Exploring conformational search protocols for ligand-based virtual screening and 3-D QSAR modeling.

3-D ligand conformations are required for most ligand-based drug design methods, such as pharmacophore modeling, shape-based screening, and 3-D QSAR model building. Many studies of conformational search methods have focused on the reproduction of crystal structures (i.e. bioactive conformations); however, for ligand-based modeling the key question is how to generate a ligand alignment that produces the best results for a given query molecule. In this work, we study different conformation generation modes of ConfGen and the impact on virtual screening (Shape Screening and e-Pharmacophore) and QSAR predictions (atom-based and field-based). In addition, we develop a new search method, called common scaffold alignment, that automatically detects the maximum common scaffold between each screening molecule and the query to ensure identical coordinates of the common core, thereby minimizing the noise introduced by analogous parts of the molecules. In general, we find that virtual screening results are relatively insensitive to the conformational search protocol; hence, a conformational search method that generates fewer conformations could be considered "better" because it is more computationally efficient for screening. However, for 3-D QSAR modeling we find that more thorough conformational sampling tends to produce better QSAR predictions. In addition, significant improvements in QSAR predictions are obtained with the common scaffold alignment protocol developed in this work, which focuses conformational sampling on parts of the molecules that are not part of the common scaffold.

Kernel-based partial least squares: application to fingerprint-based QSAR with model visualization.

Numerous regression-based and machine learning techniques are available for the development of linear and nonlinear QSAR models that can accurately predict biological endpoints. Such tools can be quite powerful in the hands of an experienced modeler, but too frequently a disconnect remains between the modeler and project chemist because the resulting QSAR models are effectively black boxes. As a result, learning methods that yield models that can be visualized in the context of chemical structures are in high demand. In this work, we combine direct kernel-based PLS with Canvas 2D fingerprints to arrive at predictive QSAR models that can be projected onto the atoms of a chemical structure, allowing immediate identification of favorable and unfavorable characteristics. The method is validated using binding affinities for ligands from 10 different protein targets covering 7 distinct protein families. Models with significant predictive ability (test set Q(2) > 0.5) are obtained for 6 of 10 data sets, and fingerprints are shown to consistently outperform large collections of classical physicochemical and topological descriptors. In addition, we demonstrate how a simple bootstrapping technique may be employed to obtain uncertainties that provide meaningful estimates of prediction accuracy.

Hole filling and library optimization: application to commercially available fragment libraries.

Compound libraries comprise an integral component of drug discovery in the pharmaceutical and biotechnology industries. While in-house libraries often contain millions of molecules, this number pales in comparison to the accessible space of drug-like molecules. Therefore, care must be taken when adding new compounds to an existing library in order to ensure that unexplored regions in the chemical space are filled efficiently while not needlessly increasing the library size. In this work, we present an automated method to fill holes in an existing library using compounds from an external source and apply it to commercially available fragment libraries. The method, called Canvas HF, uses distances computed from 2D chemical fingerprints and selects compounds that fill vacuous regions while not suffering from the problem of selecting only compounds at the edge of the chemical space. We show that the method is robust with respect to different databases and the number of requested compounds to retrieve. We also present an extension of the method where chemical properties can be considered simultaneously with the selection process to bias the compounds toward a desired property space without imposing hard property cutoffs. We compare the results of Canvas HF to those obtained with a standard sphere exclusion method and with random compound selection and find that Canvas HF performs favorably. Overall, the method presented here offers an efficient and effective hole-filling strategy to augment compound libraries with compounds from external sources. The method does not have any fit parameters and therefore it should be applicable in most hole-filling applications.

Rapid shape-based ligand alignment and virtual screening method based on atom/feature-pair similarities and volume overlap scoring.

Shape-based methods for aligning and scoring ligands have proven to be valuable in the field of computer-aided drug design. Here, we describe a new shape-based flexible ligand superposition and virtual screening method, Phase Shape, which is shown to rapidly produce accurate 3D ligand alignments and efficiently enrich actives in virtual screening. We describe the methodology, which is based on the principle of atom distribution triplets to rapidly define trial alignments, followed by refinement of top alignments to maximize the volume overlap. The method can be run in a shape-only mode or it can include atom types or pharmacophore feature encoding, the latter consistently producing the best results for database screening. We apply Phase Shape to flexibly align molecules that bind to the same target and show that the method consistently produces correct alignments when compared with crystal structures. We then illustrate the effectiveness of the method for identifying active compounds in virtual screening of eleven diverse targets. Multiple parameters are explored, including atom typing, query structure conformation, and the database conformer generation protocol. We show that Phase Shape performs well in database screening calculations when compared with other shape-based methods using a common set of actives and decoys from the literature.

Reliable and Accurate Solution to the Induced Fit Docking Problem for Protein-Ligand Binding.

We present a reliable and accurate solution to the induced fit docking problem for protein-ligand binding by combining ligand-based pharmacophore docking, rigid receptor docking, and protein structure prediction with explicit solvent molecular dynamics simulations. This novel methodology in detailed retrospective and prospective testing succeeded to determine protein-ligand binding modes with a root-mean-square deviation within 2.5 Å in over 90% of cross-docking cases. We further demonstrate these predicted ligand-receptor structures were sufficiently accurate to prospectively enable predictive structure-based drug discovery for challenging targets, substantially expanding the domain of applicability for such methods.

