STING Report: convenient web-based application for graphic and tabular presentations of protein sequence, structure and function descriptors from the STING database

Study of specific nanoenvironments containing α-helices in all-α and (α+β)+(α/β) proteins

Protein secondary structure elements (PSSEs) such as α-helices, β-strands, and turns are the primary building blocks of the tertiary protein structure. Our primary interest here is to reveal the characteristics of the nanoenvironment formed by both PSSEs and their surrounding amino acid residues (AARs), which might contribute to the general understanding of how proteins fold. The characteristics of such nanoenvironments must be specific to each secondary structure element, and we have set our goal here to gather the fullest possible description of the α-helical nanoenvironment. In general, this postulate (the existence of specific nanoenvironments for specific protein substructures/neighbourhoods/regions with distinct functionality) was already successfully explored and confirmed for some protein regions, such as protein-protein interfaces and enzyme catalytic sites. Consequently, PSSEs were the obvious next choice for additional work for further evidence showing that specific nanoenvironments (having characteristics fully describable by means of structural and physical chemical descriptors) do exist for the corresponding and determined intraprotein regions. The nanoenvironment of α-helices (nEoαH) is defined as any region of the protein where this secondary structure element type is detected. The nEoαH, therefore, includes not only the α-helix amino acid residues but also the residues immediately around the α-helix. The hypothesis that motivated this work is that it might in fact be possible to detect a postulated "signal" or "signature" that distinguishes the specific location of α-helices. This "signal" must be discernible by tracking differences in the values of physical, chemical, physicochemical, structural and geometric descriptors immediately before (or after) the PSSE from those in the region along the α-helices. The search for this specific nanoenvironment "signal" was made possible by aligning previously selected α-helices of equal length. Afterward, we calculated the average value, standard deviation and mean square error at each aligned residue position for each selected descriptor. We applied Student's t-test, the Kolmogorov-Smirnov test and MANOVA statistical tests to the dataset constructed as described above, and the results confirmed that the hypothesized "signal"/"signature" is both existing/identifiable and capable of distinguishing the presence of an α-helix inside the specific nanoenvironment, contextualized as a specific region within the whole protein. However, such conclusion might rarely be reached if only one descriptor is considered at a time. A more accurate signal with broader coverage is achieved only if one applies multivariate analysis, which means that several descriptors (usually approximately 10 descriptors) should be considered at the same time. To a limited extent (up to a maximum of 15% of cases), such conclusion is also possible with only a single descriptor, and the conclusion is also possible in general for up to 50-80% of cases when no less than 5 nonlinear descriptors are selected and considered. Using all the descriptors considered in this work, provided all assumptions about data characteristics for this analysis are met, multivariate analysis regularly reached a coverage and accuracy above 90%. Understanding how secondary structure elements are formed and maintained within a protein structure could enable a more detailed understanding of how proteins reach their final 3D structure and consequently, their function. Likewise, this knowledge may also improve the tools used to determine how good a structure is by means of comparing the "signal" around a selected PSSE with the one obtained from the best (resolution and quality wise) protein structures available.

3D deep convolutional neural networks for amino acid environment similarity analysis

Background

Central to protein biology is the understanding of how structural elements give rise to observed function. The surfeit of protein structural data enables development of computational methods to systematically derive rules governing structural-functional relationships. However, performance of these methods depends critically on the choice of protein structural representation. Most current methods rely on features that are manually selected based on knowledge about protein structures. These are often general-purpose but not optimized for the specific application of interest.

In this paper, we present a general framework that applies 3D convolutional neural network (3DCNN) technology to structure-based protein analysis. The framework automatically extracts task-specific features from the raw atom distribution, driven by supervised labels. As a pilot study, we use our network to analyze local protein microenvironments surrounding the 20 amino acids, and predict the amino acids most compatible with environments within a protein structure. To further validate the power of our method, we construct two amino acid substitution matrices from the prediction statistics and use them to predict effects of mutations in T4 lysozyme structures.

Results

Our deep 3DCNN achieves a two-fold increase in prediction accuracy compared to models that employ conventional hand-engineered features and successfully recapitulates known information about similar and different microenvironments. Models built from our predictions and substitution matrices achieve an 85% accuracy predicting outcomes of the T4 lysozyme mutation variants. Our substitution matrices contain rich information relevant to mutation analysis compared to well-established substitution matrices. Finally, we present a visualization method to inspect the individual contributions of each atom to the classification decisions.

Conclusions

End-to-end trained deep learning networks consistently outperform methods using hand-engineered features, suggesting that the 3DCNN framework is well suited for analysis of protein microenvironments and may be useful for other protein structural analyses.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-017-1702-0) contains supplementary material, which is available to authorized users.

