MOTIVATION: AlphaFold2 significantly improved the prediction of protein complex structures. However, its accuracy is lower for interactions without coevolutionary signals, such as host-pathogen and antibody-antigen interactions. Two strategies have been developed to address this limitation: massive sampling and replacing the evoformer with the pairformer, which does not rely on coevolution, as introduced in AlphaFold3, thereby enabling more structural reasoning by the network. RESULTS: In this study, we benchmark structure prediction methods on unseen antibody-antigen complexes. We found that increased sampling improves the chances of generating a correct protein model, roughly in a log-linear manner. However, the internal quality estimates by AlphaFold often cannot identify the best predicted structures for each target, resulting in a significant loss of performance for the top-ranked protein model compared with the best model. For all methods, a significant challenge remains the identification of the best model. We also show that AlphaFold3 outperforms AlphaFold2, Boltz-1, and Chai-1. Furthermore, AlphaFold3 performance declines significantly for complexes that lack structural similarity to the training set, indicating that it has to some extent learned to detect remote structural similarities. AVAILABILITY AND IMPLEMENTATION: All code is available from github.com/samuelfromm/abag-benchmark-set/ and all data from DOI: 10.5281/zenodo.17978681. The latter repository also contains the code.

MOTIVATION: G-quadruplex-binding proteins (G4BPs) play key roles in RNA metabolism and stress response, yet their identification remains experimentally challenging. Here, we present a deep learning (DL) framework for the prediction of RNA G4BPs (RG4BPs), integrating diverse encoding strategies and neural architectures. Our best-performing model, which includes ESM-2 protein language model embeddings and consists of an LSTM architecture, achieved 86% accuracy in distinguishing RG4BPs from non-binder proteins. The application of this model to the human proteome uncovered 2160 high-confidence RG4BP candidates, many of which display intrinsically disordered regions (IDRs) and enrichment in stress granule organelles. These findings reveal a potential link between G-quadruplex recognition and cellular stress responses. To enable easy and broad access to the framework, we developed G4REP, a web server for RG4BP prediction and analysis. Overall, an effective approach to explore the RG4BPs landscape and uncover novel players in RNA regulation is provided. AVAILABILITY: Source code for the G4REP Model training and evaluation is available at: https://github.com/G4REP/G4REPmodel and at https://doi.org/10.5281/zenodo.17963046. G4REP Server is hosted at: https://schubert.bio.uniroma1.it/g4/.

Recent advances in geometric deep learning and generative modeling have enabled the design of novel proteins with a wide range of desired properties. However, current state-of-the-art approaches are typically restricted to generating proteins with only static target properties, such as motifs and symmetries. In this work, we take a step towards overcoming this limitation by proposing a framework to condition structure generation on flexibility, which is crucial for key functionalities such as catalysis or molecular recognition. We first introduce BackFlip, an equivariant neural network for predicting per-residue flexibility from an input backbone structure. Relying on BackFlip, we propose FliPS, an SE(3)-equivariant conditional flow matching model that solves the inverse problem, that is, generating backbones that display a target flexibility profile. In our experiments, we show that FliPS is able to generate novel and diverse protein backbones with the desired flexibility, verified by Molecular Dynamics (MD) simulations. FliPS and BackFlip are available at https://github.com/graeter-group/flips.

Molecular structure generation is a fundamental problem that involves determining the 3D positions of molecules’ constituents. It has crucial biological applications, such as molecular docking, protein folding, and molecular design. Recent advances in generative modeling, such as diffusion models and flow matching, have made great progress on these tasks by modeling molecular conformations as a distribution. In this work, we focus on flow matching and adopt an energy-based perspective to improve training and inference of structure generation models. Our view results in a mapping function, represented by a deep network, that is directly learned to iteratively map random configurations, i.e. samples from the source distribution, to target structures, i.e. points in the data manifold. This yields a conceptually simple and empirically effective flow matching setup that is theoretically justified and has interesting connections to fundamental properties such as idempotency and stability, as well as the empirically useful techniques such as structure refinement in AlphaFold. Experiments on protein docking as well as protein backbone generation consistently demonstrate the method’s effectiveness, where it outperforms recent baselines of task-associated flow matching and diffusion models, using a similar computational budget.

MOTIVATION: Today, the prediction of structures of large protein complexes solely from their sequence information requires prior knowledge of the stoichiometry of the complex. To address this challenge, we have enhanced the Monte Carlo Tree Search algorithms in MoLPC to enable the assembly of protein complexes while simultaneously predicting their stoichiometry. RESULTS: In MoLPC2, we have improved the predictions by allowing sampling alternative AlphaFold predictions. Using MoLPC2, we accurately predicted the structures of 50 out of 175 nonredundant protein complexes (TM-score ≥ 0.8) without knowing the stoichiometry. MoLPC2 provides new opportunities for predicting protein complex structures without stoichiometry information. AVAILABILITY AND IMPLEMENTATION: MoLPC2 is freely available at https://github.com/hychim/molpc2. A notebook is also available from the repository for easy use.

MOTIVATION: Understanding metal-protein interaction can provide structural and functional insights into cellular processes. As the number of protein sequences increases, developing fast yet precise computational approaches to predict and annotate metal-binding sites becomes imperative. Quick and resource-efficient pre-trained protein language model (pLM) embeddings have successfully predicted binding sites from protein sequences despite not using structural or evolutionary features (multiple sequence alignments). Using residue-level embeddings from the pLMs, we have developed a sequence-based method (M-Ionic) to identify metal-binding proteins and predict residues involved in metal binding. RESULTS: On independent validation of recent proteins, M-Ionic reports an area under the curve (AUROC) of 0.83 (recall = 84.6%) in distinguishing metal binding from non-binding proteins compared to AUROC of 0.74 (recall = 61.8%) of the next best method. In addition to comparable performance to the state-of-the-art method for identifying metal-binding residues (Ca2+, Mg2+, Mn2+, Zn2+), M-Ionic provides binding probabilities for six additional ions (i.e. Cu2+, Po43-, So42-, Fe2+, Fe3+, Co2+). We show that the pLM embedding of a single residue contains sufficient information about its neighbours to predict its binding properties. AVAILABILITY AND IMPLEMENTATION: M-Ionic can be used on your protein of interest using a Google Colab Notebook (https://bit.ly/40FrRbK). The GitHub repository (https://github.com/TeamSundar/m-ionic) contains all code and data.