Large-scale systematic analysis of 2D fingerprint methods and parameters to improve virtual screening enrichments.

A systematic virtual screening study on 11 pharmaceutically relevant targets has been conducted to investigate the interrelation between 8 two-dimensional (2D) fingerprinting methods, 13 atom-typing schemes, 13 bit scaling rules, and 12 similarity metrics using the new cheminformatics package Canvas. In total, 157 872 virtual screens were performed to assess the ability of each combination of parameters to identify actives in a database screen. In general, fingerprint methods, such as MOLPRINT2D, Radial, and Dendritic that encode information about local environment beyond simple linear paths outperformed other fingerprint methods. Atom-typing schemes with more specific information, such as Daylight, Mol2, and Carhart were generally superior to more generic atom-typing schemes. Enrichment factors across all targets were improved considerably with the best settings, although no single set of parameters performed optimally on all targets. The size of the addressable bit space for the fingerprints was also explored, and it was found to have a substantial impact on enrichments. Small bit spaces, such as 1024, resulted in many collisions and in a significant degradation in enrichments compared to larger bit spaces that avoid collisions.

AutoQSAR: an automated machine learning tool for best-practice quantitative structure-activity relationship modeling.

AIM: We introduce AutoQSAR, an automated machine-learning application to build, validate and deploy quantitative structure-activity relationship (QSAR) models. METHODOLOGY/RESULTS: The process of descriptor generation, feature selection and the creation of a large number of QSAR models has been automated into a single workflow within AutoQSAR. The models are built using a variety of machine-learning methods, and each model is scored using a novel approach. Effectiveness of the method is demonstrated through comparison with literature QSAR models using identical datasets for six end points: protein-ligand binding affinity, solubility, blood-brain barrier permeability, carcinogenicity, mutagenicity and bioaccumulation in fish. CONCLUSION: AutoQSAR demonstrates similar or better predictive performance as compared with published results for four of the six endpoints while requiring minimal human time and expertise.

Analysis and comparison of 2D fingerprints: insights into database screening performance using eight fingerprint methods.

Virtual screening is a widely used strategy in modern drug discovery and 2D fingerprint similarity is an important tool that has been successfully applied to retrieve active compounds from large datasets. However, it is not always straightforward to select an appropriate fingerprint method and associated settings for a given problem. Here, we applied eight different fingerprint methods, as implemented in the new cheminformatics package Canvas, on a well-validated dataset covering five targets. The fingerprint methods include Linear, Dendritic, Radial, MACCS, MOLPRINT2D, Pairwise, Triplet, and Torsion. We find that most fingerprints have similar retrieval rates on average; however, each has special characteristics that distinguish its performance on different query molecules and ligand sets. For example, some fingerprints exhibit a significant ligand size dependency whereas others are more robust with respect to variations in the query or active compounds. In cases where little information is known about the active ligands, MOLPRINT2D fingerprints produce the highest average retrieval actives. When multiple queries are available, we find that a fingerprint averaged over all query molecules is generally superior to fingerprints derived from single queries. Finally, a complementarity metric is proposed to determine which fingerprint methods can be combined to improve screening results.

PHASE: a new engine for pharmacophore perception, 3D QSAR model development, and 3D database screening: 1. Methodology and preliminary results.

We introduce PHASE, a highly flexible system for common pharmacophore identification and assessment, 3D QSAR model development, and 3D database creation and searching. The primary workflows and tasks supported by PHASE are described, and details of the underlying scientific methodologies are provided. Using results from previously published investigations, PHASE is compared directly to other ligand-based software for its ability to identify target pharmacophores, rationalize structure-activity data, and predict activities of external compounds.

Induction of decision trees via evolutionary programming.

Decision trees have been used extensively in cheminformatics for modeling various biochemical endpoints including receptor-ligand binding, ADME properties, environmental impact, and toxicity. The traditional approach to inducing decision trees based upon a given training set of data involves recursive partitioning which selects partitioning variables and their values in a greedy manner to optimize a given measure of purity. This methodology has numerous benefits including classifier interpretability and the capability of modeling nonlinear relationships. The greedy nature of induction, however, may fail to elucidate underlying relationships between the data and endpoints. Using evolutionary programming, decision trees are induced which are significantly more accurate than trees induced by recursive partitioning. Furthermore, when assessed on previously unseen data in a 10-fold cross-validated manner, evolutionary programming induced trees exhibit a significantly higher accuracy on previously unseen data. This methodology is compared to single-tree and multiple-tree recursive partitioning in two domains (aerobic biodegradability and hepatotoxicity) and shown to produce less complex classifiers with average increases in predictive accuracy of 5-10% over the traditional method.

Use of robust classification techniques for the prediction of human cytochrome P450 2D6 inhibition.

A new in silico model is developed to predict cytochrome P450 2D6 inhibition from 2D chemical structure. Using a diverse training set of 100 compounds with published inhibition constants, an ensemble approach to recursive partitioning is applied to create a large number of classification trees, each of which yields a yes/no prediction about inhibition for a given compound. These binary classifications are combined to provide an overall prediction, which answers the yes/no question about inhibition and provides a measure of confidence about that prediction. Compared to single-tree models, the ensemble approach is less sensitive to noise in the experimental data as well as to changes in the training set. