Large Structural Databases Research Articles

Predicting the functional state of protein kinases using interpretable graph neural networks from sequence and structural data.

Protein kinases are central to cellular activities and are actively pursued as drug targets for several conditions including cancer and autoimmune diseases. Despite the availability of a large structural database for kinases, methodologies to elucidate the structure-function relationship of these proteins (without manual intervention) are lacking. Such techniques are essential in structural biology and to accelerate drug discovery efforts. Here, we implement an interpretable graph neural network (GNN) framework for classifying the functionally active and inactive states of a large set of protein kinases by only using their tertiary structure and amino acid sequence. We show that the GNN models can classify kinase structures with high accuracy (>97%). We implement the Gradient-weighted Class Activation Mapping for graphs (Graph Grad-CAM) to automatically identify structurally important residues and residue-residue contacts of the kinases without any a priori input. We show that the motifs identified through the Graph Grad-CAM methodology are functionally critical, consistent with the existing kinase literature. Notably, the highly conserved DFG and HRD motifs of the well-known hydrophobic spine are identified by the interpretable framework in addition to some of the lesser known motifs. Further, using Grad-CAM maps as the vector embedding of the protein structures, we identify the subtle differences in the crystal structures among different sub-classes of kinases in the Protein Data Bank (PDB). Frameworks such as the one implemented here, for high-throughput identification of protein structure-function relationships are essential in designing targeted small molecules therapies as well as in engineering new proteins for novel applications.

Analyzing of metal organic frameworks performance in CH4 adsorption using machine learning techniques: A GBRT model based on small training dataset

Metal organic frameworks (MOFs) have already been widely used in the field of gas adsorption and separation. The combination of various organic ligands and metal ions results in large MOF libraries with different topological structures. However, such large structural databases make it difficult to screen out MOF materials with high adsorption performance towards a given gas to establish statistical pictures through experiments or molecular simulations. In this work, we studied and built machine learning (ML) models based on the hypothetical MOF (hMOF) database. A gradient-boosted regression trees (GBRT) model with 1000 samples can predict the methane adsorption capacity of more than 130,000 hMOFs with high accuracy. A fast high-throughput screening on the entire database was performed using the constructed regression model, which extracts several molecular fragments from the top 1% of materials with the largest methane adsorption capacity. Careful analysis of the chemical structure of these fragments indicates that MOF materials containing cyano, hydroxyl, carboxyl and/or aromatic rings may possess high adsorption performance. The electrostatic potential distribution of these fragments was obtained at the ab initio computational level and was highlighted to be another important descriptor. We further extended the GBRT predictor to diverse MOF libraries including tobacco MOF (tobmof) and CoRE MOF. The molecular fragments extracted from the ML-based screening on the real CoRE MOF database as well as the corresponding electrostatic properties were compared with the results of hMOFs, which provides new insight into the design of MOF materials.

In Silico Design of New Dual Inhibitors of SARS-CoV-2 MPRO through Ligand- and Structure-Based Methods.

The viral main protease is one of the most attractive targets among all key enzymes involved in the life cycle of SARS-CoV-2. Considering its mechanism of action, both the catalytic and dimerization regions could represent crucial sites for modulating its activity. Dual-binding the SARS-CoV-2 main protease inhibitors could arrest the replication process of the virus by simultaneously preventing dimerization and proteolytic activity. To this aim, in the present work, we identified two series' of small molecules with a significant affinity for SARS-CoV-2 MPRO, by a hybrid virtual screening protocol, combining ligand- and structure-based approaches with multivariate statistical analysis. The Biotarget Predictor Tool was used to filter a large in-house structural database and select a set of benzo[b]thiophene and benzo[b]furan derivatives. ADME properties were investigated, and induced fit docking studies were performed to confirm the DRUDIT prediction. Principal component analysis and docking protocol at the SARS-CoV-2 MPRO dimerization site enable the identification of compounds 1b,c,i,l and 2i,l as promising drug molecules, showing favorable dual binding site affinity on SARS-CoV-2 MPRO.

