Machine Learning Models for Accurate Prediction of Kinase Inhibitors with Different Binding Modes.

Noncovalent inhibitors of protein kinases have different modes of action. They bind to the active or inactive form of kinases, compete with ATP, stabilize inactive kinase conformations, or act through allosteric sites. Accordingly, kinase inhibitors have been classified on the basis of different binding modes. For medicinal chemistry, it would be very useful to derive mechanistic hypotheses for newly discovered inhibitors. Therefore, we have applied different machine learning approaches to generate models for predicting different classes of kinase inhibitors including types I, I1/2, and II as well as allosteric inhibitors. These models were built on the basis of compounds with binding modes confirmed by X-ray crystallography and yielded unexpectedly accurate and stable predictions without the need for deep learning. The results indicate that the new machine learning models have considerable potential for practical applications. Therefore, our data sets and models are made freely available.

MCR-ALS analysis of IR spectroscopy and XRD for the investigation of ibuprofen - nicotinamide cocrystal formation.

To improve aqueous solubility, a poorly water-soluble active ingredient is classically combined with a conformer to form cocrystals. Hot melt extrusion is one preparation method for the formation of cocrystal solids. The aim of our study was to determine the optimal temperature conditions for the formation of ibuprofen and nicotinamide cocrystals using real-time infrared (IR) and X-ray diffraction (XRD) measurements. IR spectra and XRD patterns were subjected to multivariate curve resolution alternating least squares (MCR-ALS) analysis and decomposed into several components. Each component was descriptive of a specific step in the formation of the cocrystal. Cocrystal formation was followed by a separation phase between amorphous ibuprofen and crystalline nicotinamide. Our results suggest that, when using the hot melt exclusion method, careful consideration should be made towards optimizing processing temperatures in order to prevent amorphization and promote control over the process of cocrystal formation.

The X-ray crystal structure of human endothelin 1, a polypeptide hormone regulator of blood pressure.

Human endothelin is a 21-amino-acid polypeptide, constrained by two intra-chain disulfide bridges, that is made by endothelial cells. It is the most potent vasoconstrictor in the body and is crucially important in the regulation of blood pressure. It plays a major role in a host of medical conditions, including hypertension, diabetes, stroke and cancer. Endothelin was crystallized 28 years ago in the putative space group P6122, but the structure was never successfully solved by X-ray diffraction. Using X-ray diffraction data from 1992, the structure has now been solved. Assuming a unit cell belonging to space group P61 and a twin fraction of 0.28, a solution emerged with two, almost identical, closely associated molecules in the asymmetric unit. Although the data extended to beyond 1.8 Šresolution, a model containing 25 waters was refined to 1.85 Šresolution with an R of 0.216 and an Rfree of 0.284. The disulfide-constrained `core' of the molecule, amino-acid residues 1-15, has a main-chain conformation that is essentially the same as endothelin when bound to its receptor, but many side-chain rotamers are different. The carboxy-terminal `tail' comprising amino-acid residues 16-21 is extended as when receptor-bound, but it exhibits a different conformation with respect to the `core'. The dimer that comprises the asymmetric unit is maintained almost exclusively by hydrophobic interactions and may be stable in an aqueous medium.

Ab initio structure determination from experimental fluctuation X-ray scattering data.

Fluctuation X-ray scattering (FXS) is an emerging experimental technique in which X-ray solution scattering data are collected from particles in solution using ultrashort X-ray exposures generated by a free-electron laser (FEL). FXS experiments overcome the low data-to-parameter ratios associated with traditional solution scattering measurements by providing several orders of magnitude more information in the final processed data. Here we demonstrate the practical feasibility of FEL-based FXS on a biological multiple-particle system and describe data-processing techniques required to extract robust FXS data and significantly reduce the required number of snapshots needed by introducing an iterative noise-filtering technique. We showcase a successful ab initio electron density reconstruction from such an experiment, studying the Paramecium bursaria Chlorella virus (PBCV-1).

Automated evaluation of quaternary structures from protein crystals.