MOTIVATION: Protein-protein interaction (PPI) networks and transcriptional regulatory networks are critical in regulating cells and their signaling. A thorough understanding of PPIs can provide more insights into cellular physiology at normal and disease states. Although numerous methods have been proposed to predict PPIs, it is still challenging for interaction prediction between unknown proteins. In this study, a novel neural network named AFTGAN was constructed to predict multi-type PPIs. Regarding feature input, ESM-1b embedding containing much biological information for proteins was added as a protein sequence feature besides amino acid co-occurrence similarity and one-hot coding. An ensemble network was also constructed based on a transformer encoder containing an AFT module (performing the weight operation on vital protein sequence feature information) and graph attention network (extracting the relational features of protein pairs) for the part of the network framework. RESULTS: The experimental results showed that the Micro-F1 of the AFTGAN based on three partitioning schemes (BFS, DFS and the random mode) on the SHS27K and SHS148K datasets was 0.685, 0.711 and 0.867, as well as 0.745, 0.819 and 0.920, respectively, all higher than that of other popular methods. In addition, the experimental comparisons confirmed the performance superiority of the proposed model for predicting PPIs of unknown proteins on the STRING dataset. AVAILABILITY AND IMPLEMENTATION: The source code is publicly available at https://github.com/1075793472/AFTGAN. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Despite near-experimental accuracy on single-chain predictions, there is still scope for improvement among multimeric predictions. Methods like AlphaFold-Multimer and FoldDock can accurately model dimers. However, how well these methods fare on larger complexes is still unclear. Further, evaluation methods of the quality of multimeric complexes are not well established. RESULTS: We analysed the performance of AlphaFold-Multimer on a homology-reduced dataset of homo- and heteromeric protein complexes. We highlight the differences between the pairwise and multi-interface evaluation of chains within a multimer. We describe why certain complexes perform well on one metric (e.g. TM-score) but poorly on another (e.g. DockQ). We propose a new score, Predicted DockQ version 2 (pDockQ2), to estimate the quality of each interface in a multimer. Finally, we modelled protein complexes (from CORUM) and identified two highly confident structures that do not have sequence homology to any existing structures. AVAILABILITY AND IMPLEMENTATION: All scripts, models, and data used to perform the analysis in this study are freely available at https://gitlab.com/ElofssonLab/afm-benchmark.

MOTIVATION: In the last decade, de novo protein structure prediction accuracy for individual proteins has improved significantly by utilising deep learning (DL) methods for harvesting the co-evolution information from large multiple sequence alignments (MSAs). The same approach can, in principle, also be used to extract information about evolutionary-based contacts across protein-protein interfaces. However, most earlier studies have not used the latest DL methods for inter-chain contact distance prediction. This article introduces a fold-and-dock method based on predicted residue-residue distances with trRosetta. RESULTS: The method can simultaneously predict the tertiary and quaternary structure of a protein pair, even when the structures of the monomers are not known. The straightforward application of this method to a standard dataset for protein-protein docking yielded limited success. However, using alternative methods for generating MSAs allowed us to dock accurately significantly more proteins. We also introduced a novel scoring function, PconsDock, that accurately separates 98% of correctly and incorrectly folded and docked proteins. The average performance of the method is comparable to the use of traditional, template-based or ab initio shape-complementarity-only docking methods. Moreover, the results of conventional and fold-and-dock approaches are complementary, and thus a combined docking pipeline could increase overall docking success significantly. This methodology contributed to the best model for one of the CASP14 oligomeric targets, H1065. AVAILABILITY AND IMPLEMENTATION: All scripts for predictions and analysis are available from https://github.com/ElofssonLab/bioinfo-toolbox/ and https://gitlab.com/ElofssonLab/benchmark5/. All models joined alignments, and evaluation results are available from the following figshare repository https://doi.org/10.6084/m9.figshare.14654886.v2. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Proteins are ubiquitous molecules whose function in biological processes is determined by their 3D structure. Experimental identification of a protein's structure can be time-consuming, prohibitively expensive and not always possible. Alternatively, protein folding can be modeled using computational methods, which however are not guaranteed to always produce optimal results. GraphQA is a graph-based method to estimate the quality of protein models, that possesses favorable properties such as representation learning, explicit modeling of both sequential and 3D structure, geometric invariance and computational efficiency. RESULTS: GraphQA performs similarly to state-of-the-art methods despite using a relatively low number of input features. In addition, the graph network structure provides an improvement over the architecture used in ProQ4 operating on the same input features. Finally, the individual contributions of GraphQA components are carefully evaluated. AVAILABILITY AND IMPLEMENTATION: PyTorch implementation, datasets, experiments and link to an evaluation server are available through this GitHub repository: github.com/baldassarreFe/graphqa. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Contact predictions within a protein have recently become a viable method for accurate prediction of protein structure. Using predicted distance distributions has been shown in many cases to be superior to only using a binary contact annotation. Using predicted interprotein distances has also been shown to be able to dock some protein dimers. RESULTS: Here, we present pyconsFold. Using CNS as its underlying folding mechanism and predicted contact distance it outperforms regular contact prediction-based modeling on our dataset of 210 proteins. It performs marginally worse than the state-of-the-art pyRosetta folding pipeline but is on average about 20 times faster per model. More importantly pyconsFold can also be used as a fold-and-dock protocol by using predicted interprotein contacts/distances to simultaneously fold and dock two protein chains. AVAILABILITY AND IMPLEMENTATION: pyconsFold is implemented in Python 3 with a strong focus on using as few dependencies as possible for longevity. It is available both as a pip package in Python 3 and as source code on GitHub and is published under the GPLv3 license. The data underlying this article together with source code are available on github, at https://github.com/johnlamb/pyconsfold. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Repeat proteins are abundant in eukaryotic proteomes. They are involved in many eukaryotic specific functions, including signalling. For many of these proteins, the structure is not known, as they are difficult to crystallise. Today, using direct coupling analysis and deep learning it is often possible to predict a protein's structure. However, the unique sequence features present in repeat proteins have been a challenge to use direct coupling analysis for predicting contacts. Here, we show that deep learning-based methods (trRosetta, DeepMetaPsicov (DMP) and PconsC4) overcomes this problem and can predict intra- and inter-unit contacts in repeat proteins. In a benchmark dataset of 815 repeat proteins, about 90% can be correctly modelled. Further, among 48 PFAM families lacking a protein structure, we produce models of forty-one families with estimated high accuracy.