Correlations between crack initiation and crack propagation lives of notched specimens under constant and variable amplitude loading

AbstractThis paper starts with an overview of the application of the three guidelines (GL) of the German Research Association of Mechanical Engineers (FKM). Each of these provides algorithms for calculating fatigue lives of components under constant or variable amplitude loading, however, with underlying different failure criteria, that is, technical crack initiation life (GL‐nonlinear), fatigue crack growth life (GL‐fracture mechanics), and total fracture life (GL‐linear). Since the GL‐linear only considers linear‐elastic material behavior in the fracture fatigue life calculation, the guideline is not capable for the LCF and is not able to take sequence effects into account. For a better approximation of fracture fatigue lives, the elastic–plastic material behavior has to be taken into account. This paper introduces the U‐Concept which has been evaluated from a large structural durability database from the literature with more than 1,500 experiments for constant and more than 700 experiments for variable amplitude loading. The U‐Concept is a small add‐on to the local strain approach (LSA) which is the backbone of the GL‐nonlinear. It enables (1) to directly calculate the fatigue life to total fracture based on elastic–plastic material behavior according to the LSA or (2) to estimate the remaining fatigue life from crack initiation to fracture without a crack growth simulation. Furthermore, the U‐Concept needs no experimental determined crack growth parameters, the only necessary material parameter is the ultimate tensile strength Rm. The U‐Concept is therefore easy applicable for industrial users and increases the accuracy of the approximated fatigue lives by considering the elastic–plastic material behavior.

Comparative evaluation of shape retrieval methods on macromolecular surfaces: an application of computer vision methods in structural bioinformatics.

MotivationThe investigation of the structure of biological systems at the molecular level gives insights about their functions and dynamics. Shape and surface of biomolecules are fundamental to molecular recognition events. Characterizing their geometry can lead to more adequate predictions of their interactions. In the present work, we assess the performance of reference shape retrieval methods from the computer vision community on protein shapes.ResultsShape retrieval methods are efficient in identifying orthologous proteins and tracking large conformational changes. This work illustrates the interest for the protein surface shape as a higher-level representation of the protein structure that (i) abstracts the underlying protein sequence, structure or fold, (ii) allows the use of shape retrieval methods to screen large databases of protein structures to identify surficial homologs and possible interacting partners and (iii) opens an extension of the protein structure–function paradigm toward a protein structure-surface(s)-function paradigm.Availabilityand implementationAll data are available online at http://datasetmachat.drugdesign.fr.Supplementary information Supplementary data are available at Bioinformatics online.

A Database of Porous Rigid Amorphous Materials

Atomically detailed simulations have become an important complement to experimental studies of nanoporous materials. Large databases of crystalline porous material structures have been available fo...

ChemEnv: a fast and robust coordination environment identification tool.

Coordination or local environments have been used to describe, analyze and understand crystal structures for more than a century. Here, a new tool called ChemEnv, which can identify coordination environments in a fast and robust manner, is presented. In contrast to previous tools, the assessment of the coordination environments is not biased by small distortions of the crystal structure. Its robust and fast implementation enables the analysis of large databases of structures. The code is available open source within the pymatgen package and the software can also be used through a web app available on http://crystaltoolkit.org through the Materials Project.

Searching for electronically two dimensional metals in high-throughput ab initio databases

Due to the high degree of improvement and prediction success of ab initio calculations, the advance in Materials Science has been partially shifted from the empirical chemical quest of new compounds to the systematic calculation of putative structures that can yield compounds with promising properties. In particular, two dimensional (2D) materials are at present a central target, as many potential devices rely on particular characteristics of a single layer of atoms or low dimensional electronic active sheets in bulk materials. Several works have focused on this paradigm, yielding a large list of new interesting 2D candidates. Here, we introduce a technique based on an algorithmic filter for searching prospective candidates in very large electronic structure databases. It is designed to detect electronic bands with low energy dispersion along particular directions in reciprocal space, i.e. those perpendicular to certain plane or surface. We propose it as an absolute criterion for obtaining all possible two dimensional compounds. With it we show that previous explorations have missed out a whole group of possible two dimensional materials, namely only electronically 2D.

ALEPH: a network-oriented approach for the generation of fragment-based libraries and for structure interpretation.

The analysis of large structural databases reveals general features and relationships among proteins, providing useful insight. A different approach is required to characterize ubiquitous secondary-structure elements, where flexibility is essential in order to capture small local differences. The ALEPH software is optimized for the analysis and the extraction of small protein folds by relying on their geometry rather than on their sequence. The annotation of the structural variability of a given fold provides valuable information for fragment-based molecular-replacement methods, in which testing alternative model hypotheses can succeed in solving difficult structures when no homology models are available or are successful. ARCIMBOLDO_BORGES combines the use of composite secondary-structure elements as a search model with density modification and tracing to reveal the rest of the structure when both steps are successful. This phasing method relies on general fold libraries describing variations around a given pattern of β-sheets and helices extracted using ALEPH. The program introduces characteristic vectors defined from the main-chain atoms as a way to describe the geometrical properties of the structure. ALEPH encodes structural properties in a graph network, the exploration of which allows secondary-structure annotation, decomposition of a structure into small compact folds, generation of libraries of models representing a variation ofa given fold and finally superposition of these folds onto a target structure. These functions are available through a graphical interface designed to interactively show the results of structure manipulation, annotation, fold decomposition, clustering and library generation. ALEPH can produce pictures of the graphs, structures and folds for publication purposes.

