Search

Over the past few years, many machine learning-based scoring functions for predicting the binding of small molecules to proteins have been developed. Their objective is to approximate the distribution which takes two molecules as input and outputs the energy of their interaction. Only a scoring function that accounts for the interatomic interactions involved in binding can accurately predict binding affinity on unseen molecules. However, many scoring functions make predictions based on data set biases rather than an understanding of the physics of binding. These scoring functions perform well when tested on similar targets to those in the training set but fail to generalize to dissimilar targets. To test what a machine learning-based scoring function has learned, input attribution, a technique for learning which features are important to a model when making a prediction on a particular data point, can be applied. If a model successfully learns something beyond data set biases, attribution should give insight into the important binding interactions that are taking place. We built a machine learning-based scoring function that aimed to avoid the influence of bias via thorough train and test data set filtering and show that it achieves comparable performance on the Comparative Assessment of Scoring Functions, 2016 (CASF-2016) benchmark to other leading methods. We then use the CASF-2016 test set to perform attribution and find that the bonds identified as important by PointVS, unlike those extracted from other scoring functions, have a high correlation with those found by a distance-based interaction profiler. We then show that attribution can be used to extract important binding pharmacophores from a given protein target when supplied with a number of bound structures. We use this information to perform fragment elaboration and see improvements in docking scores compared to using structural information from a traditional, data-based approach. This not only provides definitive proof that the scoring function has learned to identify some important binding interactions but also constitutes the first deep learning-based method for extracting structural information from a target for molecule design.

A novel crystallographic fragment screening data set was generated and used in the SAMPL7 challenge for protein-ligands. The SAMPL challenges prospectively assess the predictive power of methods involved in computer-aided drug design. Application of various methods to fragment molecules are now widely used in the search for new drugs. However, there is little in the way of systematic validation specifically for fragment-based approaches. We have performed a large crystallographic high-throughput fragment screen against the therapeutically relevant second bromodomain of the Pleckstrin-homology domain interacting protein (PHIP2) that revealed 52 different fragments bound across 4 distinct sites, 47 of which were bound to the pharmacologically relevant acetylated lysine (Kac) binding site. These data were used to assess computational screening, binding pose prediction and follow-up enumeration. All submissions performed randomly for screening. Pose prediction success rates (defined as less than 2 Å root mean squared deviation against heavy atom crystal positions) ranged between 0 and 25% and only a very few follow-up compounds were deemed viable candidates from a medicinal-chemistry perspective based on a common molecular descriptors analysis. The tight deadlines imposed during the challenge led to a small number of submissions suggesting that the accuracy of rapidly responsive workflows remains limited. In addition, the application of these methods to reproduce crystallographic fragment data still appears to be very challenging. The results show that there is room for improvement in the development of computational tools particularly when applied to fragment-based drug design., (© 2022. The Author(s).)

Selectivity is a crucial property in small molecule development. Binding site comparisons within a protein family are a key piece of information when aiming to modulate the selectivity profile of a compound. Binding site differences can be exploited to confer selectivity for a specific target, while shared areas can provide insights into polypharmacology. As the quantity of structural data grows, automated methods are needed to process, summarize, and present these data to users. We present a computational method that provides quantitative and data-driven summaries of the available binding site information from an ensemble of structures of the same protein. The resulting ensemble maps identify the key interactions important for ligand binding in the ensemble. The comparison of ensemble maps of related proteins enables the identification of selectivity-determining regions within a protein family. We applied the method to three examples from the well-researched human bromodomain and kinase families, demonstrating that the method is able to identify selectivity-determining regions that have been used to introduce selectivity in past drug discovery campaigns. We then illustrate how the resulting maps can be used to automate comparisons across a target protein family.