A correct assessment of the quaternary structure of proteins is a fundamental prerequisite to understanding their function, physico-chemical properties and mode of interaction with other proteins. Currently about 90% of structures in the Protein Data Bank are crystal structures, in which the correct quaternary structure is embedded in the crystal lattice among a number of crystal contacts. Computational methods are required to 1) classify all protein-protein contacts in crystal lattices as biologically relevant or crystal contacts and 2) provide an assessment of how the biologically relevant interfaces combine into a biological assembly. In our previous work we addressed the first problem with our EPPIC (Evolutionary Protein Protein Interface Classifier) method. Here, we present our solution to the second problem with a new method that combines the interface classification results with symmetry and topology considerations. The new algorithm enumerates all possible valid assemblies within the crystal using a graph representation of the lattice and predicts the most probable biological unit based on the pairwise interface scoring. Our method achieves 85% precision (ranging from 76% to 90% for different oligomeric types) on a new dataset of 1,481 biological assemblies with consensus of PDB annotations. Although almost the same precision is achieved by PISA, currently the most popular quaternary structure assignment method, we show that, due to the fundamentally different approach to the problem, the two methods are complementary and could be combined to improve biological assembly assignments. The software for the automatic assessment of protein assemblies (EPPIC version 3) has been made available through a web server at http://www.eppic-web.org.

Pink-beam serial crystallography.

Serial X-ray crystallography allows macromolecular structure determination at both X-ray free electron lasers (XFELs) and, more recently, synchrotron sources. The time resolution for serial synchrotron crystallography experiments has been limited to millisecond timescales with monochromatic beams. The polychromatic, "pink", beam provides a more than two orders of magnitude increased photon flux and hence allows accessing much shorter timescales in diffraction experiments at synchrotron sources. Here we report the structure determination of two different protein samples by merging pink-beam diffraction patterns from many crystals, each collected with a single 100 ps X-ray pulse exposure per crystal using a setup optimized for very low scattering background. In contrast to experiments with monochromatic radiation, data from only 50 crystals were required to obtain complete datasets. The high quality of the diffraction data highlights the potential of this method for studying irreversible reactions at sub-microsecond timescales using high-brightness X-ray facilities.

Acknowledging Errors: Advanced Molecular Replacement with Phaser.

Molecular replacement is a method for solving the crystallographic phase problem using an atomic model for the target structure. State-of-the-art methods have moved the field significantly from when it was first envisaged as a method for solving cases of high homology and completeness between a model and target structure. Improvements brought about by application of maximum likelihood statistics mean that various errors in the model and pathologies in the data can be accounted for, so that cases hitherto thought to be intractable are standardly solvable. As a result, molecular replacement phasing now accounts for the lion's share of structures deposited in the Protein Data Bank. However, there will always be cases at the fringes of solvability. I discuss here the approaches that will help tackle challenging molecular replacement cases.

Rosetta Structure Prediction as a Tool for Solving Difficult Molecular Replacement Problems.

Molecular replacement (MR), a method for solving the crystallographic phase problem using phases derived from a model of the target structure, has proven extremely valuable, accounting for the vast majority of structures solved by X-ray crystallography. However, when the resolution of data is low, or the starting model is very dissimilar to the target protein, solving structures via molecular replacement may be very challenging. In recent years, protein structure prediction methodology has emerged as a powerful tool in model building and model refinement for difficult molecular replacement problems. This chapter describes some of the tools available in Rosetta for model building and model refinement specifically geared toward difficult molecular replacement cases.

Structure Refinement at Atomic Resolution.

X-Ray diffraction data at atomic resolution, i.e., beyond 1.2 Å, provide the most detailed and reliable information we have about the structure of macromolecules, which is especially important for validating new discoveries and resolving subtle issues of molecular mechanisms. Refinement at atomic resolution allows reliable interpretation of static disorder and solvent structure, as well as modeling of anisotropic atomic vibrations and even of H atoms. Stereochemical restraints can be relaxed or removed, providing unbiased information about macromolecular stereochemistry, which in turn can be used to define improved conformation-dependent libraries, and the surplus of data allows estimation of least-squares uncertainties in the derived parameters. At ultrahigh resolution it is possible to study charge density distribution by multipolar refinement of electrons in non-spherical orbitals.

