Search

BackgroundDespite the immense importance of transmembrane proteins (TMP) for molecular biology and medicine, experimental 3D structures for TMPs remain about 4-5 times underrepresented compared to non-TMPs. Today’s top methods such as AlphaFold2 accurately predict 3D structures for many TMPs, but annotating transmembrane regions remains a limiting step for proteome-wide predictions.ResultsHere, we present TMbed, a novel method inputting embeddings from protein Language Models (pLMs, here ProtT5), to predict for each residue one of four classes: transmembrane helix (TMH), transmembrane strand (TMB), signal peptide, or other. TMbed completes predictions for entire proteomes within hours on a single consumer-grade desktop machine at performance levels similar or better than methods, which are using evolutionary information from multiple sequence alignments (MSAs) of protein families. On the per-protein level, TMbed correctly identified 94±8% of the beta barrel TMPs (53 of 57) and 98±1% of the alpha helical TMPs (557 of 571) in a non-redundant data set, at false positive rates well below 1% (erred on 30 of 5654 non-membrane proteins). On the per-segment level, TMbed correctly placed, on average, 9 of 10 transmembrane segments within five residues of the experimental observation. Our method can handle sequences of up to 4200 residues on standard graphics cards used in desktop PCs (e.g., NVIDIA GeForce RTX 3060).ConclusionsBased on embeddings from pLMs and two novel filters (Gaussian and Viterbi), TMbed predicts alpha helical and beta barrel TMPs at least as accurately as any other method but at lower false positive rates. Given the few false positives and its outstanding speed, TMbed might be ideal to sieve through millions of 3D structures soon to be predicted, e.g., by AlphaFold2.AvailabilityOur code, method, and data sets are freely available in the GitHub repository, https://github.com/BernhoferM/TMbed.

The intricate details of how proteins bind to proteins, DNA, and RNA are crucial for the understanding of almost all biological processes. Disease-causing sequence variants often affect binding residues. Here, we described a new, comprehensive system of in silico methods that take only protein sequence as input to predict binding of protein to DNA, RNA, and other proteins. Firstly, we needed to develop several new methods to predict whether or not proteins bind (per-protein prediction). Secondly, we developed independent methods that predict which residues bind (per-residue). Not requiring three-dimensional information, the system can predict the actual binding residue. The system combined homology-based inference with machine learning and motif-based profile-kernel approaches with word-based (ProtVec) solutions to machine learning protein level predictions. This achieved an overall non-exclusive three-state accuracy of 77% ± 1% (±one standard error) corresponding to a 1.8 fold improvement over random (best classification for protein–protein with F1 = 91 ± 0.8%). Standard neural networks for per-residue binding residue predictions appeared best for DNA-binding (Q2 = 81 ± 0.9%) followed by RNA-binding (Q2 = 80 ± 1%) and worst for protein–protein binding (Q2 = 69 ± 0.8%). The new method, dubbed ProNA2020, is available as code through github ( https://github.com/Rostlab/ProNA2020.git ) and through PredictProtein ( www.predictprotein.org ).

Experimental structures are leveraged through multiple sequence alignments, or more generally through homology-based inference (HBI), facilitating the transfer of information from a protein with known annotation to a query without any annotation. A recent alternative expands the concept of HBI from sequence-distance lookup to embedding-based annotation transfer (EAT). These embeddings are derived from protein Language Models (pLMs). Here, we introduce using single protein representations from pLMs for contrastive learning. This learning procedure creates a new set of embeddings that optimizes constraints captured by hierarchical classifications of protein 3D structures defined by the CATH resource. The approach, dubbed ProtTucker, has an improved ability to recognize distant homologous relationships than more traditional techniques such as threading or fold recognition. Thus, these embeddings have allowed sequence comparison to step into the “midnight zone” of protein similarity, i.e., the region in which distantly related sequences have a seemingly random pairwise sequence similarity. The novelty of this work is in the particular combination of tools and sampling techniques that ascertained good performance comparable or better to existing state-of-the-art sequence comparison methods. Additionally, since this method does not need to generate alignments it is also orders of magnitudes faster. The code is available at https://github.com/Rostlab/EAT.

Advanced protein structure prediction requires evolutionary information from multiple sequence alignments (MSAs) from evolutionary couplings that are not always available. Artificial intelligence (AI)-based predictions inputting only single sequences are faster but so inaccurate as to render speed irrelevant. Here, we described a competitive prediction of inter-residue distances (2D structure) exclusively inputting embeddings from pre-trained protein language models (pLMs), namely ProtT5, from single sequences into a convolutional neural network (CNN) with relatively few layers. The major advance used the ProtT5 attention heads. Our new method, EMBER2, which never requires any MSAs, performed similarly to other methods that fully rely on co-evolution. Although clearly not reaching AlphaFold2, our leaner solution came somehow close at substantially lower costs. By generating protein-specific rather than family-averaged predictions, EMBER2 might better capture some features of particular protein structures. Results from using protein engineering and deep mutational scanning (DMS) experiments provided at least a proof of principle for such a speculation.

We modeled 3D structures of all SARS‐CoV‐2 proteins, generating 2,060 models that span 69% of the viral proteome and provide details not available elsewhere. We found that ˜6% of the proteome mimicked human proteins, while ˜7% was implicated in hijacking mechanisms that reverse post‐translational modifications, block host translation, and disable host defenses; a further ˜29% self‐assembled into heteromeric states that provided insight into how the viral replication and translation complex forms. To make these 3D models more accessible, we devised a structural coverage map, a novel visualization method to show what is—and is not—known about the 3D structure of the viral proteome. We integrated the coverage map into an accompanying online resource (https://aquaria.ws/covid) that can be used to find and explore models corresponding to the 79 structural states identified in this work. The resulting Aquaria‐COVID resource helps scientists use emerging structural data to understand the mechanisms underlying coronavirus infection and draws attention to the 31% of the viral proteome that remains structurally unknown or dark., 2,060 structural models spanning 69% of the SARS‐CoV‐2 proteome are generated and presented in a novel visual layout that summarises current knowledge on viral protein structures, and provides insight into viral replication, mimicry, and hijacking.