Recently, an increasing number of studies have demonstrated that miRNAs are involved in human diseases, indicating that miRNAs might be a potential pathogenic factor for various diseases. Therefore, figuring out the relationship between miRNAs and diseases plays a critical role in not only the development of new drugs, but also the formulation of individualized diagnosis and treatment. As the prediction of miRNA-disease association via biological experiments is expensive and time-consuming, computational methods have a positive effect on revealing the association. In this study, a novel prediction model integrating GCN, CNN and Squeeze-and-Excitation Networks (GCSENet) was constructed for the identification of miRNA-disease association. The model first captured features by GCN based on a heterogeneous graph including diseases, genes and miRNAs. Then, considering the different effects of genes on each type of miRNA and disease, as well as the different effects of the miRNA-gene and disease-gene relationships on miRNA-disease association, a feature weight was set and a combination of miRNA-gene and disease-gene associations was added as feature input for the convolution operation in CNN. Furthermore, the squeeze and excitation blocks of SENet were applied to determine the importance of each feature channel and enhance useful features by means of the attention mechanism, thus achieving a satisfactory prediction of miRNA-disease association. The proposed method was compared against other state-of-the-art methods. It achieved an AUROC score of 95.02% and an AUPR score of 95.55% in a 10-fold cross-validation, which led to the finding that the proposed method is superior to these popular methods on most of the performance evaluation indexes.

CPA/AT transporters are made up of scaffold and a core domain. The core domain contains two non-canonical helices (broken or reentrant) that mediate the transport of ions, amino acids or other charged compounds. During evolution, these transporters have undergone substantial changes in structure, topology and function. To shed light on these structural transitions, we create models for all families using an integrated topology annotation method. We find that the CPA/AT transporters can be classified into four fold-types based on their structure; (1) the CPA-broken fold-type, (2) the CPA-reentrant fold-type, (3) the BART fold-type, and (4) a previously not described fold-type, the Reentrant-Helix-Reentrant fold-type. Several topological transitions are identified, including the transition between a broken and reentrant helix, one transition between a loop and a reentrant helix, complete changes of orientation, and changes in the number of scaffold helices. These transitions are mainly caused by gene duplication and shuffling events. Structural models, topology information and other details are presented in a searchable database, CPAfold (cpafold.bioinfo.se).

MOTIVATION: Identifying compound-protein interaction (CPI) is a crucial task in drug discovery and chemogenomics studies, and proteins without three-dimensional structure account for a large part of potential biological targets, which requires developing methods using only protein sequence information to predict CPI. However, sequence-based CPI models may face some specific pitfalls, including using inappropriate datasets, hidden ligand bias and splitting datasets inappropriately, resulting in overestimation of their prediction performance. RESULTS: To address these issues, we here constructed new datasets specific for CPI prediction, proposed a novel transformer neural network named TransformerCPI, and introduced a more rigorous label reversal experiment to test whether a model learns true interaction features. TransformerCPI achieved much improved performance on the new experiments, and it can be deconvolved to highlight important interacting regions of protein sequences and compound atoms, which may contribute chemical biology studies with useful guidance for further ligand structural optimization. AVAILABILITY AND IMPLEMENTATION: https://github.com/lifanchen-simm/transformerCPI.

MOTIVATION: Residue contact prediction was revolutionized recently by the introduction of direct coupling analysis (DCA). Further improvements, in particular for small families, have been obtained by the combination of DCA and deep learning methods. However, existing deep learning contact prediction methods often rely on a number of external programs and are therefore computationally expensive. RESULTS: Here, we introduce a novel contact predictor, PconsC4, which performs on par with state of the art methods. PconsC4 is heavily optimized, does not use any external programs and therefore is significantly faster and easier to use than other methods. AVAILABILITY AND IMPLEMENTATION: PconsC4 is freely available under the GPL license from https://github.com/ElofssonLab/PconsC4. Installation is easy using the pip command and works on any system with Python 3.5 or later and a GCC compiler. It does not require a GPU nor special hardware. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Intrinsic disorder is more abundant in eukaryotic than prokaryotic proteins.Methods predicting intrinsic disorder are based on the amino acid sequence of a protein.Therefore, there must exist an underlying difference in the sequences between eukaryotic and prokaryotic proteins causing the (predicted) difference in intrinsic disorder.By comparing proteins, from complete eukaryotic and prokaryotic proteomes, we show that the difference in intrinsic disorder emerges from the linker regions connecting Pfam domains.Eukaryotic proteins have more extended linker regions, and in addition, the eukaryotic linkers are significantly more disordered, 38% vs. 12-16% disordered residues.Next, we examined the underlying reason for the increase in disorder in eukaryotic linkers, and we found that the changes in abundance of only three amino acids cause the increase.Eukaryotic proteins contain 8.6% serine; while prokaryotic proteins have 6.5%, eukaryotic proteins also contain 5.4% proline and 5.3% isoleucine compared with 4.0% proline and � 7.5% isoleucine in the prokaryotes.All these three differences contribute to the increased disorder in eukaryotic proteins.It is tempting to speculate that the increase in serine frequencies in eukaryotes is related to regulation by kinases, but direct evidence for this is lacking.The differences are observed in all phyla, protein families, structural regions and type of protein but are most pronounced in disordered and linker regions.The observation that differences in the abundance of three amino acids cause the difference in disorder between eukaryotic and prokaryotic proteins raises the question: Are amino acid frequencies different in eukaryotic linkers because the linkers are more disordered or do the differences cause the increased disorder? Author SummaryIntrinsic disorder is essential for various functions in eukaryotic cells and is a signature of eukaryotic proteins.Here, we try to understand the origin of the difference in disorder between eukaryotic and prokaryotic proteins.We show that eukaryotic proteins contain more extended linker regions and that these linker regions are significantly more disordered.Further, we show, for the first time, that the difference in disorder originates from a systematic difference in amino acid frequencies between eukaryotic and prokaryotic

All PLOS journals have an open data policy that, amongst other things, states that all data and related metadata underlying the findings reported in a submitted manuscript should be deposited in an appropriate public repository, or for smaller datasets, as supporting information.This should obviously apply to computational methods as well, but unfortunately this is not always applied in practice, although it is of greatest importance for the scientific quality of simulations [1] and other modeling projects [2].Molecular dynamics [3] and other type of simulations [2,4] have become a fundamental part of life sciences.The simulations are dependent on a number of parameters such as force fields, initial configurations, simulation protocols, and software.Researchers have different opinions about the types of software they prefer, and in general, we believe authors should be free to choose the tools that best fit their needs.However, as scientists, we also have a common obligation to critically test each other's statements to find mistakes (including errors in the algorithms and bugs in the code), which can be exemplified by a heated debate over simulations of supercooled water that ended up being due to a subtle algorithmic issue [5], and we believe PLOS has a particularly strong responsibility to lead this development even if it might cause some short-term grief [6].In particular, all published results should, in principle, be possible to reproduce independently by scientists in other labs using different tools.To ensure this, we propose a set of standards that any publication in PLOS Computational Biology, and hopefully, publications in other journals as well, should follow.We do believe that the sooner such policies are widely adapted, the more open and collaborative science will flourish [7].These 10 simple rules should not be limited to molecular dynamics but also include Monte Carlo simulations, quantum mechanics calculations, molecular docking, and any other computational methods involving computations on biological molecules. Rule 1: The simulation protocol should be providedThe complete set of input files that are used in the simulations should be provided, either as supplementary material or preferably through a publicly available repository.

