Idea2Data: Toward a New Paradigm for Drug Discovery.

Increasing the success rate and throughput of drug discovery will require efficiency improvements throughout the process that is currently used in the pharmaceutical community, including the crucial step of identifying hit compounds to act as drivers for subsequent optimization. Hit identification can be carried out through large compound collection screening and often involves the generation and testing of many hypotheses based on available knowledge. In practice, hypothesis generation can involve the selection of promising chemical structures from compound collections using predictive models built from previous screening/assay results. Available physical collections, typically used during hit identification, are of the order of 106 compounds but represent only a small fraction of the small molecule drug-like chemical space. In an effort to survey a larger portion of chemical space and eliminate inefficiencies during hit identification, we introduce a new process, termed Idea2Data (I2D) that tightly integrates computational and experimental components of the drug discovery process. I2D provides the ability to connect a vast virtual collection of compounds readily synthesizable on automated synthesis systems with computational predictive models for the identification of promising structures. This new paradigm enables researchers to process billions of virtual molecules and select structures that can be prepared on automated systems and made available for biological testing, allowing for timely hypothesis testing and follow-up. Since its introduction, I2D has positively impacted several portfolio efforts through identification of new chemical scaffolds and functionalization of existing scaffolds. In this Innovations paper, we describe the I2D process and present an application for the discovery of new ULK inhibitors.

Reconstruction of 3D structures of MET antibodies from electron microscopy 2D class averages.

Dynamics of three MET antibody constructs (IgG1, IgG2, and IgG4) and the IgG4-MET antigen complex was investigated by creating their atomic models with an integrative experimental and computational approach. In particular, we used two-dimensional (2D) Electron Microscopy (EM) images, image class averaging, homology modeling, Rapidly exploring Random Tree (RRT) structure sampling, and fitting of models to images, to find the relative orientations of antibody domains that are consistent with the EM images. We revealed that the conformational preferences of the constructs depend on the extent of the hinge flexibility. We also quantified how the MET antigen impacts on the conformational dynamics of IgG4. These observations allow to create testable hypothesis to investigate MET biology. Our protocol may also help describe structural diversity of other antigen systems at approximately 5 Å precision, as quantified by Root-Mean-Square Deviation (RMSD) among good-scoring models.

A Residue-Resolved Bayesian Approach to Quantitative Interpretation of Hydrogen-Deuterium Exchange from Mass Spectrometry: Application to Characterizing Protein-Ligand Interactions.

Characterization of interactions between proteins and other molecules is crucial for understanding the mechanisms of action of biological systems and, thus, drug discovery. An increasingly useful approach to mapping these interactions is measurement of hydrogen/deuterium exchange (HDX) using mass spectrometry (HDX-MS), which measures the time-resolved deuterium incorporation of peptides obtained by enzymatic digestion of the protein. Comparison of exchange rates between apo- and ligand-bound conditions results in a mapping of the differential HDX (ΔHDX) of the ligand. Residue-level analysis of these data, however, must account for experimental error, sparseness, and ambiguity due to overlapping peptides. Here, we propose a Bayesian method consisting of a forward model, noise model, prior probabilities, and a Monte Carlo sampling scheme. This method exploits a residue-resolved exponential rate model of HDX-MS data obtained from all peptides simultaneously, and explicitly models experimental error. The result is the best possible estimate of ΔHDX magnitude and significance for each residue given the data. We demonstrate the method by revealing richer structural interpretation of ΔHDX data on two nuclear receptors: vitamin D-receptor (VDR) and retinoic acid receptor gamma (RORγ). The method is implemented in HDX Workbench and as a standalone module of the open source Integrative Modeling Platform.

Biased retrieval of chemical series in receptor-based virtual screening.

Using the kinases in the DUD dataset and an in-house HTS dataset from PI3K-γ, receptor-based virtual screening experiments were performed using Glide SP docking. While significant enrichments were observed for eight of the nine targets in the set, more detailed analyses highlighted that much of the early enrichment (10-80%) is the result of retrieval of a single cluster of active compounds. This biased retrieval was not necessarily due to early enrichment of the cluster containing the co-crystallized ligand. Virtual screening validation studies could thus benefit from including cluster-based analyses to assess enrichment of diverse chemotypes.