Low Resolution Refinement of Atomic Models Against Crystallographic Data.

This review describes some of the problems encountered during low-resolution refinement and map calculation. Refinement is considered as an application of Bayes' theorem, allowing combination of information from various sources including crystallographic experimental data and prior chemical and structural knowledge. The sources of prior knowledge relevant to macromolecules include basic chemical information such as bonds and angles, structural information from reference models of known homologs, knowledge about secondary structures, hydrogen bonding patterns, and similarity of non-crystallographically related copies of a molecule. Additionally, prior information encapsulating local conformational conservation is exploited, keeping local interatomic distances similar to those in the starting atomic model. The importance of designing an accurate likelihood function-the only link between model parameters and observed data-is emphasized. The review also reemphasizes the importance of phases, and describes how the use of raw observed amplitudes could give a better correlation between the calculated and "true" maps. It is shown that very noisy or absent observations can be replaced by calculated structure factors, weighted according to the accuracy of the atomic model. This approach helps to smoothen the map. However, such replacement should be used sparingly, as the bias toward errors in the model could be too much to avoid. It is in general recommended that, whenever a new map is calculated, map quality should be judged by inspection of the parts of the map where there is no atomic model. It is also noted that it is advisable to work with multiple blurred and sharpened maps, as different parts of a crystal may exhibit different degrees of mobility. Doing so can allow accurate building of atomic models, accounting for overall shape as well as finer structural details. Some of the results described in this review have been implemented in the programs REFMAC5, ProSMART and LORESTR, which are available as part of the CCP4 software suite.

Stereochemistry and Validation of Macromolecular Structures.

Macromolecular structure is governed by the strict rules of stereochemistry. Several approaches to the validation of the correctness of the interpretation of crystallographic and NMR data that underlie the models deposited in the PDB are utilized in practice. The stereochemical rules applicable to macromolecular structures are discussed in this chapter. Practical, computer-based methods and tools of verification of how well the models adhere to those established structural principles to assure their quality are summarized.

Validation of Protein-Ligand Crystal Structure Models: Small Molecule and Peptide Ligands.

Models of target proteins in complex with small molecule ligands or peptide ligands are of significant interest to the biomedical research community. Structure-guided lead discovery and structure-based drug design make extensive use of such models. The bound ligands comprise only a small fraction of the total X-ray scattering mass, and therefore particular care must be taken to properly validate the atomic model of the ligand as experimental data can often be scarce. The ligand model must be validated against both the primary experimental data and the local environment, specifically: (1) the primary evidence in the form of the electron density, (2) examined for reasonable stereochemistry, and (3) the chemical plausibility of the binding interactions must be inspected. Tools that assist the researcher in the validation process are presented.

Protein Data Bank (PDB): The Single Global Macromolecular Structure Archive.

The Protein Data Bank (PDB)--the single global repository of experimentally determined 3D structures of biological macromolecules and their complexes--was established in 1971, becoming the first open-access digital resource in the biological sciences. The PDB archive currently houses ~130,000 entries (May 2017). It is managed by the Worldwide Protein Data Bank organization (wwPDB; wwpdb.org), which includes the RCSB Protein Data Bank (RCSB PDB; rcsb.org), the Protein Data Bank Japan (PDBj; pdbj.org), the Protein Data Bank in Europe (PDBe; pdbe.org), and BioMagResBank (BMRB; www.bmrb.wisc.edu). The four wwPDB partners operate a unified global software system that enforces community-agreed data standards and supports data Deposition, Biocuration, and Validation of ~11,000 new PDB entries annually (deposit.wwpdb.org). The RCSB PDB currently acts as the archive keeper, ensuring disaster recovery of PDB data and coordinating weekly updates. wwPDB partners disseminate the same archival data from multiple FTP sites, while operating complementary websites that provide their own views of PDB data with selected value-added information and links to related data resources. At present, the PDB archives experimental data, associated metadata, and 3D-atomic level structural models derived from three well-established methods: crystallography, nuclear magnetic resonance spectroscopy (NMR), and electron microscopy (3DEM). wwPDB partners are working closely with experts in related experimental areas (small-angle scattering, chemical cross-linking/mass spectrometry, Forster energy resonance transfer or FRET, etc.) to establish a federation of data resources that will support sustainable archiving and validation of 3D structural models and experimental data derived from integrative or hybrid methods.