MOTIVATION: In the context of genome-wide association studies (GWAS), there is a variety of statistical techniques in order to conduct the analysis, but, in most cases, the underlying genetic model is usually unknown. Under these circumstances, the classical Cochran-Armitage trend test (CATT) is suboptimal. Robust procedures that maximize the power and preserve the nominal type I error rate are preferable. Moreover, performing a meta-analysis using robust procedures is of great interest and has never been addressed in the past. The primary goal of this work is to implement several robust methods for analysis and meta-analysis in the statistical package Stata and subsequently to make the software available to the scientific community. RESULTS: The CATT under a recessive, additive and dominant model of inheritance as well as robust methods based on the Maximum Efficiency Robust Test statistic, the MAX statistic and the MIN2 were implemented in Stata. Concerning MAX and MIN2, we calculated their asymptotic null distributions relying on numerical integration resulting in a great gain in computational time without losing accuracy. All the aforementioned approaches were employed in a fixed or a random effects meta-analysis setting using summary data with weights equal to the reciprocal of the combined cases and controls. Overall, this is the first complete effort to implement procedures for analysis and meta-analysis in GWAS using Stata. AVAILABILITY AND IMPLEMENTATION: A Stata program and a web-server are freely available for academic users at http://www.compgen.org/tools/GWAR. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Accurate contact predictions can be used for predicting the structure of proteins. Until recently these methods were limited to very big protein families, decreasing their utility. However, recent progress by combining direct coupling analysis with machine learning methods has made it possible to predict accurate contact maps for smaller families. To what extent these predictions can be used to produce accurate models of the families is not known. RESULTS: We present the PconsFold2 pipeline that uses contact predictions from PconsC3, the CONFOLD folding algorithm and model quality estimations to predict the structure of a protein. We show that the model quality estimation significantly increases the number of models that reliably can be identified. Finally, we apply PconsFold2 to 6379 Pfam families of unknown structure and find that PconsFold2 can, with an estimated 90% specificity, predict the structure of up to 558 Pfam families of unknown structure. Out of these, 415 have not been reported before. AVAILABILITY AND IMPLEMENTATION: Datasets as well as models of all the 558 Pfam families are available at http://c3.pcons.net/ . All programs used here are freely available. CONTACT: [email protected].

MOTIVATION: A few years ago it was shown that by using a maximum entropy approach to describe couplings between columns in a multiple sequence alignment it is possible to significantly increase the accuracy of residue contact predictions. For very large protein families with more than 1000 effective sequences the accuracy is sufficient to produce accurate models of proteins as well as complexes. Today, for about half of all Pfam domain families no structure is known, but unfortunately most of these families have at most a few hundred members, i.e. are too small for such contact prediction methods. RESULTS: To extend accurate contact predictions to the thousands of smaller protein families we present PconsC3, a fast and improved method for protein contact predictions that can be used for families with even 100 effective sequence members. PconsC3 outperforms direct coupling analysis (DCA) methods significantly independent on family size, secondary structure content, contact range, or the number of selected contacts. AVAILABILITY AND IMPLEMENTATION: PconsC3 is available as a web server and downloadable version at http://c3.pcons.net . The downloadable version is free for all to use and licensed under the GNU General Public License, version 2. At this site contact predictions for most Pfam families are also available. We do estimate that more than 4000 contact maps for Pfam families of unknown structure have more than 50% of the top-ranked contacts predicted correctly. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Knowledge of the correct protein subcellular localization is necessary for understanding the function of a protein. Unfortunately large-scale experimental studies are limited in their accuracy. Therefore, the development of prediction methods has been limited by the amount of accurate experimental data. However, recently large-scale experimental studies have provided new data that can be used to evaluate the accuracy of subcellular predictions in human cells. Using this data we examined the performance of state of the art methods and developed SubCons, an ensemble method that combines four predictors using a Random Forest classifier. RESULTS: SubCons outperforms earlier methods in a dataset of proteins where two independent methods confirm the subcellular localization. Given nine subcellular localizations, SubCons achieves an F1-Score of 0.79 compared to 0.70 of the second best method. Furthermore, at a FPR of 1% the true positive rate (TPR) is over 58% for SubCons compared to less than 50% for the best individual predictor. AVAILABILITY AND IMPLEMENTATION: SubCons is freely available as a webserver (http://subcons.bioinfo.se) and source code from https://bitbucket.org/salvatore_marco/subcons-web-server. The golden dataset as well is available from http://subcons.bioinfo.se/pred/download. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

SUMMARY: Protein quality assessment is a long-standing problem in bioinformatics. For more than a decade we have developed state-of-art predictors by carefully selecting and optimising inputs to a machine learning method. The correlation has increased from 0.60 in ProQ to 0.81 in ProQ2 and 0.85 in ProQ3 mainly by adding a large set of carefully tuned descriptions of a protein. Here, we show that a substantial improvement can be obtained using exactly the same inputs as in ProQ2 or ProQ3 but replacing the support vector machine by a deep neural network. This improves the Pearson correlation to 0.90 (0.85 using ProQ2 input features). AVAILABILITY AND IMPLEMENTATION: ProQ3D is freely available both as a webserver and a stand-alone program at http://proq3.bioinfo.se/. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

De novo creation of protein coding genes involves the formation of short ORFs from noncoding regions; some of these ORFs might then become fixed in the population. These orphan proteins need to, at the bare minimum, not cause serious harm to the organism, meaning that they should for instance not aggregate. Therefore, although the creation of short ORFs could be truly random, the fixation should be subjected to some selective pressure. The selective forces acting on orphan proteins have been elusive, and contradictory results have been reported. In Drosophila young proteins are more disordered than ancient ones, while the opposite trend is present in yeast. To the best of our knowledge no valid explanation for this difference has been proposed. To solve this riddle we studied structural properties and age of proteins in 187 eukaryotic organisms. We find that, with the exception of length, there are only small differences in the properties between proteins of different ages. However, when we take the GC content into account we noted that it could explain the opposite trends observed for orphans in yeast (low GC) and Drosophila (high GC). GC content is correlated with codons coding for disorder promoting amino acids. This leads us to propose that intrinsic disorder is not a strong determining factor for fixation of orphan proteins. Instead these proteins largely resemble random proteins given a particular GC level. During evolution the properties of a protein change faster than the GC level causing the relationship between disorder and GC to gradually weaken.