In fragment-based drug discovery, hundreds or often thousands of compounds smaller than ~300 Da are tested against the protein of interest to identify chemical entities that can be developed into potent drug candidates. Since the compounds are small, interactions are weak, and the screening method must therefore be highly sensitive; moreover, structural information tends to be crucial for elaborating these hits into lead-like compounds. Therefore, protein crystallography has always been a gold-standard technique, yet historically too challenging to find widespread use as a primary screen. Initial XChem experiments were demonstrated in 2014 and then trialed with academic and industrial collaborators to validate the process. Since then, a large research effort and significant beamtime have streamlined sample preparation, developed a fragment library with rapid follow-up possibilities, automated and improved the capability of I04-1 beamline for unattended data collection, and implemented new tools for data management, analysis and hit identification. XChem is now a facility for large-scale crystallographic fragment screening, supporting the entire crystals-to-deposition process, and accessible to academic and industrial users worldwide. The peer-reviewed academic user program has been actively developed since 2016, to accommodate projects from as broad a scientific scope as possible, including well-validated as well as exploratory projects. Academic access is allocated through biannual calls for peer-reviewed proposals, and proprietary work is arranged by Diamond's Industrial Liaison group. This workflow has already been routinely applied to over a hundred targets from diverse therapeutic areas, and effectively identifies weak binders (1%-30% hit rate), which both serve as high-quality starting points for compound design and provide extensive structural information on binding sites. The resilience of the process was demonstrated by continued screening of SARS-CoV-2 targets during the COVID-19 pandemic, including a 3-week turn-around for the main protease.

Despite the tremendous success of X-ray cryo-crystallography in recent decades, the transfer of crystals from the drops in which they are grown to diffractometer sample mounts remains a manual process in almost all laboratories. Here, the Shifter, a motorized, interactive microscope stage that transforms the entire crystal-mounting workflow from a rate-limiting manual activity to a controllable, high-throughput semi-automated process, is described. By combining the visual acuity and fine motor skills of humans with targeted hardware and software automation, it was possible to transform the speed and robustness of crystal mounting. Control software, triggered by the operator, manoeuvres crystallization plates beneath a clear protective cover, allowing the complete removal of film seals and thereby eliminating the tedium of repetitive seal cutting. The software, either upon request or working from an imported list, controls motors to position crystal drops under a hole in the cover for human mounting at a microscope. The software automatically captures experimental annotations for uploading to the user's data repository, removing the need for manual documentation. The Shifter facilitates mounting rates of 100-240 crystals per hour in a more controlled process than manual mounting, which greatly extends the lifetime of the drops and thus allows a dramatic increase in the number of crystals retrievable from any given drop without loss of X-ray diffraction quality. In 2015, the first in a series of three Shifter devices was deployed as part of the XChem fragment-screening facility at Diamond Light Source, where they have since facilitated the mounting of over 120 000 crystals. The Shifter was engineered to have a simple design, providing a device that could be readily commercialized and widely adopted owing to its low cost. The versatile hardware design allows use beyond fragment screening and protein crystallography., (open access.)

Understanding allosteric regulation of proteins is fundamental to our study of protein structure and function. Moreover, allosteric binding pockets have become a major target of drug discovery efforts in recent years. However, even though the function of almost every protein can be influenced by allostery, it remains a challenge to discover, rationalise and validate putative allosteric binding pockets. This review examines how the discovery and analysis of putative allosteric binding sites have been influenced by the availability of centralised facilities for crystallographic fragment screening, along with newly developed computational methods for modelling low occupancy features. We discuss the experimental parameters required for success, and how new methods could influence the field in the future. Finally, we reflect on the general problem of how to translate these findings into actual ligand development programs., (Copyright © 2020. Published by Elsevier Ltd.)

The XChem facility at Diamond Light Source offers fragment screening by X-ray crystallography as a general access user program. The main advantage of X-ray crystallography as a primary fragment screen is that it yields directly the location and pose of the fragment hits, whether within pockets of interest or merely on surface sites: this is the key information for structure-based design and for enabling synthesis of follow-up molecules. Extensive streamlining of the screening experiment at XChem has engendered a very active user program that is generating large amounts of data: in 2017, 36 academic and industry groups generated 35,000 datasets of uniquely soaked crystals. It has also generated a large number of learnings concerning the main remaining bottleneck, namely, obtaining a suitable crystal system that will support a successful fragment screen. Here we discuss the practicalities of generating screen-ready crystals that have useful electron density maps, and how to ensure they will be successfully reproduced and usable at a facility outside the home lab., (© 2018 Elsevier Inc. All rights reserved.)