Visualizing protein data remains a challenging and stimulating task. Useful and intuitive visualization tools may help advance biomolecular and medical research; unintuitive tools may bar important breakthroughs. This protocol describes two use cases for the CellMap (http://cellmap.protein.properties) web tool. The tool allows researchers to visualize human protein-protein interaction data constrained by protein subcellular localizations. In the simplest form, proteins are visualized on cell images that also show protein-protein interactions (PPIs) through lines (edges) connecting the proteins across the compartments. At a glance, this simultaneously highlights spatial constraints that proteins are subject to in their physical environment and visualizes PPIs against these localizations. Visualizing two realities helps in decluttering the protein interaction visualization from "hairball" phenomena that arise when single proteins or groups thereof interact with hundreds of partners. © 2019 The Authors. Basic Protocol 1: Visualizing proteins and their interactions on cell images Basic Protocol 2: Displaying all interaction partners for a protein.

Binding small ligands such as ions or macromolecules such as DNA, RNA, and other proteins is one important aspect of the molecular function of proteins. Many binding sites remain without experimental annotations. Predicting binding sites on a per-residue level is challenging, but if 3D structures are known, information about coevolving residue pairs (evolutionary couplings) can predict catalytic residues through mutual information. Here, we predicted protein binding sites from evolutionary couplings derived from a global statistical model using maximum entropy. Additionally, we included information from sequence variation. A simple method using a weighted sum over eight scores substantially outperformed random (F1 = 19.3% ± 0.7% vs F1 = 2% for random). Training a neural network on these eight scores (along with predicted solvent accessibility and conservation in protein families) improved substantially (F1 = 26.2% ±0.8%). Although the machine learning was limited by the small data set and possibly wrong annotations of binding sites, the predicted binding sites formed spatial clusters in the protein. The source code of the binding site predictions is available through GitHub: https://github.com/Rostlab/bindPredict.

Transmembrane proteins (TMPs) are important drug targets because they are essential for signaling, regulation, and transport. Despite important breakthroughs, experimental structure determination remains challenging for TMPs. Various methods have bridged the gap by predicting transmembrane helices (TMHs), but room for improvement remains. Here, we present TMSEG, a novel method identifying TMPs and accurately predicting their TMHs and their topology. The method combines machine learning with empirical filters. Testing it on a non-redundant dataset of 41 TMPs and 285 soluble proteins, and applying strict performance measures, TMSEG outperformed the state-of-the-art in our hands. TMSEG correctly distinguished helical TMPs from other proteins with a sensitivity of 98 ± 2% and a false positive rate as low as 3 ± 1%. Individual TMHs were predicted with a precision of 87 ± 3% and recall of 84 ± 3%. Furthermore, in 63 ± 6% of helical TMPs the placement of all TMHs and their inside/outside topology was correctly predicted. There are two main features that distinguish TMSEG from other methods. First, the errors in finding all helical TMPs in an organism are significantly reduced. For example, in human this leads to 200 and 1600 fewer misclassifications compared to the second and third best method available, and 4400 fewer mistakes than by a simple hydrophobicity-based method. Second, TMSEG provides an add-on improvement for any existing method to benefit from. Proteins 2016; 84:1706-1716. © 2016 Wiley Periodicals, Inc.

Following publication of the original article [1], the author reported that an incorrect figure has been published as Figure 2. The correct Figure 2 is shown below.

Experimental structure determination continues to be challenging for membrane proteins. Computational prediction methods are therefore needed and widely used to supplement experimental data. Here, we re-examined the state of the art in transmembrane helix prediction based on a nonredundant dataset with 190 high-resolution structures. Analyzing 12 widely-used and well-known methods using a stringent performance measure, we largely confirmed the expected high level of performance. On the other hand, all methods performed worse for proteins that could not have been used for development. A few results stood out: First, all methods predicted proteins in eukaryotes better than those in bacteria. Second, methods worked less well for proteins with many transmembrane helices. Third, most methods correctly discriminated between soluble and transmembrane proteins. However, several older methods often mistook signal peptides for transmembrane helices. Some newer methods have overcome this shortcoming. In our hands, PolyPhobius and MEMSAT-SVM outperformed other methods.

Sulfur, most abundantly found in the environment as sulfate (SO42-), is an essential element in metabolites required by all living cells, including amino acids, co-factors and vitamins. However, current understanding of the cellular delivery of SO42- at the molecular level is limited. CysZ has been described as a SO42- permease, but its sequence family is without known structural precedent. Based on crystallographic structure information, SO42- binding and flux experiments, we provide insight into the molecular mechanism of CysZ-mediated translocation of SO42- across membranes. CysZ structures from three different bacterial species display a hitherto unknown fold and have subunits organized with inverted transmembrane topology. CysZ from Pseudomonas denitrificans assembles as a trimer of antiparallel dimers and the CysZ structures from two other species recapitulate dimers from this assembly. Mutational studies highlight the functional relevance of conserved CysZ residues.

Some mutations of protein residues matter more than others, and these are often conserved evolutionarily. The explosion of deep sequencing and genotyping increasingly requires the distinction between effect and neutral variants. The simplest approach predicts all mutations of conserved residues to have an effect; however, this works poorly, at best. Many computational tools that are optimized to predict the impact of point mutations provide more detail. Here, we expand the perspective from the view of single variants to the level of sketching the entire mutability landscape. This landscape is defined by the impact of substituting every residue at each position in a protein by each of the 19 non-native amino acids. We review some of the powerful conclusions about protein function, stability and their robustness to mutation that can be drawn from such an analysis. Large-scale experimental and computational mutagenesis experiments are increasingly furthering our understanding of protein function and of the genotype–phenotype associations. We also discuss how these can be used to improve predictions of protein function and pathogenicity of missense variants.