High-Throughput Crystallography: Reliable and Efficient Identification of Fragment Hits.

Today the identification of lead structures for drug development often starts from small fragment-like molecules raising the chances to find compounds that successfully pass clinical trials. At the heart of the screening for fragments binding to a specific target, crystallography delivers structural information essential for subsequent drug design. While it is common to search for bound ligands in electron densities calculated directly after an initial refinement cycle, we raise the important question whether this strategy is viable for fragments characterized by low affinities. Here, we describe and provide a collection of high-quality diffraction data obtained from 364 protein crystals treated with diverse fragments. Subsequent data analysis showed that ∼25% of all hits would have been missed without further refining the resulting structures. To enable fast and reliable hit identification, we have designed an automated refinement pipeline that will inspire the development of optimized tools facilitating the successful application of fragment-based methods.

100 Years later: Celebrating the contributions of x-ray crystallography to allergy and clinical immunology.

Current knowledge of molecules involved in immunology and allergic disease results from the significant contributions of x-ray crystallography, a discipline that just celebrated its 100th anniversary. The histories of allergens and x-ray crystallography are intimately intertwined. The first enzyme structure to be determined was lysozyme, also known as the chicken food allergen Gal d 4. Crystallography determines the exact 3-dimensional positions of atoms in molecules. Structures of molecular complexes in the disciplines of immunology and allergy have revealed the atoms involved in molecular interactions and mechanisms of disease. These complexes include peptides presented by MHC class II molecules, cytokines bound to their receptors, allergen-antibody complexes, and innate immune receptors with their ligands. The information derived from crystallographic studies provides insights into the function of molecules. Allergen function is one of the determinants of environmental exposure, which is essential for IgE sensitization. Proteolytic activity of allergens or their capacity to bind LPSs can also contribute to allergenicity. The atomic positions define the molecular surface that is accessible to antibodies. In turn, this surface determines antibody specificity and cross-reactivity, which are important factors for the selection of allergen panels used for molecular diagnosis and the interpretation of clinical symptoms. This review celebrates the contributions of x-ray crystallography to clinical immunology and allergy, focusing on new molecular perspectives that influence the diagnosis and treatment of allergic diseases.

Protein crystallography for aspiring crystallographers or how to avoid pitfalls and traps in macromolecular structure determination.

The number of macromolecular structures deposited in the Protein Data Bank now approaches 100,000, with the vast majority of them determined by crystallographic methods. Thousands of papers describing such structures have been published in the scientific literature, and 20 Nobel Prizes in chemistry or medicine have been awarded for discoveries based on macromolecular crystallography. New hardware and software tools have made crystallography appear to be an almost routine (but still far from being analytical) technique and many structures are now being determined by scientists with very limited experience in the practical aspects of the field. However, this apparent ease is sometimes illusory and proper procedures need to be followed to maintain high standards of structure quality. In addition, many noncrystallographers may have problems with the critical evaluation and interpretation of structural results published in the scientific literature. The present review provides an outline of the technical aspects of crystallography for less experienced practitioners, as well as information that might be useful for users of macromolecular structures, aiming to show them how to interpret (but not overinterpret) the information present in the coordinate files and in their description. A discussion of the extent of information that can be gleaned from the atomic coordinates of structures solved at different resolution is provided, as well as problems and pitfalls encountered in structure determination and interpretation.

The utility of comparative models and the local model quality for protein crystal structure determination by Molecular Replacement.