Chemical space sampling in virtual screening by different crystal structures.

Retrospective virtual screening experiments were carried out to investigate the effects of combining hit lists from different crystal structures of the same target using consensus scoring. An in-house High Throughput Screening (HTS) dataset from PI3K-γ was used and docked against five diverse PI3K-γ crystal structures. The results show that consensus scoring prioritizes compounds that score moderately against individual crystal structures and is thus complementary to individual crystal structure screening leading to an increase in the diversity of hits. Enrichment factors (EFs) of the consensus score for two or three structures are often as high as or higher than the EF of the individual structures used in the consensus score. Combining four or five structures in the consensus score generally yields lower enrichments. Compounds in the top 500 of the consensus score that are also found in the top 500 of an individual X-ray structure used in the consensus score calculations yield the largest number of hits with the lowest number of false positives.

A protein relational database and protein family knowledge bases to facilitate structure-based design analyses.

The Protein Data Bank is the most comprehensive source of experimental macromolecular structures. It can, however, be difficult at times to locate relevant structures with the Protein Data Bank search interface. This is particularly true when searching for complexes containing specific interactions between protein and ligand atoms. Moreover, searching within a family of proteins can be tedious. For example, one cannot search for some conserved residue as residue numbers vary across structures. We describe herein three databases, Protein Relational Database, Kinase Knowledge Base, and Matrix Metalloproteinase Knowledge Base, containing protein structures from the Protein Data Bank. In Protein Relational Database, atom-atom distances between protein and ligand have been precalculated allowing for millisecond retrieval based on atom identity and distance constraints. Ring centroids, centroid-centroid and centroid-atom distances and angles have also been included permitting queries for pi-stacking interactions and other structural motifs involving rings. Other geometric features can be searched through the inclusion of residue pair and triplet distances. In Kinase Knowledge Base and Matrix Metalloproteinase Knowledge Base, the catalytic domains have been aligned into common residue numbering schemes. Thus, by searching across Protein Relational Database and Kinase Knowledge Base, one can easily retrieve structures wherein, for example, a ligand of interest is making contact with the gatekeeper residue.

Insights for predicting blood-brain barrier penetration of CNS targeted molecules using QSPR approaches.

Due to the high attrition rate of central nervous system drug candidates during clinical trials, the assessment of blood-brain barrier (BBB) penetration in early research is particularly important. A genetic approximation (GA)-based regression model was developed for predicting in vivo blood-brain partitioning data, expressed as logBB (log[brain]/[blood]). The model was built using an in-house data set of 193 compounds assembled from 22 different therapeutic projects. The final model (cross-validated r(2) = 0.72) with five molecular descriptors was selected based on validation using several large internal and external test sets. We demonstrate the potential utility of the model by applying it to a set of literature reported secretase inhibitors. In addition, we describe a rule-based approach for rapid assessment of brain penetration with several simple molecular descriptors.

Computational alanine scanning with linear scaling semiempirical quantum mechanical methods.

Alanine scanning is a powerful experimental tool for understanding the key interactions in protein-protein interfaces. Linear scaling semiempirical quantum mechanical calculations are now sufficiently fast and robust to allow meaningful calculations on large systems such as proteins, RNA and DNA. In particular, they have proven useful in understanding protein-ligand interactions. Here we ask the question: can these linear scaling quantum mechanical methods developed for protein-ligand scoring be useful for computational alanine scanning? To answer this question, we assembled 15 protein-protein complexes with available crystal structures and sufficient alanine scanning data. In all, the data set contains Delta Delta Gs for 400 single point alanine mutations of these 15 complexes. We show that with only one adjusted parameter the quantum mechanics-based methods outperform both buried accessible surface area and a potential of mean force and compare favorably to a variety of published empirical methods. Finally, we closely examined the outliers in the data set and discuss some of the challenges that arise from this examination.