This study uses the Pfam database to show that the sequence redundancy of protein structures deposited in the PDB is increasing. The possible reasons behind this trend are discussed., High-resolution structural knowledge is key to understanding how proteins function at the molecular level. The number of entries in the Protein Data Bank (PDB), the repository of all publicly available protein structures, continues to increase, with more than 8000 structures released in 2012 alone. The authors of this article have studied how structural coverage of the protein-sequence space has changed over time by monitoring the number of Pfam families that acquired their first representative structure each year from 1976 to 2012. Twenty years ago, for every 100 new PDB entries released, an estimated 20 Pfam families acquired their first structure. By 2012, this decreased to only about five families per 100 structures. The reasons behind the slower pace at which previously uncharacterized families are being structurally covered were investigated. It was found that although more than 50% of current Pfam families are still without a structural representative, this set is enriched in families that are small, functionally uncharacterized or rich in problem features such as intrinsically disordered and transmembrane regions. While these are important constraints, the reasons why it may not yet be time to give up the pursuit of a targeted but more comprehensive structural coverage of the protein-sequence space are discussed.

Precision medicine and personalized health efforts propose leveraging complex molecular, medical and family history, along with other types of personal data toward better life. We argue that this ambitious objective will require advanced and specialized machine learning solutions. Simply skimming some low-hanging results off the data wealth might have limited potential. Instead, we need to better understand all parts of the system to define medically relevant causes and effects: how do particular sequence variants affect particular proteins and pathways? How do these effects, in turn, cause the health or disease-related phenotype? Toward this end, deeper understanding will not simply diffuse from deeper machine learning, but from more explicit focus on understanding protein function, context-specific protein interaction networks, and impact of variation on both.

Recent years have seen the establishment of structural genomics centers that explicitly target integral membrane proteins. Here, we review the advances in targeting these extremely high-hanging fruits of structural biology in high-throughput mode. We observe that the experimental determination of high-resolution structures of integral membrane proteins is increasingly successful both in terms of getting structures and of covering important protein families, for example, from Pfam. Structural genomics has begun to contribute significantly toward this progress. An important component of this contribution is the set up of robotic pipelines that generate a wealth of experimental data for membrane proteins. We argue that prediction methods for the identification of membrane regions and for the comparison of membrane proteins largely suffice to meet the challenges of target selection for structural genomics of membrane proteins. In contrast, we need better methods to prioritize the most promising members in a family of closely related proteins and to annotate protein function from sequence and structure in absence of homology.

The ribosome consists of small and large subunits each composed of dozens of proteins and RNA molecules. However, the functions of many of the individual protomers within the ribosome are still unknown. In this article, we describe the solution NMR structure of the ribosomal protein RP-L35Ae from the archaeon Pyrococcus furiosus. RP-L35Ae is buried within the large subunit of the ribosome and belongs to Pfam protein domain family PF01247, which is highly conserved in eukaryotes, present in a few archaeal genomes, but absent in bacteria. The protein adopts a six-stranded anti-parallel β-barrel analogous to the “tRNA binding motif” fold. The structure of the P. furiosus RP-L35Ae presented in this article constitutes the first structural representative from this protein domain family. Proteins 2012. © 2012 Wiley Periodicals, Inc.

As an operational definition, we refer to regions in proteins that do not adopt regular three-dimensional structures in isolation, as disordered regions. An antipode to disorder would be 'well-structured' rather than 'ordered'. Here, we argue for the following three hypotheses. Firstly, it is more useful to picture disorder as a distinct phenomenon in structural biology than as an extreme example of protein flexibility. Secondly, there are many very different flavors of protein disorder, nevertheless, it seems advantageous to portray the universe of all possible proteins in terms of two main types: well-structured, disordered. There might be a third type 'other' but we have so far no positive evidence for this. Thirdly, nature uses protein disorder as a tool to adapt to different environments. Protein disorder is evolutionarily conserved and this maintenance of disorder is highly nontrivial. Increasingly integrating protein disorder into the toolbox of a living cell was a crucial step in the evolution from simple bacteria to complex eukaryotes. We need new advanced computational methods to study this new milestone in the advance of protein biology.

Saccharides have a central role in the nutrition of all living organisms. Whereas several saccharide uptake systems are shared between the different phylogenetic kingdoms, the phosphoenolpyruvate-dependent phosphotransferase system exists almost exclusively in bacteria. This multi-component system includes an integral membrane protein EIIC that transports saccharides and assists in their phosphorylation. Here we present the crystal structure of an EIIC from Bacillus cereus that transports diacetylchitobiose. The EIIC is a homodimer, with an expansive interface formed between the amino-terminal halves of the two protomers. The carboxy-terminal half of each protomer has a large binding pocket that contains a diacetylchitobiose, which is occluded from both sides of the membrane with its site of phosphorylation near the conserved His250 and Glu334 residues. The structure shows the architecture of this important class of transporters, identifies the determinants of substrate binding and phosphorylation, and provides a framework for understanding the mechanism of sugar translocation.

