDB) has naturally made statistical analyses feasible in studying complex high-order inter-residue correlations. In this paper, we report a context-based secondary structure potential (CSSP) for assessing the quality of predicted protein secondary structures generated by various prediction servers. CSSP is a sequence-position-specific knowledge-based potential generated based on the potentials of mean force approach, where high-order inter-residue interactions are taken into consideration. The CSSP potential is effective in identifying secondary structure predictions with good quality. In 56% of the targets in the CB513 benchmark, the optimal CSSP potential is able to recognize the native secondary structure or a prediction with Q3 accuracy higher than 90% as best scored in the predicted secondary structures generated by 10 popularly used secondary structure prediction servers. In more than 80% of the CB513 targets, the predicted secondary structures with the lowest CSSP potential values yield higher than 80% Q3 accuracy. Similar performance of CSSP is found on the CASP9 targets as well. Moreover, our computational results also show that the CSSP potential using triplets outperforms the CSSP potential using doublets and is currently better than the CSSP potential using quartets.

BCL::Fold--de novo prediction of complex and large protein topologies by assembly of secondary structure elements

Computational de novo protein structure prediction is limited to small proteins of simple topology. The present work explores an approach to extend beyond the current limitations through assembling protein topologies from idealized α-helices and β-strands. The algorithm performs a Monte Carlo Metropolis simulated annealing folding simulation. It optimizes a knowledge-based potential that analyzes radius of gyration, β-strand pairing, secondary structure element (SSE) packing, amino acid pair distance, amino acid environment, contact order, secondary structure prediction agreement and loop closure. Discontinuation of the protein chain favors sampling of non-local contacts and thereby creation of complex protein topologies. The folding simulation is accelerated through exclusion of flexible loop regions further reducing the size of the conformational search space. The algorithm is benchmarked on 66 proteins with lengths between 83 and 293 amino acids. For 61 out of these proteins, the best SSE-only models obtained have an RMSD100 below 8.0 Å and recover more than 20% of the native contacts. The algorithm assembles protein topologies with up to 215 residues and a relative contact order of 0.46. The method is tailored to be used in conjunction with low-resolution or sparse experimental data sets which often provide restraints for regions of defined secondary structure.

Recursive protein modeling: a divide and conquer strategy for Protein Structure Prediction and its case study in CASP9

After decades of research, protein structure prediction remains a very challenging problem. In order to address the different levels of complexity of structural modeling, two types of modeling techniques--template-based modeling and template-free modeling--have been developed. Template-based modeling can often generate a moderate- to high-resolution model when a similar, homologous template structure is found for a query protein but fails if no template or only incorrect templates are found. Template-free modeling, such as fragment-based assembly, may generate models of moderate resolution for small proteins of low topological complexity. Seldom have the two techniques been integrated together to improve protein modeling. Here we develop a recursive protein modeling approach to selectively and collaboratively apply template-based and template-free modeling methods to model template-covered (i.e. certain) and template-free (i.e. uncertain) regions of a protein. A preliminary implementation of the approach was tested on a number of hard modeling cases during the 9th Critical Assessment of Techniques for Protein Structure Prediction (CASP9) and successfully improved the quality of modeling in most of these cases. Recursive modeling can significantly reduce the complexity of protein structure modeling and integrate template-based and template-free modeling to improve the quality and efficiency of protein structure prediction.

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/.

Ab initio protein structure assembly using continuous structure fragments and optimized knowledge-based force field

Ab initio protein folding is one of the major unsolved problems in computational biology owing to the difficulties in force field design and conformational search. We developed a novel program, QUARK, for template-free protein structure prediction. Query sequences are first broken into fragments of 1-20 residues where multiple fragment structures are retrieved at each position from unrelated experimental structures. Full-length structure models are then assembled from fragments using replica-exchange Monte Carlo simulations, which are guided by a composite knowledge-based force field. A number of novel energy terms and Monte Carlo movements are introduced and the particular contributions to enhancing the efficiency of both force field and search engine are analyzed in detail. QUARK prediction procedure is depicted and tested on the structure modeling of 145 nonhomologous proteins. Although no global templates are used and all fragments from experimental structures with template modeling score >0.5 are excluded, QUARK can successfully construct 3D models of correct folds in one-third cases of short proteins up to 100 residues. In the ninth community-wide Critical Assessment of protein Structure Prediction experiment, QUARK server outperformed the second and third best servers by 18 and 47% based on the cumulative Z-score of global distance test-total scores in the FM category. Although ab initio protein folding remains a significant challenge, these data demonstrate new progress toward the solution of the most important problem in the field.

Target highlights in CASP9: Experimental target structures for the critical assessment of techniques for protein structure prediction

One goal of the CASP community wide experiment on the critical assessment of techniques for protein structure prediction is to identify the current state of the art in protein structure prediction and modeling. A fundamental principle of CASP is blind prediction on a set of relevant protein targets, that is, the participating computational methods are tested on a common set of experimental target proteins, for which the experimental structures are not known at the time of modeling. Therefore, the CASP experiment would not have been possible without broad support of the experimental protein structural biology community. In this article, several experimental groups discuss the structures of the proteins which they provided as prediction targets for CASP9, highlighting structural and functional peculiarities of these structures: the long tail fiber protein gp37 from bacteriophage T4, the cyclic GMP-dependent protein kinase Iβ dimerization/docking domain, the ectodomain of the JTB (jumping translocation breakpoint) transmembrane receptor, Autotaxin in complex with an inhibitor, the DNA-binding J-binding protein 1 domain essential for biosynthesis and maintenance of DNA base-J (β-D-glucosyl-hydroxymethyluracil) in Trypanosoma and Leishmania, an so far uncharacterized 73 residue domain from Ruminococcus gnavus with a fold typical for PDZ-like domains, a domain from the phycobilisome core-membrane linker phycobiliprotein ApcE from Synechocystis, the heat shock protein 90 activators PFC0360w and PFC0270w from Plasmodium falciparum, and 2-oxo-3-deoxygalactonate kinase from Klebsiella pneumoniae.

Assessment of template based protein structure predictions in CASP9

In the Ninth Edition of the Critical Assessment of Techniques for Protein Structure Prediction (CASP9), 61,665 models submitted by 176 groups were assessed for their accuracy in the template based modeling category. The models were evaluated numerically in comparison to their experimental control structures using two global measures (GDT and GDC), and a novel local score evaluating the correct modeling of local interactions (lDDT). Overall, the state of the art of template based modeling in CASP9 is high, with many groups performing well. Among the methods registered as prediction "servers", six independent groups are performing on average better than the rest. The submissions by "human" groups are dominated by meta-predictors, with one group performing noticeably better than the others. Most of the participating groups failed to assign realistic confidence estimates to their predictions, and only a very small fraction of the assessed methods have provided highly accurate models and realistic error estimates at the same time. Also, the accuracy of predictions for homo-oligomeric assemblies was overall poor, and only one group performed better than a naïve control predictor. Here, we present the results of our assessment of the CASP9 predictions in the category of template based modeling, documenting the state of the art and highlighting areas for future developments.