Chemical space sampling by different scoring functions and crystal structures.

Virtual screening has become a popular tool to identify novel leads in the early phases of drug discovery. A variety of docking and scoring methods used in virtual screening have been the subject of active research in an effort to gauge limitations and articulate best practices. However, how to best utilize different scoring functions and various crystal structures, when available, is not yet well understood. In this work we use multiple crystal structures of PI3 K-gamma in both prospective and retrospective virtual screening experiments. Both Glide SP scoring and Prime MM-GBSA rescoring are utilized in the prospective and retrospective virtual screens, and consensus scoring is investigated in the retrospective virtual screening experiments. The results show that each of the different crystal structures that was used, samples a different chemical space, i.e. different chemotypes are prioritized by each structure. In addition, the different (re)scoring functions prioritize different chemotypes as well. Somewhat surprisingly, the Prime MM-GBSA scoring function generally gives lower enrichments than Glide SP. Finally we investigate the impact of different ligand preparation protocols on virtual screening enrichment factors. In summary, different crystal structures and different scoring functions are complementary to each other and allow for a wider variety of chemotypes to be considered for experimental follow-up.

Using a homology model of cytochrome P450 2D6 to predict substrate site of metabolism.

CYP2D6 is an important enzyme that is involved in first pass metabolism and is responsible for metabolizing ~25% of currently marketed drugs. A homology model of CYP2D6 was built using X-ray structures of ligand-bound CYP2C5 complexes as templates. This homology model was used in docking studies to rationalize and predict the site of metabolism of known CYP2D6 substrates. While the homology model was generally found to be in good agreement with the recently solved apo (ligand-free) X-ray structure of CYP2D6, significant differences between the structures were observed in the B' and F-G helical region. These structural differences are similar to those observed between ligand-free and ligand-bound structures of other CYPs and suggest that these conformational changes result from induced-fit adaptations upon ligand binding. By docking to the homology model using Glide, it was possible to identify the correct site of metabolism for a set of 16 CYP2D6 substrates 85% of the time when the 5 top scoring poses were examined. On the other hand, docking to the apo CYP2D6 X-ray structure led to a loss in accuracy in predicting the sites of metabolism for many of the CYP2D6 substrates considered in this study. These results demonstrate the importance of describing substrate-induced conformational changes that occur upon binding. The best results were obtained using Glide SP with van der Waals scaling set to 0.8 for both the receptor and ligand atoms. A discussion of putative binding modes that explain the distribution of metabolic sites for substrates, as well as a relationship between the number of metabolic sites and substrate size, are also presented. In addition, analysis of these binding modes enabled us to rationalize the typical hydroxylation and O-demethylation reactions catalyzed by CYP2D6 as well as the less common N-dealkylation.

An enriched structural kinase database to enable kinome-wide structure-based analyses and drug discovery.

The development of a kinase structural database, the kinase knowledge base (KKB), is described. It covers all human kinase domain structures that have been deposited in the Protein Data Bank. All structures are renumbered using a common scheme, which enables efficient cross-comparisons and multiple queries of interest to the kinase field. The common numbering scheme is also used to automatically annotate conserved residues and motifs, and conformationally classify the structures based on the DFG-loop and Helix C. Analyses of residue conservation in the ATP binding site using the full human-kinome-sequence alignment lead to the identification of a conserved hydrogen bond between the hinge region backbone and a glycine in the specificity surface. Furthermore, 90% of kinases are found to have at least one stabilizing interaction for the hinge region, which has not been described before.

A structural informatics approach to mine kinase knowledge bases.

In this paper, we describe a combination of structural informatics approaches developed to mine data extracted from existing structure knowledge bases (Protein Data Bank and the GVK database) with a focus on kinase ATP-binding site data. In contrast to existing systems that retrieve and analyze protein structures, our techniques are centered on a database of ligand-bound geometries in relation to residues lining the binding site and transparent access to ligand-based SAR data. We illustrate the systems in the context of the Abelson kinase and related inhibitor structures.