Oxygenic photosynthesis is initiated by photosystem II (PSII) in the thylakoid membranes of plants, algae and cyanobacteria. PSII is a multi-subunit pigment-protein complex responsible for splitting water into oxygen gas, hydrogen ions and electrons transferred to electron acceptors during photosynthesis.1 Two homologous membrane-spanning proteins D1 (PsbA) and D2 (PsbD) form the PSII complex core.1 Peripherally, two chlorophyll (Chl)-binding inner antenna proteins CP47 (PsbB) and CP43 (PsbC) are bound to the D1-D2 PSII complex core.1 These four large proteins are surrounded by a large number of smaller membrane proteins.2 Most of these small proteins have been observed in the crystal structures of the PSII complex from cyanobacteria.3,4 However, one small protein, Psb28, previously detected as a nonstoichiometric component of PSII,5 was not observed in the crystal structures indicating that Psb28 might not be a true PSII subunit. Recent studies revealed that Psb28 was preferentially bound to PSII core complex lacking CP43 (RC47) and involved in the biogenesis of CP47.6 Understanding the association of Psb28 with the PSII core complex should provide additional insight into its role in PSII-mediated function. However, the structure of Psb28 has remained unknown up until now. In this note, we report the solution NMR structure of Psb28 protein encoded by gene sll1398 [gi|952386] of Synechocystis sp. strain PCC 6803 (SWISS-PROT ID: PSB28_SYNY3, NESG target ID: SgR171).7 This protein, also named Psb13 or ycf79, belongs to the Psb28 protein family (Pfam ID: PF03912), which is currently made up of ~48 protein sequences (E score less than 0.001 using PSI-BLAST, Table S1). Both PSI-BLAST sequence similarity and Dali8 structure similarity searches indicate that this is the first atomic resolution structure available for the Psb28 family. ConSurf9 was used to identify conserved surface residues potentially involved in binding to the PSII core complex.10

The New York Consortium on Membrane Protein Structure (NYCOMPS) was formed to accelerate the acquisition of structural information on membrane proteins by applying a structural genomics approach. NY-COMPS comprises a bioinformatics group, a centralized facility operating a high-throughput cloning and screening pipeline, a set of associated wet labs that perform high-level protein production and structure determination by x-ray crystallography and NMR, and a set of investigators focused on methods development. In the first three years of operation, the NYCOMPS pipeline has so far produced and screened 7,250 expression constructs for 8,045 target proteins. Approximately 600 of these verified targets were scaled up to levels required for structural studies, so far yielding 24 membrane protein crystals. Here we describe the overall structure of NYCOMPS and provide details on the high-throughput pipeline.

Lin0431 protein from Listeria innocua (UniProtKB/TrEMBL ID {"type":"entrez-protein","attrs":{"text":"Q92EM7","term_id":"81527521","term_text":"Q92EM7"}}Q92EM7/Q92EM7_LISIN) was selected as a target of the Northeast Structural Genomics Consortium (target ID: LkR112). Here, we present the high-quality NMR solution structure of this protein which is the first representative for a member of DUF1312 domain family. Lin0431 protein exhibits a β-sandwich topology. Four anti-parallel β-strands form one face of the sandwich and the other three anti-parallel β-strands together with a short α-helix form the other face of the sandwich. Structure alignment by Dali reveals an unexpected structural similarity with domain II of NusG from Aquifex aeolicus. Analyses of the electrostatic protein surface potential and searches for protein surface cavities reveal the conserved basic charged surface cavities of both the Lin0431 and domain II of AaeNusG, suggesting they may bind the negatively charged nucleic acids and/or and other binding partners. The high structural similarity and similar surface features, despite the lack of recognizable sequence similarity, between Lin0431 and AaeNusG domain II suggest that the domain II of NusG and DUF1312 domains have a homologous relationship and may share similar biochemical functions.

Protein with 83 residues, VPA0419 (residues 17–99, numbered 1–83) (gi|81726230, SwissProt/TrEMBL ID Q87J34_VIBPA, accession number {"type":"entrez-protein","attrs":{"text":"Q87J34","term_id":"81726230","term_text":"Q87J34"}}Q87J34)1 from Vibrio parahaemolyticus and 99-residue protein yiiS (gi|81722782, SwissProt/TrEMBL ID Q83IT9_SHIFL, accession number {"type":"entrez-protein","attrs":{"text":"Q83IT9","term_id":"81722782","term_text":"Q83IT9"}}Q83IT9)2, 3 from Shigella flexneri were selected as targets for the Protein Structure Initiative-2 and assigned to the Northeast Structural Genomics Consortium (NESG) for structure determination (NESG target ID VpR68 for VPA0419 and SfR90 for yiiS). VPA0419 and yiiS share 36% sequence identity, but show no significant sequence identity with any protein with known three-dimensional structure in the Protein Data Bank (PDB4). The two proteins belong to Pfam5 domain family PF04175 which currently contains 123 members with unknown three-dimensional structures and functional annotation, all of which appear to be found in gamma proteobacteria (for a sequence alignment, see Fig. S1 in Supporting Information). The NMR structures of VPA0419 and yiiS were solved using a protocol for high-throughput protein structure determination6 and represent the first ones for protein family PF04175. As these structures are the first for PF04175, “high leverage”7 of the experimental structures can be expected for calculating homology models.8, 9

The New York Consortium on Membrane Protein Structure (NYCOMPS), a part of the Protein Structure Initiative (PSI) in the USA, has as its mission to establish a high-throughput pipeline for determination of novel integral membrane protein structures. Here we describe our current target selection protocol, which applies structural genomics approaches informed by the collective experience of our team of investigators. We first extract all annotated proteins from our reagent genomes, i.e. the 96 fully sequenced prokaryotic genomes from which we clone DNA. We filter this initial pool of sequences and obtain a list of valid targets. NYCOMPS defines valid targets as those that, among other features, have at least two predicted transmembrane helices, no predicted long disordered regions and, except for community nominated targets, no significant sequence similarity in the predicted transmembrane region to any known protein structure. Proteins that feed our experimental pipeline are selected by defining a protein seed and searching the set of all valid targets for proteins that are likely to have a transmembrane region structurally similar to that of the seed. We require sequence similarity aligning at least half of the predicted transmembrane region of seed and target. Seeds are selected according to their feasibility and/or biological interest, and they include both centrally selected targets and community nominated targets. As of December 2008, over 6,000 targets have been selected and are currently being processed by the experimental pipeline. We discuss how our target list may impact structural coverage of the membrane protein space. Electronic supplementary material The online version of this article (doi:10.1007/s10969-009-9071-1) contains supplementary material, which is available to authorized users.