Discovery of small molecule inhibitors for the snake venom metalloprotease BaP1 using in silico and in vitro tests

Snakebites represent an important public health problem, with a great number of victims with permanent sequelae or fatal outcomes, particularly in rural, agriculturally active areas. The snake venom metalloproteases (SVMPs) are the principal proteins responsible for some clinically-relevant effects, such as local and systemic hemorrhage, dermonecrosis, and myonecrosis. Because of the difficulties in neutralizing them rapidly and locally by antivenoms, the search and design of small molecules as inhibitors of SVMPs are proposed. The Bothrops asper metalloprotease P1 (BaP1) is hereby used as a target protein and by High Throughput Virtual Screening (HTVS) approach, the free access virtual libraries: ZINC, PubChem and ChEMBL, were searched for potent small molecule inhibitors. Results from the aforementioned approaches provided strong evidences on the structural requirements for the efficient BaP1 inhibition such as the presence of the pyrimidine-2,4,6-trione moiety. The two proposed compounds have also shown excellent results in performed in vitro interaction studies against BaP1.

Interactions between Transmembrane Helices within Monomers of the Aquaporin AtPIP2;1 Play a Crucial Role in Tetramer Formation

Aquaporin (AQP) is a water channel protein found in various subcellular membranes of both prokaryotic and eukaryotic cells. The physiological functions of AQPs have been elucidated in many organisms. However, understanding their biogenesis remains elusive, particularly regarding how they assemble into tetramers. Here, we investigated the amino acid residues involved in the tetramer formation of the Arabidopsis plasma membrane AQP AtPIP2;1 using extensive amino acid substitution mutagenesis. The mutant proteins V41A/E44A, F51A/L52A, F87A/I91A, F92A/I93A, V95A/Y96A, and H216A/L217A, harboring alanine substitutions in the transmembrane (TM) helices of AtPIP2;1 polymerized into multiple oligomeric complexes with a variable number of subunits greater than four. Moreover, these mutant proteins failed to traffic to the plasma membrane, instead of accumulating in the endoplasmic reticulum (ER). Structure-based modeling revealed that these residues are largely involved in interactions between TM helices within monomers. These results suggest that inter-TM interactions occurring both within and between monomers play crucial roles in tetramer formation in the AtPIP2;1 complex. Moreover, the assembly of AtPIP2;1 tetramers is critical for their trafficking from the ER to the plasma membrane, as well as water permeability.

Improving predictions of protein-protein interfaces by combining amino acid-specific classifiers based on structural and physicochemical descriptors with their weighted neighbor averages

Protein-protein interactions are involved in nearly all regulatory processes in the cell and are considered one of the most important issues in molecular biology and pharmaceutical sciences but are still not fully understood. Structural and computational biology contributed greatly to the elucidation of the mechanism of protein interactions. In this paper, we present a collection of the physicochemical and structural characteristics that distinguish interface-forming residues (IFR) from free surface residues (FSR). We formulated a linear discriminative analysis (LDA) classifier to assess whether chosen descriptors from the BlueStar STING database (http://www.cbi.cnptia.embrapa.br/SMS/) are suitable for such a task. Receiver operating characteristic (ROC) analysis indicates that the particular physicochemical and structural descriptors used for building the linear classifier perform much better than a random classifier and in fact, successfully outperform some of the previously published procedures, whose performance indicators were recently compared by other research groups. The results presented here show that the selected set of descriptors can be utilized to predict IFRs, even when homologue proteins are missing (particularly important for orphan proteins where no homologue is available for comparative analysis/indication) or, when certain conformational changes accompany interface formation. The development of amino acid type specific classifiers is shown to increase IFR classification performance. Also, we found that the addition of an amino acid conservation attribute did not improve the classification prediction. This result indicates that the increase in predictive power associated with amino acid conservation is exhausted by adequate use of an extensive list of independent physicochemical and structural parameters that, by themselves, fully describe the nano-environment at protein-protein interfaces. The IFR classifier developed in this study is now integrated into the BlueStar STING suite of programs. Consequently, the prediction of protein-protein interfaces for all proteins available in the PDB is possible through STING_interfaces module, accessible at the following website: (http://www.cbi.cnptia.embrapa.br/SMS/predictions/index.html).