UNLABELLED: : Accurate topology prediction of transmembrane β-barrels is still an open question. Here, we present BOCTOPUS2, an improved topology prediction method for transmembrane β-barrels that can also identify the barrel domain, predict the topology and identify the orientation of residues in transmembrane β-strands. The major novelty of BOCTOPUS2 is the use of the dyad-repeat pattern of lipid and pore facing residues observed in transmembrane β-barrels. In a cross-validation test on a benchmark set of 42 proteins, BOCTOPUS2 predicts the correct topology in 69% of the proteins, an improvement of more than 10% over the best earlier method (BOCTOPUS) and in addition, it produces significantly fewer erroneous predictions on non-transmembrane β-barrel proteins. AVAILABILITY AND IMPLEMENTATION: BOCTOPUS2 webserver along with full dataset and source code is available at http://boctopus.bioinfo.se/ CONTACT: : [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: The translocon recognizes sufficiently hydrophobic regions of a protein and inserts them into the membrane. Computational methods try to determine what hydrophobic regions are recognized by the translocon. Although these predictions are quite accurate, many methods still fail to distinguish marginally hydrophobic transmembrane (TM) helices and equally hydrophobic regions in soluble protein domains. In vivo, this problem is most likely avoided by targeting of the TM-proteins, so that non-TM proteins never see the translocon. Proteins are targeted to the translocon by an N-terminal signal peptide. The targeting is also aided by the fact that the N-terminal helix is more hydrophobic than other TM-helices. In addition, we also recently found that the C-terminal helix is more hydrophobic than central helices. This information has not been used in earlier topology predictors. RESULTS: Here, we use the fact that the N- and C-terminal helices are more hydrophobic to develop a new version of the first-principle-based topology predictor, SCAMPI. The new predictor has two main advantages; first, it can be used to efficiently separate membrane and non-membrane proteins directly without the use of an extra prefilter, and second it shows improved performance for predicting the topology of membrane proteins that contain large non-membrane domains. AVAILABILITY AND IMPLEMENTATION: The predictor, a web server and all datasets are available at http://scampi.bioinfo.se/ CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: The PRED-TMBB method is based on Hidden Markov Models and is capable of predicting the topology of beta-barrel outer membrane proteins and discriminate them from water-soluble ones. Here, we present an updated version of the method, PRED-TMBB2, with several newly developed features that improve its performance. The inclusion of a properly defined end state allows for better modeling of the beta-barrel domain, while different emission probabilities for the adjacent residues in strands are used to incorporate knowledge concerning the asymmetric amino acid distribution occurring there. Furthermore, the training was performed using newly developed algorithms in order to optimize the labels of the training sequences. Moreover, the method is retrained on a larger, non-redundant dataset which includes recently solved structures, and a newly developed decoding method was added to the already available options. Finally, the method now allows the incorporation of evolutionary information in the form of multiple sequence alignments. RESULTS: The results of a strict cross-validation procedure show that PRED-TMBB2 with homology information performs significantly better compared to other available prediction methods. It yields 76% in correct topology predictions and outperforms the best available predictor by 7%, with an overall SOV of 0.9. Regarding detection of beta-barrel proteins, PRED-TMBB2, using just the query sequence as input, achieves an MCC value of 0.92, outperforming even predictors designed for this task and are much slower. AVAILABILITY AND IMPLEMENTATION: The method, along with all datasets used, is freely available for academic users at http://www.compgen.org/tools/PRED-TMBB2 CONTACT: [email protected].

MOTIVATION: Recently it has been shown that the quality of protein contact prediction from evolutionary information can be improved significantly if direct and indirect information is separated. Given sufficiently large protein families, the contact predictions contain sufficient information to predict the structure of many protein families. However, since the first studies contact prediction methods have improved. Here, we ask how much the final models are improved if improved contact predictions are used. RESULTS: In a small benchmark of 15 proteins, we show that the TM-scores of top-ranked models are improved by on average 33% using PconsFold compared with the original version of EVfold. In a larger benchmark, we find that the quality is improved with 15-30% when using PconsC in comparison with earlier contact prediction methods. Further, using Rosetta instead of CNS does not significantly improve global model accuracy, but the chemistry of models generated with Rosetta is improved. AVAILABILITY: PconsFold is a fully automated pipeline for ab initio protein structure prediction based on evolutionary information. PconsFold is based on PconsC contact prediction and uses the Rosetta folding protocol. Due to its modularity, the contact prediction tool can be easily exchanged. The source code of PconsFold is available on GitHub at https://www.github.com/ElofssonLab/pcons-fold under the MIT license. PconsC is available from http://c.pcons.net/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Given sufficient large protein families, and using a global statistical inference approach, it is possible to obtain sufficient accuracy in protein residue contact predictions to predict the structure of many proteins. However, these approaches do not consider the fact that the contacts in a protein are neither randomly, nor independently distributed, but actually follow precise rules governed by the structure of the protein and thus are interdependent. Here, we present PconsC2, a novel method that uses a deep learning approach to identify protein-like contact patterns to improve contact predictions. A substantial enhancement can be seen for all contacts independently on the number of aligned sequences, residue separation or secondary structure type, but is largest for β-sheet containing proteins. In addition to being superior to earlier methods based on statistical inferences, in comparison to state of the art methods using machine learning, PconsC2 is superior for families with more than 100 effective sequence homologs. The improved contact prediction enables improved structure prediction.

SUMMARY: Recently, several new contact prediction methods have been published. They use (i) large sets of multiple aligned sequences and (ii) assume that correlations between columns in these alignments can be the results of indirect interaction. These methods are clearly superior to earlier methods when it comes to predicting contacts in proteins. Here, we demonstrate that combining predictions from two prediction methods, PSICOV and plmDCA, and two alignment methods, HHblits and jackhmmer at four different e-value cut-offs, provides a relative improvement of 20% in comparison with the best single method, exceeding 70% correct predictions for one contact prediction per residue. AVAILABILITY: The source code for PconsC along with supplementary data is freely available at http://c.pcons.net/ CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

SUMMARY: Clustering methods are often needed for accurately assessing the quality of modeled protein structures. Recent blind evaluation of quality assessment methods in CASP10 showed that there is little difference between many different methods as far as ranking models and selecting best model are concerned. When comparing many models, the computational cost of the model comparison can become significant. Here, we present PconsD, a fast, stream-computing method for distance-driven model quality assessment that runs on consumer hardware. PconsD is at least one order of magnitude faster than other methods of comparable accuracy. AVAILABILITY: The source code for PconsD is freely available at http://d.pcons.net/. Supplementary benchmarking data are also available there. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Transmembrane β barrel proteins (TMBs) are found in the outer membrane of Gram-negative bacteria, chloroplast and mitochondria. They play a major role in the translocation machinery, pore formation, membrane anchoring and ion exchange. TMBs are also promising targets for antimicrobial drugs and vaccines. Given the difficulty in membrane protein structure determination, computational methods to identify TMBs and predict the topology of TMBs are important. RESULTS: Here, we present BOCTOPUS; an improved method for the topology prediction of TMBs by employing a combination of support vector machines (SVMs) and Hidden Markov Models (HMMs). The SVMs and HMMs account for local and global residue preferences, respectively. Based on a 10-fold cross-validation test, BOCTOPUS performs better than all existing methods, reaching a Q3 accuracy of 87%. Further, BOCTOPUS predicted the correct number of strands for 83% proteins in the dataset. BOCTOPUS might also help in reliable identification of TMBs by using it as an additional filter to methods specialized in this task. AVAILABILITY: BOCTOPUS is freely available as a web server at: http://boctopus.cbr.su.se/. The datasets used for training and evaluations are also available from this site.