Summary One major objective of structural genomics efforts, including the NIH-funded Protein Structure Initiative (PSI), has been to increase the structural coverage of protein sequence space. Here, we present the target selection strategy used during the second phase of PSI (PSI-2). This strategy, jointly devised by the bioinformatics groups associated with the PSI-2 large-scale production centers, targets representatives from large, structurally uncharacterized protein domain families, and from structurally uncharacterized subfamilies in very large and diverse families with incomplete structural coverage. These very large families are extremely diverse both structurally and functionally, and are highly overrepresented in known proteomes. On the basis of several metrics, we then discuss to what extent PSI-2, during its first 3 years, has increased the structural coverage of genomes, and contributed structural and functional novelty. Together, the results presented here suggest that PSI-2 is successfully meeting its objectives and provides useful insights into structural and functional space.

The strategy for evaluating template-based models submitted to CASP has continuously evolved from CASP1 to CASP5, leading to a standard procedure that has been used in all subsequent editions. The established approach includes methods for calculating the quality of each individual model, for assigning scores based on the distribution of the results for each target and for computing the statistical significance of the differences in scores between prediction methods. These data are made available to the assessor of the template-based modeling category, who uses them as a starting point for further evaluations and analyses. This article describes the detailed workflow of the procedure, provides justifications for a number of choices that are customarily made for CASP data evaluation, and reports the results of the analysis of template-based predictions at CASP8. Proteins 2009. © 2009 Wiley-Liss, Inc.

For cell regulation, E2-like ubiquitin-fold modifier conjugating enzyme 1 (Ufc1) is involved in the transfer of ubiquitin-fold modifier 1 (Ufm1), a ubiquitin like protein which is activated by E1-like enzyme Uba5, to various target proteins. Thereby, Ufc1 participates in the very recently discovered Ufm1-Uba5-Ufc1 ubiquination pathway which is found in metazoan organisms. The structure of human Ufc1 was solved by using both NMR spectroscopy and X-ray crystallography. The complementary insights obtained with the two techniques provided a unique basis for understanding the function of Ufc1 at atomic resolution. The Ufc1 structure consists of the catalytic core domain conserved in all E2-like enzymes and an additional N-terminal helix. The active site Cys(116), which forms a thio-ester bond with Ufm1, is located in a flexible loop that is highly solvent accessible. Based on the Ufc1 and Ufm1 NMR structures, a model could be derived for the Ufc1-Ufm1 complex in which the C-terminal Gly(83) of Ufm1 may well form the expected thio-ester with Cys(116), suggesting that Ufm1-Ufc1 functions as described for other E1-E2-E3 machineries. alpha-helix 1 of Ufc1 adopts different conformations in the crystal and in solution, suggesting that this helix plays a key role to mediate specificity.

We report on several proteins recently solved by structural genomics consortia, in particular by the Northeast Structural Genomics consortium (NESG). The proteins considered in this study differ substantially in their sequences but they share a similar structural core, characterized by a pseudobarrel five-stranded beta sheet. This core corresponds to the PUA domain-like architecture in the SCOP database. By connecting sequence information with structural knowledge, we characterize a new subgroup of these proteins that we propose to be distinctly different from previously described PUA domain-like domains such as PUA proper or ASCH. We refer to these newly defined domains as EVE. Although EVE may have retained the ability of PUA domains to bind RNA, the available experimental and computational data suggests that both the details of its molecular function and its cellular function differ from those of other PUA domain-like domains. This study of EVE and its relatives illustrates how the combination of structure and genomics creates new insights by connecting a cornucopia of structures that map to the same evolutionary potential. Primary sequence information alone would have not been sufficient to reveal these evolutionary links.

Many important characteristics of proteins such as biochemical activity and subcellular localization present a challenge to machine-learning methods: it is often difficult to encode the appropriate input features at the residue level for the purpose of making a prediction for the entire protein. The problem is usually that the biophysics of the connection between a machine-learning method's input (sequence feature) and its output (observed phenomenon to be predicted) remains unknown; in other words, we may only know that a certain protein is an enzyme (output) without knowing which region may contain the active site residues (input). The goal then becomes to dissect a protein into a vast set of sequence-derived features and to correlate those features with the desired output. We introduce a framework that begins with a set of global sequence features and then vastly expands the feature space by generically encoding the coexistence of residue-based features. It is this combination of individual features, that is the step from the fractions of serine and buried (input space 20 + 2) to the fraction of buried serine (input space 20 ☆ 2) that implicitly shifts the search space from global feature inputs to features that can capture very local evidence such as a the individual residues of a catalytic triad. The vast feature space created is explored by a genetic algorithm (GA) paired with neural networks and support vector machines. We find that the GA is critical for selecting combinations of features that are neither too general resulting in poor performance, nor too specific, leading to overtraining. The final framework manages to effectively sample a feature space that is far too large for exhaustive enumeration. We demonstrate the power of the concept by applying it to prediction of protein enzymatic activity. Proteins 2009. © 2008 Wiley-Liss, Inc.