The effect of rare-earth substitution on the Debye temperature of inorganic phosphors

In the quest to predict photoluminescent efficiency in rare-earth substituted inorganic phosphors, research has shown that materials consisting of ordered, rigid crystal structures tend to possess the highest photoluminescent quantum yields. A compound's Debye temperature (ΘD), which can be calculated using ab initio calculations, is an ideal proxy for quantitatively comparing structural rigidity among different inorganic compounds, allowing potentially efficient phosphors to be selected from large crystal structure databases. However, the high computational cost of these calculations limits estimating ΘD for unsubstituted host crystal structures only. It is assumed that the low substitution concentration of the rare-earth luminescent center does not significantly influence a material's Debye temperature. This work evaluates the validity of this approximation by examining the effect of luminescent center substitution on a host structure's ΘD. Two well-known phosphors, (Y1–xCex)3Al5O12 (x = 0 − 0.05) and Ba1-δEuδMgAl10O17 (δ = 0 − 0.15), were synthesized with varying rare-earth concentrations, while ΘD was computationally estimated and then determined by ultrasonic pulse-echo speed-of-sound and low-temperature heat capacity measurements. The ensuing results provide key implications for using ΘD as a proxy for structural rigidity in substituted inorganic compounds.

Modulation of the Double-Helical Cores: A New Strategy for Structural Predictions of Thiolate-Protected Gold Nanoclusters.

A fundamental understanding of the structural growth of thiolate-protected gold nanoclusters not only benefits experimental synthesis but also will advance the methodology for structural predictions and for rational design of highly stable nanoclusters. Herein, we report numerous new structures (11 total) of thiolate-protected gold nanoclusters predicted from theoretical modulation of the double-helical cores of experimentally determined nanoclusters. Among these newly predicted structures, Au32(SR)22, Au40(SR)26, and Au48(SR)30 are obtained by adding a defective layer containing 4 gold atoms on a structural sequence of experimentally crystallized nanoclusters, namely, Au28(SR)20, Au36(SR)24, and Au44(SR)28. The generic growth pattern underlying this sequence of nanoclusters can be viewed as adding the highly stable tetrahedral Au4 unit on the double-helical cores. Likewise, the other eight newly predicted structures, including two groups of isomeric structures corresponding to the sequence of experimentally determined Au28(SR)20, Au36(SR)24, Au44(SR)28, and Au52(SR)32 nanoclusters, are successfully predicted. Density functional theory calculations show that these 11 newly predicted nanoclusters exhibit large highest occupied molecular orbital-lowest unoccupied molecular orbital gaps and all-positive harmonic vibrational frequencies, suggesting their high chemical stabilities. Additional analyses on the structures and properties suggest that these newly predicted nanoclusters are very likely to be synthesized in the laboratory. Confirmation by experiments would validate the new strategy for structural prediction of thiolate-protected gold nanoclusters by taking advantage of a large structure database of crystallized ligand-protected gold nanoclusters with a variety of gold cores.

Full-profile search-match by the Rietveld method.

A new search-match procedure has been developed and tested which, in contrast to previously existing methods, does not use a set of lines identified from a diffraction pattern, but an optimized Rietveld fitting on the raw data. Modern computers with multicore processors allow the routine to be fast enough to perform the entire search in a reasonable time using quite large databases of crystal structures. The search-match is done using the crystal structures for all phases and the instrumental geometry, and as such can be applied to every kind of diffraction experiment, including X-rays, thermal/time-of-flight neutrons and electrons. The methodology can also be applied to nanocrystalline samples for which peak identification may be a problem. A web interface has been developed to permit easy testing and evaluation of the procedure. The quality of the results mainly depends on the availability of the sought phase in the structure database. The method permits not only phase identification but also a rapid quantification of the phases and their gross microstructural features, provided the instrumental function is known.