MOTIVATION: Transmembrane β-barrels exist in the outer membrane of gram-negative bacteria as well as in chloroplast and mitochondria. They are often involved in transport processes and are promising antimicrobial drug targets. Structures of only a few β-barrel protein families are known. Therefore, a method that could automatically generate such models would be valuable. The symmetrical arrangement of the barrels suggests that an approach based on idealized geometries may be successful. RESULTS: Here, we present tobmodel; a method for generating 3D models of β-barrel transmembrane proteins. First, alternative topologies are obtained from the BOCTOPUS topology predictor. Thereafter, several 3D models are constructed by using different angles of the β-sheets. Finally, the best model is selected based on agreement with a novel predictor, ZPRED3, which predicts the distance from the center of the membrane for each residue, i.e. the Z-coordinate. The Z-coordinate prediction has an average error of 1.61 Å. Tobmodel predicts the correct topology for 75% of the proteins in the dataset which is a slight improvement over BOCTOPUS alone. More importantly, however, tobmodel provides a Cα template with an average RMSD of 7.24 Å from the native structure. AVAILABILITY: Tobmodel is freely available as a web server at: http://tobmodel.cbr.su.se/. The datasets used for training and evaluations are also available from this site.

UNLABELLED: State-of-the-art methods for topology of α-helical membrane proteins are based on the use of time-consuming multiple sequence alignments obtained from PSI-BLAST or other sources. Here, we examine if it is possible to use the consensus of topology prediction methods that are based on single sequences to obtain a similar accuracy as the more accurate multiple sequence-based methods. Here, we show that TOPCONS-single performs better than any of the other topology prediction methods tested here, but ~6% worse than the best method that is utilizing multiple sequence alignments. AVAILABILITY AND IMPLEMENTATION: TOPCONS-single is available as a web server from http://single.topcons.net/ and is also included for local installation from the web site. In addition, consensus-based topology predictions for the entire international protein index (IPI) is available from the web server and will be updated at regular intervals.

UNLABELLED: Multiple templates can often be used to build more accurate homology models than models built from a single template. Here we introduce PconsM, an automated protocol that uses multiple templates to build protein models. PconsM has been among the top-performing methods in the recent CASP experiments and consistently perform better than the single template models used in Pcons.net. In particular for the easier targets with many alternative templates with a high degree of sequence identity, quality is readily improved with a few percentages over the highest ranked model built on a single template. PconsM is available as an additional pipeline within the Pcons.net protein structure prediction server. AVAILABILITY AND IMPLEMENTATION: PconsM is freely available from http://pcons.net/.

SUMMARY: Kalign2 is one of the fastest and most accurate methods for multiple alignments. However, in contrast to other methods Kalign2 does not allow externally supplied position specific gap penalties. Here, we present a modification to Kalign2, KalignP, so that it accepts such penalties. Further, we show that KalignP using position specific gap penalties obtained from predicted secondary structures makes steady improvement over Kalign2 when tested on Balibase 3.0 as well as on a dataset derived from Pfam-A seed alignments. AVAILABILITY AND IMPLEMENTATION: KalignP is freely available at http://kalignp.cbr.su.se. The source code of KalignP is available under the GNU General Public License, Version 2 or later from the same website.

BACKGROUND: In water-soluble proteins it is energetically favorable to bury hydrophobic residues and to expose polar and charged residues. In contrast to water soluble proteins, transmembrane proteins face three distinct environments; a hydrophobic lipid environment inside the membrane, a hydrophilic water environment outside the membrane and an interface region rich in phospholipid head-groups. Therefore, it is energetically favorable for transmembrane proteins to expose different types of residues in the different regions. RESULTS: Investigations of a set of structurally determined transmembrane proteins showed that the composition of solvent exposed residues differs significantly inside and outside the membrane. In contrast, residues buried within the interior of a protein show a much smaller difference. However, in all regions exposed residues are less conserved than buried residues. Further, we found that current state-of-the-art predictors for surface area are optimized for one of the regions and perform badly in the other regions. To circumvent this limitation we developed a new predictor, MPRAP, that performs well in all regions. In addition, MPRAP performs better on complete membrane proteins than a combination of specialized predictors and acceptably on water-soluble proteins. A web-server of MPRAP is available at http://mprap.cbr.su.se/ CONCLUSION: By including complete a-helical transmembrane proteins in the training MPRAP is able to predict surface accessibility accurately both inside and outside the membrane. This predictor can aid in the prediction of 3D-structure, and in the identification of erroneous protein structures.

SUMMARY: SPOCTOPUS is a method for combined prediction of signal peptides and membrane protein topology, suitable for genome-scale studies. Its objective is to minimize false predictions of transmembrane regions as signal peptides and vice versa. We provide a description of the SPOCTOPUS algorithm together with a performance evaluation where SPOCTOPUS compares favorably with state-of-the-art methods for signal peptide and topology predictions. AVAILABILITY: SPOCTOPUS is available as a web server and both the source code and benchmark data are available for download at http://octopus.cbr.su.se/

MOTIVATION: As alpha-helical transmembrane proteins constitute roughly 25% of a typical genome and are vital parts of many essential biological processes, structural knowledge of these proteins is necessary for increasing our understanding of such processes. Because structural knowledge of transmembrane proteins is difficult to attain experimentally, improved methods for prediction of structural features of these proteins are important. RESULTS: OCTOPUS, a new method for predicting transmembrane protein topology is presented and benchmarked using a dataset of 124 sequences with known structures. Using a novel combination of hidden Markov models and artificial neural networks, OCTOPUS predicts the correct topology for 94% of the sequences. In particular, OCTOPUS is the first topology predictor to fully integrate modeling of reentrant/membrane-dipping regions and transmembrane hairpins in the topological grammar. AVAILABILITY: OCTOPUS is available as a web server at http://octopus.cbr.su.se.