Solution NMR structure of the SOS response protein YnzC from Bacillus subtilis James M. Aramini,* Seema Sharma, Yuanpeng J. Huang, G. V. T. Swapna, Chi Kent Ho, Karishma Shetty, Kellie Cunningham, Li-Chung Ma, Li Zhao, Leah A. Owens, Mei Jiang, Rong Xiao, Jinfeng Liu, Michael C. Baran, Thomas B. Acton, Burkhard Rost, and Gaetano T. Montelione* 1Department of Molecular Biology and Biochemistry, Center for Advanced Biotechnology and Medicine, Rutgers

NMR structure of the peptidyl-tRNA hydrolase domain from Pseudomonas syringae expands the structural coverage of the hydrolysis domains of class 1 peptide chain release factors Kiran Kumar Singarapu, Rong Xiao, Thomas Acton, Burkhard Rost, Gaetano T. Montelione, and Thomas Szyperski* 1Department of Chemistry, State University of New York at Buffalo, Buffalo, New York 14260-3000 2Northeast Structural Genomics Consortium 3Department of Molecular Biology and Biochemistry, Center of Advanced Biotechnology and Medicine, Rutgers University, Piscataway, New Jersey 08854-5638 4Department of Biochemistry and Molecular Biophysics, Columbia University, New York, New York 10032

Solution NMR structure of Escherichia coli ytfP expands the structural coverage of the UPF0131 protein domain family James M. Aramini,* Yuanpeng J. Huang, G.V.T. Swapna, John R. Cort, P.K. Rajan, Rong Xiao, Ritu Shastry, Thomas B. Acton, Jinfeng Liu, Burkhard Rost, Michael A. Kennedy, and Gaetano T. Montelione* 1 Center for Advanced Biotechnology and Medicine, Department of Molecular Biology and Biochemistry, Rutgers University, Piscataway, New Jersey 08854 2Northeast Structural Genomics Consortium, Rutgers University, Piscataway, New Jersey 08854 3 Biological Sciences Division, Pacific Northwest National Laboratory, Richland, Washington 99352 4Department of Biochemistry and Molecular Biophysics, Columbia University, New York, New York 10032 5Department of Biochemistry, Robert Wood Johnson Medical School, University of Medicine and Dentistry of New Jersey, Piscataway, New Jersey 08854

This paper is an introduction to the supplemental issue of the journal PROTEINS, dedicated to the seventh CASP experiment to assess the state of the art in protein structure prediction. The paper describes the conduct of the experiment, the categories of prediction included, and outlines the evaluation and assessment procedures. Highlights are improvements in model accuracy relative to that obtainable from knowledge of a single best template structure; convergence of the accuracy of models produced by automatic servers toward that produced by human modeling teams; the emergence of methods for predicting the quality of models; and rapidly increasing practical applications of the methods.

Accurate predictions of metal-binding sites in proteins by using sequence as the only source of information can significantly help in the prediction of protein structure and function, genome annotation, and in the experimental determination of protein structure. Here, we introduce a method for identifying histidines and cysteines that participate in binding of several transition metals and iron complexes. The method predicts histidines as being in either of two states (free or metal bound) and cysteines in either of three states (free, metal bound, or in disulfide bridges). The method uses only sequence information by utilizing position-specific evolutionary profiles as well as more global descriptors such as protein length and amino acid composition. Our solution is based on a two-stage machine-learning approach. The first stage consists of a support vector machine trained to locally classify the binding state of single histidines and cysteines. The second stage consists of a bidirectional recurrent neural network trained to refine local predictions by taking into account dependencies among residues within the same protein. A simple finite state automaton is employed as a postprocessing in the second stage in order to enforce an even number of disulfide-bonded cysteines. We predict histidines and cysteines in transition-metal-binding sites at 73% precision and 61% recall. We observe significant differences in performance depending on the ligand (histidine or cysteine) and on the metal bound. We also predict cysteines participating in disulfide bridges at 86% precision and 87% recall. Results are compared to those that would be obtained by using expert information as represented by PROSITE motifs and, for disulfide bonds, to state-of-the-art methods.

This paper describes the outcome of a “Workshop on Biological Macromolecular Structure Models” held in November 2005 in which experimentalists and modelers discussed the best way to archive models of biological macromolecules.

Structural flexibility has been associated with various biological processes such as molecular recognition and catalytic activity. In silico studies of protein flexibility have attempted to characterize and predict flexible regions based on simple principles. B-values derived from experimental data are widely used to measure residue flexibility. Here, we present the most comprehensive large-scale analysis of B-values. We used this analysis to develop a neural network-based method that predicts flexible-rigid residues from amino acid sequence. The system uses both global and local information (i.e., features from the entire protein such as secondary structure composition, protein length, and fraction of surface residues, and features from a local window of sequence-consecutive residues). The most important local feature was the evolutionary exchange profile reflecting sequence conservation in a family of related proteins. To illustrate its potential, we applied our method to 4 different case studies, each of which related our predictions to aspects of function. The first 2 were the prediction of regions that undergo conformational switches upon environmental changes (switch II region in Ras) and the prediction of surface regions, the rigidity of which is crucial for their function (tunnel in propeller folds). Both were correctly captured by our method. The third study established that residues in active sites of enzymes are predicted by our method to have unexpectedly low B-values. The final study demonstrated how well our predictions correlated with NMR order parameters to reflect motion. Our method had not been set up to address any of the tasks in those 4 case studies. Therefore, we expect that this method will assist in many attempts at inferring aspects of function.