Computational analysis and identification of battery materials

Abstract Crystallography is a powerful descriptor of the atomic structure of solid-state matter and can be applied to analyse the phenomena present in functional materials. Especially for ion diffusion – one of the main processes found in electrochemical energy storage materials – crystallography can describe and evaluate the elementary steps for the hopping of mobile species from one crystallographic site to another. By translating this knowledge into parameters and search for similar numbers in other materials, promising compounds for future energy storage materials can be identified. Large crystal structure databases like the ICSD, CSD, and PCD have accumulated millions of measured crystal structures and thus represent valuable sources for future data mining and big-data approaches. In this work we want to present, on the one hand, crystallographic approaches based on geometric and crystal-chemical descriptors that can be easily applied to very large databases. On the other hand, we want to show methodologies based on ab initio and electronic modelling which can simulate the structure features more realistically, incorporating also dynamic processes. Their theoretical background, applicability, and selected examples are presented.

Learning from Correlations Based on Local Structure: Rare-Earth Nickelates Revisited

Statistical analysis of local atomic distortions in crystalline materials is a powerful tool for understanding coupled electronic and structural phase transitions in transition metal compounds. The analyses of such complex materials, however, often require significant domain knowledge to recognize limitations in the available data, whether it be experimentally reported crystal structures, property measurements, or computed quantities, and to understand when additional experiments or simulations may be necessary. Here we show how additional descriptive statistics and computational experiments can help researchers explicitly recognize these limitations and fill in missing gaps by constructing amplitude ( a) and normalized-amplitude ( n) distortion-mode property correlation-coefficient heat maps, aCCHMs and nCCHMs, respectively. We demonstrate this utility within the rare-earth nickelate perovskites RNiO3 (R = rare earth ≠ La), which exhibit antiferromagnetic and metal-insulator transitions with crystallographic symmetry breaking, and analyze the CCHMs obtained from experimental and first-principles derived symmetry modes. In contrast with the crystallographic trends gleaned from the reported experimental structures, the equilibrium structures obtained from density functional theory indicate that the Jahn-Teller distortion mode plays a negligible role in affecting the Néel temperature. We explain this discrepancy and discuss how different researchers might draw disparate conclusions from the same evidence, in particular from aCCHMs and nCCHMs. Last, we propose a general method for utilizing CCHMs for screening large databases of structures.

P2Rank: machine learning based tool for rapid and accurate prediction of ligand binding sites from protein structure

BackgroundLigand binding site prediction from protein structure has many applications related to elucidation of protein function and structure based drug discovery. It often represents only one step of many in complex computational drug design efforts. Although many methods have been published to date, only few of them are suitable for use in automated pipelines or for processing large datasets. These use cases require stability and speed, which disqualifies many of the recently introduced tools that are either template based or available only as web servers.ResultsWe present P2Rank, a stand-alone template-free tool for prediction of ligand binding sites based on machine learning. It is based on prediction of ligandability of local chemical neighbourhoods that are centered on points placed on the solvent accessible surface of a protein. We show that P2Rank outperforms several existing tools, which include two widely used stand-alone tools (Fpocket, SiteHound), a comprehensive consensus based tool (MetaPocket 2.0), and a recent deep learning based method (DeepSite). P2Rank belongs to the fastest available tools (requires under 1 s for prediction on one protein), with additional advantage of multi-threaded implementation.ConclusionsP2Rank is a new open source software package for ligand binding site prediction from protein structure. It is available as a user-friendly stand-alone command line program and a Java library. P2Rank has a lightweight installation and does not depend on other bioinformatics tools or large structural or sequence databases. Thanks to its speed and ability to make fully automated predictions, it is particularly well suited for processing large datasets or as a component of scalable structural bioinformatics pipelines.

Structural insights into suppressor of cytokine signaling 1 protein- identification of new leads for type 2 diabetes mellitus.

The study considers the Suppressor of cytokine signaling 1 (SOCS1) protein as a novel Type 2 diabetes mellitus (T2DM) drug target. T2DM in human beings is also triggered by the over expression of SOCS proteins. The SOCS1 acts as a ubiquitin ligase (E3), degrades Insulin Receptor Substrate 1 and 2 (IRS1 and IRS2) proteins, and causes insulin resistance. Therefore, the structure of the SOCS1 protein was evaluated using homology-modeling and molecular dynamics methods and validated using standard computational protocols. The Protein-Protein docking study of SOCS1 with its natural substrates, IRS1 and IRS2, and subsequent solvent accessible surface area analysis gave insight into the binding region of the SOCS1 protein. The in silico active site prediction tools highlight the residues Val155 to Ile211 in SOCS1 being implicated in the ubiquitin mediated protein degradation of the proteins IRS1 and IRS2. Virtual screening in the active site region, using large structural databases, results in selective lead structures with 3-Pyridinol, Xanthine, and Alanine moieties as Pharmacophore. The virtual screening study shows that the residues Glu149, Gly187, Arg188, Leu191, and Ser205 of the SOCS1 are important for binding. The docking study with current anti-diabetic therapeutics shows that the drugs Glibenclamide and Glyclopyramide have a partial affinity towards SOCS1. The predicted ADMET and IC50 properties for the identified ligands are within the acceptable range with drug-like properties. The structural data of SOCS1, its active site, and the identified lead structures are expedient in the development of new T2DM therapeutics.