Abstract Background Protein sequence alignment is one of the basic tools in bioinformatics. Correct alignments are required for a range of tasks including the derivation of phylogenetic trees and protein structure prediction. Numerous studies have shown that the incorporation of predicted secondary structure information into alignment algorithms improves their performance. Secondary structure predictors have to be trained on a set of somewhat arbitrarily defined states (e.g. helix, strand, coil), and it has been shown that the choice of these states has some effect on alignment quality. However, it is not unlikely that prediction of other structural features also could provide an improvement. In this study we use an unsupervised clustering method, the self-organizing map, to assign sequence profile windows to "structural states" and assess their use in sequence alignment. Results The addition of self-organizing map locations as inputs to a profile-profile scoring function improves the alignment quality of distantly related proteins slightly. The improvement is slightly smaller than that gained from the inclusion of predicted secondary structure. However, the information seems to be complementary as the two prediction schemes can be combined to improve the alignment quality by a further small but significant amount. Conclusion It has been observed in many studies that predicted secondary structure significantly improves the alignments. Here we have shown that the addition of self-organizing map locations can further improve the alignments as the self-organizing map locations seem to contain some information that is not captured by the predicted secondary structure.

Many proteins, especially in eukaryotes, contain tandem repeats of several domains from the same family. These repeats have a variety of binding properties and are involved in protein-protein interactions as well as binding to other ligands such as DNA and RNA. The rapid expansion of protein domain repeats is assumed to have evolved through internal tandem duplications. However, the exact mechanisms behind these tandem duplications are not well-understood. Here, we have studied the evolution, function, protein structure, gene structure, and phylogenetic distribution of domain repeats. For this purpose we have assigned Pfam-A domain families to 24 proteomes with more sensitive domain assignments in the repeat regions. These assignments confirmed previous findings that eukaryotes, and in particular vertebrates, contain a much higher fraction of proteins with repeats compared with prokaryotes. The internal sequence similarity in each protein revealed that the domain repeats are often expanded through duplications of several domains at a time, while the duplication of one domain is less common. Many of the repeats appear to have been duplicated in the middle of the repeat region. This is in strong contrast to the evolution of other proteins that mainly works through additions of single domains at either terminus. Further, we found that some domain families show distinct duplication patterns, e.g., nebulin domains have mainly been expanded with a unit of seven domains at a time, while duplications of other domain families involve varying numbers of domains. Finally, no common mechanism for the expansion of all repeats could be detected. We found that the duplication patterns show no dependence on the size of the domains. Further, repeat expansion in some families can possibly be explained by shuffling of exons. However, exon shuffling could not have created all repeats.

MOTIVATION: The success of the consensus approach to the protein structure prediction problem has led to development of several different consensus methods. Most of them only rely on a structural comparison of a number of different models. However, there are other types of information that might be useful such as the score from the server and structural evaluation. RESULTS: Pcons5 is a new and improved version of the consensus predictor Pcons. Pcons5 integrates information from three different sources: the consensus analysis, structural evaluation and the score from the fold recognition servers. We show that Pcons5 is better than the previous version of Pcons and that it performs better than using only the consensus analysis. In addition, we also present a version of Pmodeller based on Pcons5, which performs significantly better than Pcons5. AVAILABILITY: Pcons5 is the first Pcons version available as a standalone program from http://www.sbc.su.se/~bjorn/Pcons5. It should be easy to implement in local meta-servers.

BACKGROUND: Profile-profile methods have been used for some years now to detect and align homologous proteins. The best such methods use information from the background distribution of amino acids and substitution tables either when constructing the profiles or in the scoring. This makes the methods dependent on the quality and choice of substitution table as well as the construction of the profiles. Here, we introduce a novel method called ProfNet that is used to derive a profile-profile scoring function. The method optimizes the discrimination between scores of related and unrelated residues and it is fast and straightforward to use. This new method derives a scoring function that is mainly dependent on the actual alignment of residues from a training set, and it does not use any additional information about the background distribution. RESULTS: It is shown that ProfNet improves the discrimination of related and unrelated residues. Further it can be used to improve the alignment of distantly related proteins. CONCLUSION: The best performance is obtained using superfamily related proteins in the training of ProfNet, and a classifier that is related to the distance between the structurally aligned residues. The main difference between the new scoring function and a traditional profile-profile scoring function is that conserved residues on average score higher with the new function.

MOTIVATION: Consensus structure prediction methods (meta-predictors) have higher accuracy than individual structure prediction algorithms (their components). The goal for the development of the 3D-Jury system is to create a simple but powerful procedure for generating meta-predictions using variable sets of models obtained from diverse sources. The resulting protocol should help to improve the quality of structural annotations of novel proteins. RESULTS: The 3D-Jury system generates meta-predictions from sets of models created using variable methods. It is not necessary to know prior characteristics of the methods. The system is able to utilize immediately new components (additional prediction providers). The accuracy of the system is comparable with other well-tuned prediction servers. The algorithm resembles methods of selecting models generated using ab initio folding simulations. It is simple and offers a portable solution to improve the accuracy of other protein structure prediction protocols. AVAILABILITY: The 3D-Jury system is available via the Structure Prediction Meta Server (http://BioInfo.PL/Meta/) to the academic community. SUPPLEMENTARY INFORMATION: 3D-Jury is coupled to the continuous online server evaluation program, LiveBench (http://BioInfo.PL/LiveBench/)

BACKGROUND: T-cells are key players in regulating a specific immune response. Activation of cytotoxic T-cells requires recognition of specific peptides bound to Major Histocompatibility Complex (MHC) class I molecules. MHC-peptide complexes are potential tools for diagnosis and treatment of pathogens and cancer, as well as for the development of peptide vaccines. Only one in 100 to 200 potential binders actually binds to a certain MHC molecule, therefore a good prediction method for MHC class I binding peptides can reduce the number of candidate binders that need to be synthesized and tested. RESULTS: Here, we present a novel approach, SVMHC, based on support vector machines to predict the binding of peptides to MHC class I molecules. This method seems to perform slightly better than two profile based methods, SYFPEITHI and HLA_BIND. The implementation of SVMHC is quite simple and does not involve any manual steps, therefore as more data become available it is trivial to provide prediction for more MHC types. SVMHC currently contains prediction for 26 MHC class I types from the MHCPEP database or alternatively 6 MHC class I types from the higher quality SYFPEITHI database. The prediction models for these MHC types are implemented in a public web service available at http://www.sbc.su.se/svmhc/. CONCLUSIONS: Prediction of MHC class I binding peptides using Support Vector Machines, shows high performance and is easy to apply to a large number of MHC class I types. As more peptide data are put into MHC databases, SVMHC can easily be updated to give prediction for additional MHC class I types. We suggest that the number of binding peptides needed for SVM training is at least 20 sequences.