TenA (transcriptional enhancer A) has been proposed to function as a transcriptional regulator based on observed changes in gene-expression patterns when overexpressed in Bacillus subtilis. However, studies of the distribution of proteins involved in thiamine biosynthesis in different fully sequenced genomes have suggested that TenA may be an enzyme involved in thiamine biosynthesis, with a function related to that of the ThiC protein. The crystal structure of PF1337, the TenA homolog from Pyrococcus furiosus, is presented here. The protomer comprises a bundle of alpha-helices with a similar tertiary structure and topology to that of human heme oxygenase-1, even though there is no significant sequence homology. A solvent-sequestered cavity lined by phylogenetically conserved residues is found at the core of this bundle in PF1337 and this cavity is observed to contain electron density for 4-amino-5-hydroxymethyl-2-methylpyrimidine phosphate, the product of the ThiC enzyme. In contrast, the modestly acidic surface of PF1337 shows minimal levels of sequence conservation and a dearth of the basic residues that are typically involved in DNA binding in transcription factors. Without significant conservation of its surface properties, TenA is unlikely to mediate functionally important protein-protein or protein-DNA interactions. Therefore, the crystal structure of PF1337 supports the hypothesis that TenA homologs have an indirect effect in altering gene-expression patterns and function instead as enzymes involved in thiamine metabolism. journal: Acta Crystallographica Section D: Biological Crystallography content_type: research papers peer_reviewed: Yes review_process: Single blind received: 22 November 2004 accepted: 15 February 2005 published_online: 20 April 2005 supplementary_materials: This article has supplementary materials copyright: © 2005 International Union of Crystallography ispartof: Acta Crystallographica D vol:61 issue:Pt 5 pages:589-98 ispartof: location:United States status: published

Predicting the native subcellular compartment of a protein is an important step toward elucidating its function. Here we introduce LOCtree, a hierarchical system combining support vector machines (SVMs) and other prediction methods. LOCtree predicts the subcellular compartment of a protein by mimicking the mechanism of cellular sorting and exploiting a variety of sequence and predicted structural features in its input. Currently LOCtree does not predict localization for membrane proteins, since the compositional properties of membrane proteins significantly differ from those of non-membrane proteins. While any information about function can be used by the system, we present estimates of performance that are valid when only the amino acid sequence of a protein is known. When evaluated on a non-redundant test set, LOCtree achieved sustained levels of 74% accuracy for non-plant eukaryotes, 70% for plants, and 84% for prokaryotes. We rigorously benchmarked LOCtree in comparison to the best alternative methods for localization prediction. LOCtree outperformed all other methods in nearly all benchmarks. Localization assignments using LOCtree agreed quite well with data from recent large-scale experiments. Our preliminary analysis of a few entirely sequenced organisms, namely human (Homo sapiens), yeast (Saccharomyces cerevisiae), and weed (Arabidopsis thaliana) suggested that over 35% of all non-membrane proteins are nuclear, about 20% are retained in the cytosol, and that every fifth protein in the weed resides in the chloroplast.

This article is an introduction to the special issue of the journal Proteins, dedicated to the sixth CASP experiment to assess the state of the art in protein structure prediction. The article describes the conduct of the experiment and the categories of prediction included, and outlines the evaluation and assessment procedures. A brief summary of progress over the decade of CASP experiments is also provided.

The most reliable way to align two proteins of unknown structure is through sequence–profile and profile–profile alignment methods. If the structure for one of the two is known, fold recognition methods outperform purely sequence-based alignments. Here, we introduced a novel method that aligns generalised sequence and predicted structure profiles. Using predicted 1D structure (secondary structure and solvent accessibility) significantly improved over sequence-only methods, both in terms of correctly recognising pairs of proteins with different sequences and similar structures and in terms of correctly aligning the pairs. The scores obtained by our generalised scoring matrix followed an extreme value distribution; this yielded accurate estimates of the statistical significance of our alignments. We found that mistakes in 1D structure predictions correlated between proteins from different sequence-structure families. The impact of this surprising result was that our method succeeded in significantly out-performing sequence-only methods even without explicitly using structural information from any of the two. Since AGAPE also outperformed established methods that rely on 3D information, we made it available through http://www.predictprotein.org . If we solved the problem of CPU-time required to apply AGAPE on millions of proteins, our results could also impact everyday database searches.

A central goal of structural genom- ics is to experimentally determine representative structures for all protein families. At least 14 struc- tural genomics pilot projects are currently investi- gating the feasibility of high-throughput structure determination; the National Institutes of Health funded nine of these in the United States. Initiatives differ in the particular subset of "all families" on which they focus. At the NorthEast Structural Genomics consortium (NESG), we target eukaryotic protein domain families. The automatic target selec- tion procedure has three aims: 1) identify all protein domain families from currently five entirely se- quenced eukaryotic target organisms based on their sequence homology, 2) discard those families that can be modeled on the basis of structural informa- tion already present in the PDB, and 3) target representatives of the remaining families for struc- ture determination. To guarantee that all members of one family share a common foldlike region, we had to begin by dissecting proteins into structural domain-like regions before clustering. Our hierarchi- cal approach, CHOP, utilizing homology to PrISM, Pfam-A, and SWISS-PROT chopped the 103,796 eu- karyotic proteins/ORFs into 247,222 fragments. Of these fragments, 122,999 appeared suitable targets that were grouped into >27,000 singletons and >18,000 multifragment clusters. Thus, our results suggested that it might be necessary to determine >40,000 structures to minimally cover the subset of five eukaryotic proteomes. Proteins 2004;56:188 -200.

Non-covalent residue side-chain interactions occur in many different types of proteins and facilitate many biological functions. Are these differences manifested in the sequence compositions and/or the residue-residue contact preferences of the interfaces? Previous studies analysed small data sets and gave contradictory answers. Here, we introduced a new data-mining method that yielded the largest high-resolution data set of interactions analysed. We introduced an information theory-based analysis method. On the basis of sequence features, we were able to differentiate six types of protein interfaces, each corresponding to a different functional or structural association between residues. Particularly, we found significant differences in amino acid composition and residue-residue preferences between interactions of residues within the same structural domain and between different domains, between permanent and transient interfaces, and between interactions associating homo-oligomers and hetero-oligomers. The differences between the six types were so substantial that, using amino acid composition alone, we could predict statistically to which of the six types of interfaces a pool of 1000 residues belongs at 63-100% accuracy. All interfaces differed significantly from the background of all residues in SWISS-PROT, from the group of surface residues, and from internal residues that were not involved in non-trivial interactions. Overall, our results suggest that the interface type could be predicted from sequence and that interface-type specific mean-field potentials may be adequate for certain applications.