Modeling Sequence-Dependent Peptide Fluctuations in Immunologic Recognition.

In cellular immunity, T cells recognize peptide antigens bound and presented by major histocompatibility complex (MHC) proteins. The motions of peptides bound to MHC proteins play a significant role in determining immunogenicity. However, existing approaches for investigating peptide/MHC motional dynamics are challenging or of low throughput, hindering the development of algorithms for predicting immunogenicity from large databases, such as those of tumor or genetically unstable viral genomes. We addressed this by performing extensive molecular dynamics simulations on a large structural database of peptides bound to the most commonly expressed human class-I MHC protein, HLA-A*0201. The simulations reproduced experimental indicators of motion and were used to generate simple models for predicting site-specific, rapid motions of bound peptides through differences in their sequence and chemical composition alone. The models can easily be applied on their own or incorporated into immunogenicity prediction algorithms. Beyond their predictive power, the models provide insight into how amino acid substitutions can influence peptide and protein motions and how dynamic information is communicated across peptides. They also indicate a link between peptide rigidity and hydrophobicity, two features known to be important in influencing cellular immune responses.

Bioinformatics and Drug Discovery.

Bioinformatic analysis can not only accelerate drug target identification and drug candidate screening and refinement, but also facilitate characterization of side effects and predict drug resistance. High-throughput data such as genomic, epigenetic, genome architecture, cistromic, transcriptomic, proteomic, and ribosome profiling data have all made significant contribution to mechanism-based drug discovery and drug repurposing. Accumulation of protein and RNA structures, as well as development of homology modeling and protein structure simulation, coupled with large structure databases of small molecules and metabolites, paved the way for more realistic protein-ligand docking experiments and more informative virtual screening. I present the conceptual framework that drives the collection of these high-throughput data, summarize the utility and potential of mining these data in drug discovery, outline a few inherent limitations in data and software mining these data, point out news ways to refine analysis of these diverse types of data, and highlight commonly used software and databases relevant to drug discovery.

Mapping and classifying molecules from a high-throughput structural database

High-throughput computational materials design promises to greatly accelerate the process of discovering new materials and compounds, and of optimizing their properties. The large databases of structures and properties that result from computational searches, as well as the agglomeration of data of heterogeneous provenance leads to considerable challenges when it comes to navigating the database, representing its structure at a glance, understanding structure–property relations, eliminating duplicates and identifying inconsistencies. Here we present a case study, based on a data set of conformers of amino acids and dipeptides, of how machine-learning techniques can help addressing these issues. We will exploit a recently-developed strategy to define a metric between structures, and use it as the basis of both clustering and dimensionality reduction techniques—showing how these can help reveal structure–property relations, identify outliers and inconsistent structures, and rationalise how perturbations (e.g. binding of ions to the molecule) affect the stability of different conformers.

Targeting the ubiquitin-conjugating enzyme E2D4 for cancer drug discovery-a structure-based approach.

Cancer progression is a global burden. The incidence and mortality now reach 30 million deaths per year. Several pathways of cancer are under investigation for the discovery of effective therapeutics. The present study highlights the structural details of the ubiquitin protein 'Ubiquitin-conjugating enzyme E2D4' (UBE2D4) for the novel lead structure identification in cancer drug discovery process. The evaluation of 3D structure of UBE2D4 was carried out using homology modelling techniques. The optimized structure was validated by standard computational protocols. The active site region of the UBE2D4 was identified using computational tools like CASTp, Q-site Finder and SiteMap. The hydrophobic pocket which is responsible for binding with its natural receptor ubiquitin ligase CHIP (C-terminal of Hsp 70 interacting protein) was identified through protein-protein docking study. Corroborating the results obtained from active site prediction tools and protein-protein docking study, the domain of UBE2D4 which is responsible for cancer cell progression is sorted out for further docking study. Virtual screening with large structural database like CB_Div Set and Asinex BioDesign small molecular structural database was carried out. The obtained new ligand molecules that have shown affinity towards UBE2D4 were considered for ADME prediction studies. The identified new ligand molecules with acceptable parameters of docking, ADME are considered as potent UBE2D4 enzyme inhibitors for cancer therapy.