The ongoing explosion in genomics data has long since outpaced the capacity of conventional biochemical methodology to verify the large number of hypotheses that emerge from the analysis of such data. In contrast, it is still a gold-standard for early phenotypic validation towards small-molecule drug discovery to use probe molecules (or tool compounds), notwithstanding the difficulty and cost of generating them. Rational structure-based approaches to ligand discovery have long promised the efficiencies needed to close this divergence; in practice, however, this promise remains largely unfulfilled, for a host of well-rehearsed reasons and despite the huge technical advances spearheaded by the structural genomics initiatives of the noughties. Therefore the current, fourth funding phase of the Structural Genomics Consortium (SGC), building on its extensive experience in structural biology of novel targets and design of protein inhibitors, seeks to redefine what it means to do structural biology for drug discovery. We developed the concept of a Target Enabling Package (TEP) that provides, through reagents, assays and data, the missing link between genetic disease linkage and the development of usefully potent compounds. There are multiple prongs to the ambition: rigorously assessing targets' genetic disease linkages through crowdsourcing to a network of collaborating experts; establishing a systematic approach to generate the protocols and data that comprise each target's TEP; developing new, X-ray-based fragment technologies for generating high quality chemical matter quickly and cheaply; and exploiting a stringently open access model to build multidisciplinary partnerships throughout academia and industry. By learning how to scale these approaches, the SGC aims to make structures finally serve genomics, as originally intended, and demonstrate how 3D structures systematically allow new modes of druggability to be discovered for whole classes of targets., (© 2017 The Author(s).)

Although noncovalent binding by small molecules cannot be assumed a priori to be stoichiometric in the crystal lattice, occupancy refinement of ligands is often avoided by convention. Occupancies tend to be set to unity, requiring the occupancy error to be modelled by the B factors, and residual weak density around the ligand is necessarily attributed to `disorder'. Where occupancy refinement is performed, the complementary, superposed unbound state is rarely modelled. Here, it is shown that superior accuracy is achieved by modelling the ligand as partially occupied and superposed on a ligand-free `ground-state' model. Explicit incorporation of this model of the crystal, obtained from a reference data set, allows constrained occupancy refinement with minimal fear of overfitting. Better representation of the crystal also leads to more meaningful refined atomic parameters such as the B factor, allowing more insight into dynamics in the crystal. An outline of an approach for algorithmically generating ensemble models of crystals is presented, assuming that data sets representing the ground state are available. The applicability of various electron-density metrics to the validation of the resulting models is assessed, and it is concluded that ensemble models consistently score better than the corresponding single-state models. Furthermore, it appears that ignoring the superposed ground state becomes the dominant source of model error, locally, once the overall model is accurate enough; modelling the local ground state properly is then more meaningful than correcting all remaining model errors globally, especially for low-occupancy ligands. Implications for the simultaneous refinement of B factors and occupancies, and for future evaluation of the limits of the approach, in particular its behaviour at lower data resolution, are discussed.

XChemExplorer (XCE) is a data-management and workflow tool to support large-scale simultaneous analysis of protein-ligand complexes during structure-based ligand discovery (SBLD). The user interfaces of established crystallographic software packages such as CCP4 [Winn et al. (2011), Acta Cryst. D67, 235-242] or PHENIX [Adams et al. (2010), Acta Cryst. D66, 213-221] have entrenched the paradigm that a `project' is concerned with solving one structure. This does not hold for SBLD, where many almost identical structures need to be solved and analysed quickly in one batch of work. Functionality to track progress and annotate structures is essential. XCE provides an intuitive graphical user interface which guides the user from data processing, initial map calculation, ligand identification and refinement up until data dissemination. It provides multiple entry points depending on the need of each project, enables batch processing of multiple data sets and records metadata, progress and annotations in an SQLite database. XCE is freely available and works on any Linux and Mac OS X system, and the only dependency is to have the latest version of CCP4 installed. The design and usage of this tool are described here, and its usefulness is demonstrated in the context of fragment-screening campaigns at the Diamond Light Source. It is routinely used to analyse projects comprising 1000 data sets or more, and therefore scales well to even very large ligand-design projects.

