A. Bradley

OOMMPPAA: a tool to aid directed synthesis by the combined analysis of activity and structural data.

There is an ever increasing resource in terms of both structural information and activity data for many protein targets. In this paper we describe OOMMPPAA, a novel computational tool designed to inform compound design by combining such data. OOMMPPAA uses 3D matched molecular pairs to generate 3D ligand conformations. It then identifies pharmacophoric transformations between pairs of compounds and associates them with their relevant activity changes. OOMMPPAA presents this data in an interactive application providing the user with a visual summary of important interaction regions in the context of the binding site. We present validation of the tool using openly available data for CDK2 and a GlaxoSmithKline data set for a SAM-dependent methyl-transferase. We demonstrate OOMMPPAA’s application in optimizing both potency and cell permeability and use OOMMPPAA to highlight nuanced and cross-series SAR. OOMMPPAA is freely available to download at http://oommppaa.sgc.ox.ac.uk/OOMMPPAA/.

WONKA: objective novel complex analysis for ensembles of protein-ligand structures

WONKA is a tool for the systematic analysis of an ensemble of protein–ligand structures. It makes the identification of conserved and unusual features within such an ensemble straightforward. WONKA uses an intuitive workflow to process structural co-ordinates. Ligand and protein features are summarised and then presented within an interactive web application. WONKA’s power in consolidating and summarising large amounts of data is described through the analysis of three bromodomain datasets. Furthermore, and in contrast to many current methods, WONKA relates analysis to individual ligands, from which we find unusual and erroneous binding modes. Finally the use of WONKA as an annotation tool to share observations about structures is demonstrated. WONKA is freely available to download and install locally or can be used online at http://wonka.sgc.ox.ac.uk.

Partial-occupancy binders identified by the Pan-Dataset Density Analysis method offer new chemical opportunities and reveal cryptic binding sites

Crystallographic fragment screening uses low molecular weight compounds to probe the protein surface and although individual protein-fragment interactions are high quality, fragments commonly bind at low occupancy, historically making identification difficult. However, our new Pan-Dataset Density Analysis method readily identifies binders missed by conventional analysis: for fragment screening data of lysine-specific demethylase 4D (KDM4D), the hit rate increased from 0.9% to 10.6%. Previously unidentified fragments reveal multiple binding sites and demonstrate: the versatility of crystallographic fragment screening; that surprisingly large conformational changes are possible in crystals; and that low crystallographic occupancy does not by itself reflect a protein-ligand complexs significance.

A Multi-Crystal Method for Extracting Obscured Signal from Crystallographic Electron Density

Macromolecular crystallography is relied on to reveal subtle atomic difference between samples (e.g. ligand binding); yet their detection and modelling is subjective and ambiguous density is experimentally common, since molecular states of interest are generally only fractionally present. The existing approach relies on careful modelling for maximally accurate maps to make contributions of the minor fractions visible (1); in practice, this is time-consuming and non-objective (2–4). Instead, our PanDDA method automatically reveals clear electron density for only the changed state, even from poor models and inaccurate maps, by subtracting a proportion of the confounding ground state, accurately estimated by averaging many ground state crystals. Changed states are objectively identifiable from statistical distributions of density values; arbitrarily large searches are thus automatable. The method is completely general, implying new best practice for all changed-state studies. Finally, we demonstrate the incompleteness of current atomic models, and the need for new multi-crystal deconvolution paradigms. One Sentence Summary Normally uninterpretable map regions are reliably modelled by deconvoluting superposed crystal states, even with poor starting models.

Rapid covalent-probe discovery by electrophile fragment screening

Covalent probes can display unmatched potency, selectivity and duration of action, however, their discovery is challenging. In principle, fragments that can irreversibly bind their target can overcome the low affinity that limits reversible fragment screening. Such electrophilic fragments were considered non-selective and were rarely screened. We hypothesized that mild electrophiles might overcome the selectivity challenge, and constructed a library of 993 mildly electrophilic fragments. We characterized this library by a new high-throughput thiol-reactivity assay and screened them against ten cysteine-containing proteins. Highly reactive and promiscuous fragments were rare and could be easily eliminated. By contrast, we found selective hits for most targets. Combination with high-throughput crystallography allowed rapid progression to potent and selective probes for two enzymes, the deubiquitinase OTUB2, and the pyrophosphatase NUDT7. No inhibitors were previously known for either. This study highlights the potential of electrophile fragment screening as a practical and efficient tool for covalent ligand discovery.

Protein Family-Specific Models Using Deep Neural Networks and Transfer Learning Improve Virtual Screening and Highlight the Need for More Data

Machine learning has shown enormous potential for computer-aided drug discovery. Here we show how modern convolutional neural networks (CNNs) can be applied to structure-based virtual screening. We have coupled our densely connected CNN (DenseNet) with a transfer learning approach which we use to produce an ensemble of protein family-specific models. We conduct an in-depth empirical study and provide the first guidelines on the minimum requirements for adopting a protein family-specific model. Our method also highlights the need for additional data, even in data-rich protein families. Our approach outperforms recent benchmarks on the DUD-E data set and an independent test set constructed from the ChEMBL database. Using a clustered cross-validation on DUD-E, we achieve an average AUC ROC of 0.92 and a 0.5% ROC enrichment factor of 79. This represents an improvement in early enrichment of over 75% compared to a recent machine learning benchmark. Our results demonstrate that the continued improvements in machine learning architecture for computer vision apply to structure-based virtual screening.

WONKA and OOMMPPAA: analysis of protein-ligand interaction data to direct structure-based drug design.

The background to and development of WONKA and OOMMPPAA, tools for structure-based drug design, are described.