Identification of new sphingomyelinases D in pathogenic fungi and other pathogenic organisms

Sphingomyelinases D (SMases D) or dermonecrotic toxins are well characterized in Loxosceles spider venoms and have been described in some strains of pathogenic microorganisms, such as Corynebacterium sp. After spider bites, the SMase D molecules cause skin necrosis and occasional severe systemic manifestations, such as acute renal failure. In this paper, we identified new SMase D amino acid sequences from various organisms belonging to 24 distinct genera, of which, 19 are new. These SMases D share a conserved active site and a C-terminal motif. We suggest that the C-terminal tail is responsible for stabilizing the entire internal structure of the SMase D Tim barrel and that it can be considered an SMase D hallmark in combination with the amino acid residues from the active site. Most of these enzyme sequences were discovered from fungi and the SMase D activity was experimentally confirmed in the fungus Aspergillus flavus. Because most of these novel SMases D are from organisms that are endowed with pathogenic properties similar to those evoked by these enzymes alone, they might be associated with their pathogenic mechanisms.

Electrostatic potential studies as a consequence of cation-π interaction in cytochrome c fold of alpha proteins

Interactions between cationic and aromatic side chains of amino acid residues, the so-called cation-π interactions, are thought to contribute to the overall stability of the folded structure of peptides and proteins. We have analyzed the electrostatic behavior of residues involved in cation-π interactions for understanding the consequences of these non-covalent interactions. The average value of electrostatic potential for Arg and Lys were found to be positive which signifies their donor nature whereas Phe, Tyr and Trp showed negative values as they are acceptors. Similar trends were observed at the alpha carbon atom. We also observed that there is an opposite behavior of Lys as compared to Arg, Phe, Tyr and Trp towards electrostatic potential development on the last heavy atom. Furthermore the structural parameters like hydrophobicity and conservation score of interacting residues show that Lys to be acting totally different as compared to other residues and hence was found to be most influenced.

Analysis of binding properties and specificity through identification of the interface forming residues (IFR) for serine proteases in silico docked to different inhibitors

Background

Enzymes belonging to the same super family of proteins in general operate on variety of substrates and are inhibited by wide selection of inhibitors. In this work our main objective was to expand the scope of studies that consider only the catalytic and binding pocket amino acids while analyzing enzyme specificity and instead, include a wider category which we have named the Interface Forming Residues (IFR). We were motivated to identify those amino acids with decreased accessibility to solvent after docking of different types of inhibitors to sub classes of serine proteases and then create a table (matrix) of all amino acid positions at the interface as well as their respective occupancies. Our goal is to establish a platform for analysis of the relationship between IFR characteristics and binding properties/specificity for bi-molecular complexes.

Results

We propose a novel method for describing binding properties and delineating serine proteases specificity by compiling an exhaustive table of interface forming residues (IFR) for serine proteases and their inhibitors. Currently, the Protein Data Bank (PDB) does not contain all the data that our analysis would require. Therefore, an in silico approach was designed for building corresponding complexes

The IFRs are obtained by "rigid body docking" among 70 structurally aligned, sequence wise non-redundant, serine protease structures with 3 inhibitors: bovine pancreatic trypsin inhibitor (BPTI), ecotine and ovomucoid third domain inhibitor. The table (matrix) of all amino acid positions at the interface and their respective occupancy is created. We also developed a new computational protocol for predicting IFRs for those complexes which were not deciphered experimentally so far, achieving accuracy of at least 0.97.

Conclusions

The serine proteases interfaces prefer polar (including glycine) residues (with some exceptions). Charged residues were found to be uniquely prevalent at the interfaces between the "miscellaneous-virus" subfamily and the three inhibitors. This prompts speculation about how important this difference in IFR characteristics is for maintaining virulence of those organisms.

Our work here provides a unique tool for both structure/function relationship analysis as well as a compilation of indicators detailing how the specificity of various serine proteases may have been achieved and/or could be altered. It also indicates that the interface forming residues which also determine specificity of serine protease subfamily can not be presented in a canonical way but rather as a matrix of alternative populations of amino acids occupying variety of IFR positions.

The Interaction of Heparin/polyanions with Bovine, Porcine, and Human Growth Hormone**Sangeeta B. Joshi and Tim J. Kamerzell contributed equally to this work

The interaction of polyanions with proteins is of potential pharmaceutical and cellular significance. A partial thermodynamic description of the interaction of four representative polyanions with human, bovine, and porcine growth hormone is described. A heparin bead-binding assay confirms all growth hormones bind to heparin but to varying extents. Moderate-binding constants and high ratios of bound protein to the more extended polyanions, heparin, and dextran sulfate were measured by isothermal titration calorimetry and dynamic light scattering. The binding constants and ratio of protein bound to ligand were significantly smaller for the low molecular weight polyanions phytic acid and sucrose octasulfate (SOS). The effect of polyanion binding on the bovine, porcine, and human growth hormone's (hGH) structural and colloidal stability was also explored. Heparin and dextran sulfate inhibit porcine somatotropin (pST) and bovine somatotropin (bST) aggregation to the greatest extent, as compared to phytic acid and SOS, while decreasing secondary and tertiary structural stability as measured by the temperature dependence of their circular dichroism and intrinsic fluorescence. Somewhat surprisingly, the polyanions do not appear to affect the structure or stability of hGH. The potential biological significance of growth hormone polyanion interactions is discussed.

Flexible mapping of homology onto structure with homolmapper

Background

Over the past decade, a number of tools have emerged for the examination of homology relationships among protein sequences in a structural context. Most recent software implementations for such analysis are tied to specific molecular viewing programs, which can be problematic for collaborations involving multiple viewing environments. Incorporation into larger packages also adds complications for users interested in adding their own scoring schemes or in analyzing proteins incorporating unusual amino acid residues such as selenocysteine.

Results

We describe homolmapper, a command-line application for mapping information from a multiple protein sequence alignment onto a protein structure for analysis in the viewing software of the user's choice. Homolmapper is small (under 250 K for the application itself) and is written in Python to ensure portability. It is released for non-commercial use under a modified University of California BSD license. Homolmapper permits facile import of additional scoring schemes and can incorporate arbitrary additional amino acids to allow handling of residues such as selenocysteine or pyrrolysine. Homolmapper also provides tools for defining and analyzing subfamilies relative to a larger alignment, for mutual information analysis, and for rapidly visualizing the locations of mutations and multi-residue motifs.

Conclusion

Homolmapper is a useful tool for analysis of homology relationships among proteins in a structural context. There is also extensive, example-driven documentation available. More information about homolmapper is available at .

Molecular adaptation during adaptive radiation in the Hawaiian endemic genus Schiedea

BACKGROUND

"Explosive" adaptive radiations on islands remain one of the most puzzling evolutionary phenomena. The rate of phenotypic and ecological adaptations is extremely fast during such events, suggesting that many genes may be under fairly strong selection. However, no evidence for adaptation at the level of protein coding genes was found, so it has been suggested that selection may work mainly on regulatory elements. Here we report the first evidence that positive selection does operate at the level of protein coding genes during rapid adaptive radiations. We studied molecular adaptation in Hawaiian endemic plant genus Schiedea (Caryophyllaceae), which includes closely related species with a striking range of morphological and ecological forms, varying from rainforest vines to woody shrubs growing in desert-like conditions on cliffs. Given the remarkable difference in photosynthetic performance between Schiedea species from different habitats, we focused on the "photosynthetic" Rubisco enzyme, the efficiency of which is known to be a limiting step in plant photosynthesis.

RESULTS

We demonstrate that the chloroplast rbcL gene, encoding the large subunit of Rubisco enzyme, evolved under strong positive selection in Schiedea. Adaptive amino acid changes occurred in functionally important regions of Rubisco that interact with Rubisco activase, a chaperone which promotes and maintains the catalytic activity of Rubisco. Interestingly, positive selection acting on the rbcL might have caused favorable cytotypes to spread across several Schiedea species.

SIGNIFICANCE

We report the first evidence for adaptive changes at the DNA and protein sequence level that may have been associated with the evolution of photosynthetic performance and colonization of new habitats during a recent adaptive radiation in an island plant genus. This illustrates how small changes at the molecular level may change ecological species performance and helps us to understand the molecular bases of extremely fast rate of adaptation during island adaptive radiations.

The Diamond STING server

Diamond STING is a new version of the STING suite of programs for a comprehensive analysis of a relationship between protein sequence, structure, function and stability. We have added a number of new functionalities by both providing more structure parameters to the STING Database and by improving/expanding the interface for enhanced data handling. The integration among the STING components has also been improved. A new key feature is the ability of the STING server to handle local files containing protein structures (either modeled or not yet deposited to the Protein Data Bank) so that they can be used by the principal STING components: (Java)Protein Dossier ((J)PD) and STING Report. The current capabilities of the new STING version and a couple of biologically relevant applications are described here. We have provided an example where Diamond STING identifies the active site amino acids and folding essential amino acids (both previously determined by experiments) by filtering out all but those residues by selecting the numerical values/ranges for a set of corresponding parameters. This is the fundamental step toward a more interesting endeavor-the prediction of such residues. Diamond STING is freely accessible at http://sms.cbi.cnptia.embrapa.br and http://trantor.bioc.columbia.edu/SMS.