We have analysed fold recogni- tion, secondary structure and contact prediction servers from CAFASP3. This assessment was car- ried out in the framework of the fully automated, web-based evaluation server EVA. Detailed results are available at http://cubic.bioc.columbia.edu/eva/ cafasp3/. We observed that the sequence-unique tar- gets from CAFASP3/CASP5 were not fully represen- tative for evaluating performance. For all three categories, we showed how careless ranking might be misleading. We compared methods from all cat- egories to experts in secondary structure and con- tact prediction and homology modellers to fold recognisers. While the secondary structure experts clearly outperformed all others, the contact experts appeared to outperform only novel fold methods. Automatic evaluation servers are good at getting statistics right and at using these to discard mislead- ing ranking schemes. We challenge that to let ma- chines rule where they are best might be the best way for the community to enjoy the tremendous benefit of CASP as a unique opportunity for brain- storming. Proteins 2003;53:548 -560.

The genomes of over 60 organisms from all three kingdoms of life are now entirely sequenced. In many respects, the inventory of proteins used in different kingdoms appears surprisingly similar. However, eukaryotes differ from other kingdoms in that they use many long proteins, and have more proteins with coiled-coil helices and with regions abundant in regular secondary structure. Particular structural domains are used in many pathways. Nevertheless, one domain tends to occur only once in one particular pathway. Many proteins do not have close homologues in different species (orphans) and there could even be folds that are specific to one species. This view implies that protein fold space is discrete. An alternative model suggests that structure space is continuous and that modern proteins evolved by aggregating fragments of ancient proteins. Either way, after having harvested proteomes by applying standard tools, the challenge now seems to be to develop better methods for comparative proteomics.

The level of sequence similarity that implies similarity in protein structure is well established. Recently, many groups proposed thresholds for similarity in sequence implying similarity in enzymatic function. All previous results suggest the strong conservation of enzymatic function above levels of 50% pairwise sequence identity. Here, I argue that all groups substantially overestimated the conservation of enzyme function because their data sets were either too biased, or too small. An unbiased analysis suggested that less than 30% of the pair fragments above 50% sequence identity have entirely identical EC numbers. Another surprising finding was that even BLAST E-values below 10(-50) did not suffice to automatically transfer enzyme function without errors. As expected, most misclassifications originated from similarities in relatively short regions and/or from transferring annotations for different domains. Both problems cannot be corrected easily by adjusting the thresholds for automatic transfer of genome annotations. A score relating sequence identity to alignment length (distance from HSSP-threshold) outperformed statistical BLAST scores for high sequence similarity. In particular, the distance score allowed error-free transfer of enzyme function for the 10% most similar enzyme pairs. The results illustrated how difficult it is to assess the conservation of protein function and to guarantee error-free genome annotations, in general: sets with millions of pair comparisons might not suffice to arrive at statistically significant conclusions. In practice, the revised detailed estimates for the sequence conservation of enzyme function may provide important benchmarks for everyday sequence analysis and for more cautious automatic genome annotations.

Secondary structure predictions are increasingly becoming the workhorse for several methods aiming at predicting protein structure and function. Here we use ensembles of bidirectional recurrent neural network architectures, PSI-BLAST-derived profiles, and a large nonredundant training set to derive two new predictors: (a) the second version of the SSpro program for secondary structure classification into three categories and (b) the first version of the SSpro8 program for secondary structure classification into the eight classes produced by the DSSP program. We describe the results of three different test sets on which SSpro achieved a sustained performance of about 78% correct prediction. We report confusion matrices, compare PSI-BLAST to BLAST-derived profiles, and assess the corresponding performance improvements. SSpro and SSpro8 are implemented as web servers, available together with other structural feature predictors at: http://promoter.ics.uci.edu/BRNN-PRED/. Proteins 2002;47:228–235. © 2002 Wiley-Liss, Inc.

The reliability of ranking of protein structure modeling methods is assessed. The assessment is based on the parametric Student's t test and the nonparametric Wilcox signed rank test of statistical significance of the difference between paired samples. The approach is applied to the ranking of the comparative modeling methods tested at the fourth meeting on Critical Assessment of Techniques for Protein Structure Prediction (CASP). It is shown that the 14 CASP4 test sequences may not be sufficient to reliably distinguish between the top eight methods, given the model quality differences and their standard deviations. We suggest that CASP needs to be supplemented by an assessment of protein structure prediction methods that is automated, continuous in time, based on several criteria applied to a large number of models, and with quantitative statistical reliability assigned to each characterization.

Protein families are a rich source of information; sequence conservation and sequence correlation are two of the main properties that can be derived from the analysis of multiple sequence alignments. Sequence conservation is related to the direct evolutionary pressure to retain the chemical characteristics of some positions in order to maintain a given function. Sequence correlation is attributed to the small sequence adjustments needed to maintain protein stability against constant mutational drift. Here, we showed that sequence conservation and correlation were each frequently informative enough to detect incorrectly folded proteins. Furthermore, combining conservation, correlation, and polarity, we achieved an almost perfect discrimination between native and incorrectly folded proteins. Thus, we made use of this information for threading by evaluating the models suggested by a threading method according to the degree of proximity of the corresponding correlated, conserved, and apolar residues. The results showed that the fold recognition capacity of a given threading approach could be improved almost fourfold by selecting the alignments that score best under the three different sequence-based approaches.