The steady expansion in the capacity of modern beamlines for high-throughput data collection, enabled by increasing X-ray brightness, capacity of robotics and detector speeds, has pushed the bottleneck upstream towards sample preparation. Even in ligand-binding studies using crystal soaking, the experiment best able to exploit beamline capacity, a primary limitation is the need for gentle and nontrivial soaking regimens such as stepwise concentration increases, even for robust and well characterized crystals. Here, the use of acoustic droplet ejection for the soaking of protein crystals with small molecules is described, and it is shown that it is both gentle on crystals and allows very high throughput, with 1000 unique soaks easily performed in under 10 min. In addition to having very low compound consumption (tens of nanolitres per sample), the positional precision of acoustic droplet ejection enables the targeted placement of the compound/solvent away from crystals and towards drop edges, allowing gradual diffusion of solvent across the drop. This ensures both an improvement in the reproducibility of X-ray diffraction and increased solvent tolerance of the crystals, thus enabling higher effective compound-soaking concentrations. The technique is detailed here with examples from the protein target JMJD2D, a histone lysine demethylase with roles in cancer and the focus of active structure-based drug-design efforts.

Heavy-atom derivatization is one of the oldest techniques for obtaining phase information for protein crystals and, although it is no longer the first choice, it remains a useful technique for obtaining phases for unknown structures and for low-resolution data sets. It is also valuable for confirming the chain trace in low-resolution electron-density maps. This overview provides a summary of the technique and is aimed at first-time users of the method. It includes guidelines on when to use it, which heavy atoms are most likely to work, how to prepare heavy-atom solutions, how to derivatize crystals and how to determine whether a crystal is in fact a derivative.

Although protein crystallization is generally considered more art than science and remains significantly trial-and-error, large-scale data sets hold the promise of providing general learning. Observations are presented here from retrospective analyses of the strategies actively deployed for the extensive crystallization experiments at the Oxford site of the Structural Genomics Consortium (SGC), where comprehensive annotations by SGC scientists were recorded on a customized database infrastructure. The results point to the importance of using redundancy in crystallizing conditions, specifically by varying the mixing ratios of protein sample and precipitant, as well as incubation temperatures. No meaningful difference in performance could be identified between the four most widely used sparse-matrix screens, judged by the yield of crystals leading to deposited structures; this suggests that in general any comparison of screens will be meaningless without extensive cross-testing. Where protein sample is limiting, exploring more conditions has a higher likelihood of being informative by yielding hits than does redundancy of either mixing ratio or temperature. Finally, on the logistical question of how long experiments should be stored, 98% of all crystals that led to deposited structures appeared within 30 days. Overall, these analyses serve as practical guidelines for the design of initial screening experiments for new crystallization targets.

The quest for an optimal limited set of effective crystallization conditions remains a challenge in macromolecular crystallography, an issue that is complicated by the large number of chemicals which have been deemed to be suitable for promoting crystal growth. The lack of rational approaches towards the selection of successful chemical space and representative combinations has led to significant overlapping conditions, which are currently present in a multitude of commercially available crystallization screens. Here, an alternative approach to the sampling of widely used PEG precipitants is suggested through the use of PEG smears, which are mixtures of different PEGs with a requirement of either neutral or cooperatively positive effects of each component on crystal growth. Four newly defined smears were classified by molecular-weight groups and enabled the preservation of specific properties related to different polymer sizes. These smears not only allowed a wide coverage of properties of these polymers, but also reduced PEG variables, enabling greater sampling of other parameters such as buffers and additives. The efficiency of the smear-based screens was evaluated on more than 220 diverse recombinant human proteins, which overall revealed a good initial crystallization success rate of nearly 50%. In addition, in several cases successful crystallizations were only obtained using PEG smears, while various commercial screens failed to yield crystals. The defined smears therefore offer an alternative approach towards PEG sampling, which will benefit the design of crystallization screens sampling a wide chemical space of this key precipitant.

Bromodomains are epigenetic reader domains that have recently become popular targets. In contrast to BET bromodomains, which have proven druggable, bromodomains from other regions of the phylogenetic tree have shallower pockets. We describe successful targeting of the challenging BAZ2B bromodomain using biophysical fragment screening and structure-based optimization of high ligand-efficiency fragments into a novel series of low-micromolar inhibitors. Our results provide attractive leads for development of BAZ2B chemical probes and indicate the whole family may be tractable.