BACKGROUND: Computational models of protein structures were proved to be useful as search models in Molecular Replacement (MR), a common method to solve the phase problem faced by macromolecular crystallography. The success of MR depends on the accuracy of a search model. Unfortunately, this parameter remains unknown until the final structure of the target protein is determined. During the last few years, several Model Quality Assessment Programs (MQAPs) that predict the local accuracy of theoretical models have been developed. In this article, we analyze whether the application of MQAPs improves the utility of theoretical models in MR. RESULTS: For our dataset of 615 search models, the real local accuracy of a model increases the MR success ratio by 101% compared to corresponding polyalanine templates. On the contrary, when local model quality is not utilized in MR, the computational models solved only 4.5% more MR searches than polyalanine templates. For the same dataset of the 615 models, a workflow combining MR with predicted local accuracy of a model found 45% more correct solution than polyalanine templates. To predict such accuracy MetaMQAPclust, a "clustering MQAP" was used. CONCLUSIONS: Using comparative models only marginally increases the MR success ratio in comparison to polyalanine structures of templates. However, the situation changes dramatically once comparative models are used together with their predicted local accuracy. A new functionality was added to the GeneSilico Fold Prediction Metaserver in order to build models that are more useful for MR searches. Additionally, we have developed a simple method, AmIgoMR (Am I good for MR?), to predict if an MR search with a template-based model for a given template is likely to find the correct solution.

Radiation damage in protein crystals is reduced with a micron-sized X-ray beam.

Radiation damage is a major limitation in crystallography of biological macromolecules, even for cryocooled samples, and is particularly acute in microdiffraction. For the X-ray energies most commonly used for protein crystallography at synchrotron sources, photoelectrons are the predominant source of radiation damage. If the beam size is small relative to the photoelectron path length, then the photoelectron may escape the beam footprint, resulting in less damage in the illuminated volume. Thus, it may be possible to exploit this phenomenon to reduce radiation-induced damage during data measurement for techniques such as diffraction, spectroscopy, and imaging that use X-rays to probe both crystalline and noncrystalline biological samples. In a systematic and direct experimental demonstration of reduced radiation damage in protein crystals with small beams, damage was measured as a function of micron-sized X-ray beams of decreasing dimensions. The damage rate normalized for dose was reduced by a factor of three from the largest (15.6 µm) to the smallest (0.84 µm) X-ray beam used. Radiation-induced damage to protein crystals was also mapped parallel and perpendicular to the polarization direction of an incident 1-µm X-ray beam. Damage was greatest at the beam center and decreased monotonically to zero at a distance of about 4 µm, establishing the range of photoelectrons. The observed damage is less anisotropic than photoelectron emission probability, consistent with photoelectron trajectory simulations. These experimental results provide the basis for data collection protocols to mitigate with micron-sized X-ray beams the effects of radiation damage.

Workshop on the validation and modeling of electron cryo-microscopy structures of biological nanomachines.

Electron cryo-microscopy (cryoEM) is a rapidly maturing methodology in structural biology, which now enables the determination of 3D structures of molecules, macromolecular complexes and cellular components at resolutions as high as 3.5Å, bridging the gap between light microscopy and X-ray crystallography/NMR. In recent years structures of many complex molecular machines have been visualized using this method. Single particle reconstruction, the most widely used technique in cryoEM, has recently demonstrated the capability of producing structures at resolutions approaching those of X-ray crystallography, with over a dozen structures at better than 5 Å resolution published to date. This method represents a significant new source of experimental data for molecular modeling and simulation studies. CryoEM derived maps and models are archived through EMDataBank.org joint deposition services to the EM Data Bank (EMDB) and Protein Data Bank (PDB), respectively. CryoEM maps are now being routinely produced over the 3 - 30 Å resolution range, and a number of computational groups are developing software for building coordinate models based on this data and developing validation techniques to better assess map and model accuracy. In this workshop we will present the results of the first cryoEM modeling challenge, in which computational groups were asked to apply their tools to a selected set of published cryoEM structures. We will also compare the results of the various applied methods, and discuss the current state of the art and how we can most productively move forward.