Development of QSAR models for microsomal stability: identification of good and bad structural features for rat, human and mouse microsomal stability.

High throughput microsomal stability assays have been widely implemented in drug discovery and many companies have accumulated experimental measurements for thousands of compounds. Such datasets have been used to develop in silico models to predict metabolic stability and guide the selection of promising candidates for synthesis. This approach has proven most effective when selecting compounds from proposed virtual libraries prior to synthesis. However, these models are not easily interpretable at the structural level, and thus provide little insight to guide traditional synthetic efforts. We have developed global classification models of rat, mouse and human liver microsomal stability using in-house data. These models were built with FCFP_6 fingerprints using a Naïve Bayesian classifier within Pipeline Pilot. The test sets were correctly classified as stable or unstable with satisfying accuracies of 78, 77 and 75% for rat, human and mouse models, respectively. The prediction confidence was assigned using the Bayesian score to assess the applicability of the models. Using the resulting models, we developed a novel data mining strategy to identify structural features associated with good and bad microsomal stability. We also used this approach to identify structural features which are good for one species but bad for another. With these findings, the structure-metabolism relationships are likely to be understood faster and earlier in drug discovery.

Modeling G protein-coupled receptors for structure-based drug discovery using low-frequency normal modes for refinement of homology models: application to H3 antagonists.

G Protein-Coupled Receptors (GPCRs) are integral membrane proteins that play important role in regulating key physiological functions, and are targets of about 50% of all recently launched drugs. High-resolution experimental structures are available only for very few GPCRs. As a result, structure-based drug design efforts for GPCRs continue to rely on in silico modeling, which is considered to be an extremely difficult task especially for these receptors. Here, we describe Gmodel, a novel approach for building 3D atomic models of GPCRs using a normal mode-based refinement of homology models. Gmodel uses a small set of relevant low-frequency vibrational modes derived from Random Elastic Network model to efficiently sample the large-scale receptor conformation changes and generate an ensemble of alternative models. These are used to assemble receptor-ligand complexes by docking a known active into each of the alternative models. Each of these is next filtered using restraints derived from known mutation and binding affinity data and is refined in the presence of the active ligand. In this study, Gmodel was applied to generate models of the antagonist form of histamine 3 (H3) receptor. The validity of this novel modeling approach is demonstrated by performing virtual screening (using the refined models) that consistently produces highly enriched hit lists. The models are further validated by analyzing the available SAR related to classical H3 antagonists, and are found to be in good agreement with the available experimental data, thus providing novel insights into the receptor-ligand interactions.

Computation of 3D queries for ROCS based virtual screens.

Rapid overlay of chemical structures (ROCS) is a method that aligns molecules based on shape and/or chemical similarity. It is often used in 3D ligand-based virtual screening. Given a query consisting of a single conformation of an active molecule ROCS can generate highly enriched hit lists. Typically the chosen query conformation is a minimum energy structure. Can better enrichment be obtained using conformations other than the minimum energy structure? To answer this question a methodology has been developed called CORAL (COnformational analysis, Rocs ALignment). For a given set of molecule conformations it computes optimized conformations for ROCS screening. It does so by clustering all conformations of a chosen molecule set using pairwise ROCS combo scores. The best representative conformation is that which has the highest average overlap with the rest of the conformations in the cluster. It is these best representative conformations that are then used for virtual screening. CORAL was tested by performing virtual screening experiments with the 40 DUD (Directory of Useful Decoys) data sets. Both CORAL and minimum energy queries were used. The recognition capability of each query was quantified as the area under the ROC curve (AUC). Results show that the CORAL AUC values are on average larger than the minimum energy AUC values. This demonstrates that one can indeed obtain better ROCS enrichments with conformations other than the minimum energy structure. As a result, CORAL analysis can be a valuable first step in virtual screening workflows using ROCS.

Escape from flatland: increasing saturation as an approach to improving clinical success.