The radiation damage behaviour in 43 datasets of 34 different proteins collected over a year was examined, in order to gauge the reliability of decay metrics in practical situations, and to assess how these datasets, optimized only empirically for decay, would have benefited from the precise and automatic prediction of decay now possible with the programs RADDOSE [Murray, Garman & Ravelli (2004). J. Appl. Cryst. 37, 513-522] and BEST [Bourenkov & Popov (2010). Acta Cryst. D66, 409-419]. The results indicate that in routine practice the diffraction experiment is not yet characterized well enough to support such precise predictions, as these depend fundamentally on three interrelated variables which cannot yet be determined robustly and practically: the flux density distribution of the beam; the exact crystal volume; the sensitivity of the crystal to dose. The former two are not satisfactorily approximated from typical beamline information such as nominal beam size and transmission, or two-dimensional images of the beam and crystal; the discrepancies are particularly marked when using microfocus beams (<20 µm). Empirically monitoring decay with the dataset scaling B factor (Bourenkov & Popov, 2010) appears more robust but is complicated by anisotropic and/or low-resolution diffraction. These observations serve to delineate the challenges, scientific and logistic, that remain to be addressed if tools for managing radiation damage in practical data collection are to be conveniently robust enough to be useful in real time.

As many of the structural genomics centers have ended their first phase of operation, it is a good point to evaluate the scientific impact of this endeavour. The Structural Genomics Consortium (SGC), operating from three centers across the Atlantic, investigates human proteins involved in disease processes and proteins from Plasmodium falciparum and related organisms. We present here some of the scientific output of the Oxford node of the SGC, where the target areas include protein kinases, phosphatases, oxidoreductases and other metabolic enzymes, as well as signal transduction proteins. The SGC has aimed to achieve extensive coverage of human gene families with a focus on protein-ligand interactions. The methods employed for effective protein expression, crystallization and structure determination by X-ray crystallography are summarized. In addition to the cumulative impact of accelerated delivery of protein structures, we demonstrate how family coverage, generic screening methodology, and the availability of abundant purified protein samples, allow a level of discovery that is difficult to achieve otherwise. The contribution of NMR to structure determination and protein characterization is discussed. To make this information available to a wide scientific audience, a new tool for disseminating annotated structural information was created that also represents an interactive platform allowing for a continuous update of the annotation by the scientific community.

Understanding allosteric regulation of proteins is fundamental to our study of protein structure and function. Moreover, allosteric binding pockets have become a major target of drug discovery efforts in recent years. However, even though the function of almost every protein can be influenced by allostery, it remains a challenge to discover, rationalise and validate putative allosteric binding pockets. This review examines how the discovery and analysis of putative allosteric binding sites have been influenced by the availability of centralised facilities for crystallographic fragment screening, along with newly developed computational methods for modelling low occupancy features. We discuss the experimental parameters required for success, and how new methods could influence the field in the future. Finally, we reflect on the general problem of how to translate these findings into actual ligand development programs.

Fragment-based drug design has introduced a bottom-up process for drug development, with improved sampling of chemical space and increased effectiveness in early drug discovery. Here, we combine the use of pharmacophores, the most general concept of representing drug-target interactions with the theory of protein hotspots, to develop a design protocol for fragment libraries. The SpotXplorer approach compiles small fragment libraries that maximize the coverage of experimentally confirmed binding pharmacophores at the most preferred hotspots. The efficiency of this approach is demonstrated with a pilot library of 96 fragment-sized compounds (SpotXplorer0) that is validated on popular target classes and emerging drug targets. Biochemical screening against a set of GPCRs and proteases retrieves compounds containing an average of 70% of known pharmacophores for these targets. More importantly, SpotXplorer0 screening identifies confirmed hits against recently established challenging targets such as the histone methyltransferase SETD2, the main protease (3CLPro) and the NSP3 macrodomain of SARS-CoV-2. Fragment-based drug discovery employs screening of small polar compounds typically exhibiting low affinity towards protein targets. Here, the authors combine the use of protein-based binding pharmacophores with the theory of protein hotspots to develop a design protocol for fragment libraries, called SpotXplorer, and validate their approach on common and emerging drug targets. [ABSTRACT FROM AUTHOR]