UNLABELLED: The Structure Prediction Meta Server offers a convenient way for biologists to utilize various high quality structure prediction servers available worldwide. The meta server translates the results obtained from remote services into uniform format, which are consequently used to request a jury prediction from a remote consensus server Pcons. AVAILABILITY: The structure prediction meta server is freely available at http://BioInfo.PL/meta/, some remote servers have however restrictions for non-academic users, which are respected by the meta server. SUPPLEMENTARY INFORMATION: Results of several sessions of the CAFASP and LiveBench programs for assessment of performance of fold-recognition servers carried out via the meta server are available at http://BioInfo.PL/services.html.

BACKGROUND: Prediction of protein structures is one of the fundamental challenges in biology today. To fully understand how well different prediction methods perform, it is necessary to use measures that evaluate their performance. Every two years, starting in 1994, the CASP (Critical Assessment of protein Structure Prediction) process has been organized to evaluate the ability of different predictors to blindly predict the structure of proteins. To capture different features of the models, several measures have been developed during the CASP processes. However, these measures have not been examined in detail before. In an attempt to develop fully automatic measures that can be used in CASP, as well as in other type of benchmarking experiments, we have compared twenty-one measures. These measures include the measures used in CASP3 and CASP2 as well as have measures introduced later. We have studied their ability to distinguish between the better and worse models submitted to CASP3 and the correlation between them. RESULTS: Using a small set of 1340 models for 23 different targets we show that most methods correlate with each other. Most pairs of measures show a correlation coefficient of about 0.5. The correlation is slightly higher for measures of similar types. We found that a significant problem when developing automatic measures is how to deal with proteins of different length. Also the comparisons between different measures is complicated as many measures are dependent on the size of the target. We show that the manual assessment can be reproduced to about 70% using automatic measures. Alignment independent measures, detects slightly more of the models with the correct fold, while alignment dependent measures agree better when selecting the best models for each target. Finally we show that using automatic measures would, to a large extent, reproduce the assessors ranking of the predictors at CASP3. CONCLUSIONS: We show that given a sufficient number of targets the manual and automatic measures would have given almost identical results at CASP3. If the intent is to reproduce the type of scoring done by the manual assessor in in CASP3, the best approach might be to use a combination of alignment independent and alignment dependent measures, as used in several recent studies.

Abstract Motivation: Evaluating the accuracy of predicted models is critical for assessing structure prediction methods. Because this problem is not trivial, a large number of different assessment measures have been proposed by various authors, and it has already become an active subfield of research (Moult et al. (1999) Critical assessment of methods of protein structure prediction (CASP): round III. Proteins , Suppl32–6). The CASP (Moult et al. (1997) Critical assessment of methods of proteins structure prediction (CASP): round II. Proteins , Suppl3, Dedicated Issue, Moult et al. (1999) Critical assessment of methods of protein structure prediction (CASP): round III. Proteins , Suppl32–6) and CAFASP (Fischer et al. (1999) CAFASP-1: critical assessment of fully automated structure prediction methods. Proteins , Suppl3, 209–217) prediction experiments have demonstrated that it has been difficult to choose one single, ‘best’ method to be used in the evaluation. Consequently, the CASP3 evaluation was carried out using an extensive set of especially developed numerical measures, coupled with human-expert intervention. As part of our efforts towards a higher level of automation in the structure prediction field, here we investigate the suitability of a fully automated, simple, objective, quantitative and reproducible method that can be used in the automatic assessment of models in the upcoming CAFASP2 experiment. Such a method should (a) produce one single number that measures the quality of a predicted model and (b) perform similarly to human-expert evaluations. Results: MaxSub is a new and independently developed method that further builds and extends some of the evaluation methods introduced at CASP3. MaxSub aims at identifying the largest subset of \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(C_{{\alpha}}\) \end{document}atoms of a model that superimpose ‘well’ over the experimental structure, and produces a single normalized score that represents the quality of the model. Because there exists no evaluation method for assessment measures of predicted models, it is not easy to evaluate how good our new measure is. Even though an exact comparison of MaxSub and the CASP3 assessment is not straightforward, here we use a test-bed extracted from the CASP3 fold-recognition models. A rough qualitative comparison of the performance of MaxSub vis-a-vis the human-expert assessment carried out at CASP3 shows that there is a good agreement for the more accurate models and for the better predicting groups. As expected, some differences were observed among the medium to poor models and groups. Overall, the top six predicting groups ranked using the fully automated MaxSub are also the top six groups ranked at CASP3. We conclude that MaxSub is a suitable method for the automatic evaluation of models. Availability: MaxSub is available at: http://www.cs.bgu.ac.il/~dfischer/MaxSub/MaxSub.html Contact: {nomsiew,dfischer}@cs.bgu.ac.il; [email protected]; [email protected] Supplementary Information: Full tables are available at: http://www.cs.bgu.ac.il/\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \({\sim}\) \end{document}dfischer/MaxSub/MaxSub.html CASP web site: http://PredictionCenter.llnl.gov/casp3/CAFASP web site: http://www.cs.bgu.ac.il/~dfischer/CAFASP2 To whom correspondence should be addressed.

MOTIVATION: Protein families can be defined based on structure or sequence similarity. We wanted to compare two protein family databases, one based on structural and one on sequence similarity, to investigate to what extent they overlap, the similarity in definition of corresponding families, and to create a list of large protein families with unknown structure as a resource for structural genomics. We also wanted to increase the sensitivity of fold assignment by exploiting protein family HMMs. RESULTS: We compared Pfam, a protein family database based on sequence similarity, to Scop, which is based on structural similarity. We found that 70% of the Scop families exist in Pfam while 57% of the Pfam families exist in Scop. Most families that occur in both databases correspond well to each other, but in some cases they are different. Such cases highlight situations in which structure and sequence approaches differ significantly. The comparison enabled us to compile a list of the largest families that do not occur in Scop; these are suitable targets for structure prediction and determination, and may be useful to guide projects in structural genomics. It can be noted that 13 out of the 20 largest protein families without a known structure are likely transmembrane proteins. We also exploited Pfam to increase the sensitivity of detecting homologs of proteins with known structure, by comparing query sequences to Pfam HMMs that correspond to Scop families. For SWISSPROT+TREMBL, this yielded an increase in fold assignment from 31% to 42% compared to using FASTA only. This method assigned a structure to 22% of the proteins in Saccharomyces cerevisiae, 24% in Escherichia coli, and 16% in Methanococcus jannaschii.