The medicinal chemistry community has become increasingly aware of the value of tracking calculated physical properties such as molecular weight, topological polar surface area, rotatable bonds, and hydrogen bond donors and acceptors. We hypothesized that the shift to high-throughput synthetic practices over the past decade may be another factor that may predispose molecules to fail by steering discovery efforts toward achiral, aromatic compounds. We have proposed two simple and interpretable measures of the complexity of molecules prepared as potential drug candidates. The first is carbon bond saturation as defined by fraction sp(3) (Fsp(3)) where Fsp(3) = (number of sp(3) hybridized carbons/total carbon count). The second is simply whether a chiral carbon exists in the molecule. We demonstrate that both complexity (as measured by Fsp(3)) and the presence of chiral centers correlate with success as compounds transition from discovery, through clinical testing, to drugs. In an attempt to explain these observations, we further demonstrate that saturation correlates with solubility, an experimental physical property important to success in the drug discovery setting.

GARD: a Generally Applicable Replacement for RMSD.

The root-mean-squared deviation (rmsd) is a widely used measure of distance between two aligned objects -- often chemical structures. However, rmsd has a number of known limitations including difficulty of interpretation, no limit on weighting for any portion of the alignment, and a lack of normalization. In this work, a Generally Applicable Replacement for rmsD (GARD) is proposed. In this implementation atomic contributions are weighted by their relative importance to binding, as determined statistically by Andrews et al. (1) , and as such this method is 'chemically aware'. This novel measure is normalized and does not have many of the failings of traditional rmsd. It is, thus, perfectly suited for a wide variety of uses, including the assessment of the quality of poses produced from molecular docking programs and the comparison of conformers. Rmsd and GARD are compared in their ability to assess docking software and multiple examples of the use of GARD to rescue essentially correct poses with a high rmsd are presented.

Comparison of several molecular docking programs: pose prediction and virtual screening accuracy.

Molecular docking programs are widely used modeling tools for predicting ligand binding modes and structure based virtual screening. In this study, six molecular docking programs (DOCK, FlexX, GLIDE, ICM, PhDOCK, and Surflex) were evaluated using metrics intended to assess docking pose and virtual screening accuracy. Cognate ligand docking to 68 diverse, high-resolution X-ray complexes revealed that ICM, GLIDE, and Surflex generated ligand poses close to the X-ray conformation more often than the other docking programs. GLIDE and Surflex also outperformed the other docking programs when used for virtual screening, based on mean ROC AUC and ROC enrichment values obtained for the 40 protein targets in the Directory of Useful Decoys (DUD). Further analysis uncovered general trends in accuracy that are specific for particular protein families. Modifying basic parameters in the software was shown to have a significant effect on docking and virtual screening results, suggesting that expert knowledge is critical for optimizing the accuracy of these methods.

Outcome of a workshop on applications of protein models in biomedical research.

We describe the proceedings and conclusions from the "Workshop on Applications of Protein Models in Biomedical Research" (the Workshop) that was held at the University of California, San Francisco on 11 and 12 July, 2008. At the Workshop, international scientists involved with structure modeling explored (i) how models are currently used in biomedical research, (ii) the requirements and challenges for different applications, and (iii) how the interaction between the computational and experimental research communities could be strengthened to advance the field.

CONFIRM: connecting fragments found in receptor molecules.

A novel algorithm for the connecting of fragment molecules is presented and validated for a number of test systems. Within the CONFIRM (Connecting Fragments Found in Receptor Molecules) approach a pre-prepared library of bridges is searched to extract those which match a search criterion derived from known experimental or computational binding information about fragment molecules within a target binding site. The resulting bridge 'hits' are then connected, in an automated fashion, to the fragments and docked into the target receptor. Docking poses are assessed in terms of root-mean-squared deviation from the known positions of the fragment molecules, as well as docking score should known inhibitors be available. The creation of the bridge library, the full details and novelty of the CONFIRM algorithm, and the general applicability of this approach within the field of fragment-based de novo drug design are discussed.