The productive exploration of chemical space is an enduring challenge in chemical biology and medicinal chemistry. Natural products are biologically relevant, and their frameworks have facilitated chemical tool and drug discovery. A 'top-down' synthetic approach is described that enabled a range of complex bridged intermediates to be converted with high step efficiency into 26 diverse sp3-rich scaffolds. The scaffolds have local natural product-like features, but are only distantly related to specific natural product frameworks. To assess biological relevance, a set of 52 fragments was prepared, and screened by high-throughput crystallography against three targets from two protein families (ATAD2, BRD1 and JMJD2D). In each case, 3D fragment hits were identified that would serve as distinctive starting points for ligand discovery. This demonstrates that frameworks that are distantly related to natural products can facilitate discovery of new biologically relevant regions within chemical space. [ABSTRACT FROM AUTHOR]

Soluble guanylate cyclase (sGC) catalyses the synthesis of cyclic GMP in response to nitric oxide. The enzyme is a heterodimer of homologous α and β subunits, each of which is composed of multiple domains. We present here crystal structures of a heterodimer of the catalytic domains of the α and β subunits, as well as an inactive homodimer of β subunits. This first structure of a metazoan, heteromeric cyclase provides several observations. First, the structures resemble known structures of adenylate cyclases and other guanylate cyclases in overall fold and in the arrangement of conserved active-site residues, which are contributed by both subunits at the interface. Second, the subunit interaction surface is promiscuous, allowing both homodimeric and heteromeric association; the preference of the full-length enzyme for heterodimer formation must derive from the combined contribution of other interaction interfaces. Third, the heterodimeric structure is in an inactive conformation, but can be superposed onto an active conformation of adenylate cyclase by a structural transition involving a 26° rigid-body rotation of the α subunit. In the modelled active conformation, most active site residues in the subunit interface are precisely aligned with those of adenylate cyclase. Finally, the modelled active conformation also reveals a cavity related to the active site by pseudo-symmetry. The pseudosymmetric site lacks key active site residues, but may bind allosteric regulators in a manner analogous to the binding of forskolin to adenylate cyclase. This indicates the possibility of developing a new class of small-molecule modulators of guanylate cyclase activity targeting the catalytic domain. [ABSTRACT FROM AUTHOR]

Fragment based methods are now widely used to identify starting points in drug discovery and generation of tools for chemical biology. A significant challenge is optimization of these weak binding fragments to hit and lead compounds. We have developed an approach where individual reaction mixtures of analogues of hits can be evaluated without purification of the product. Here, we describe experiments to optimise the processes and then assess such mixtures in the high throughput crystal structure determination facility, XChem. Diffraction data for crystals of the proteins Hsp90 and PDHK2 soaked individually with 83 crude reaction mixtures are analysed manually or with the automated XChem procedures. The results of structural analysis are compared with binding measurements from other biophysical techniques. This approach can transform early hit to lead optimisation and the lessons learnt from this study provide a protocol that can be used by the community. Fragment-based screening offers a valuable method for identifying ligands and can be expanded to crude reaction mixtures. Here, high-throughput crystallographic screening of crystals soaked with crude reaction mixtures is assessed as a method to rapidly obtain ligands with improved affinity. [ABSTRACT FROM AUTHOR]

Cullin-RING ligases are multisubunit E3 ubiquitin ligases that recruit substrate-specific adaptors to catalyze protein ubiquitylation. Cul3-based Cullin-RING ligases are uniquely associated with BTB adaptors that incorporate homodimerization, Cul3 assembly, and substrate recognition into a single multidomain protein, of which the best known are BTB-BACK-Kelch domain proteins, including KEAP1. Cul3 assembly requires a BTB protein "3-box" motif, analogous to the F-box and SOCS box motifs of other Cullin-based E3s. To define the molecular basis for this assembly and the overall architecture of the E3, we determined the crystal structures of the BTB-BACK domains of KLHL11 both alone and in complex with Cul3, along with the Kelch domain structures of KLHL2 (Mayven), KLHL7, KLHL12, and KBTBD5. We show that Cul3 interaction is dependent on a unique N-terminal extension sequence that packs against the 3-box in a hydrophobic groove centrally located between the BTB and BACK domains. Deletion of this N-terminal region results in a 30-fold loss in affinity. The presented data offer a model for the quaternary assembly of this E3 class that supports the bivalent capture of Nrf2 and reveals potential new sites for E3 inhibitor design. [ABSTRACT FROM AUTHOR]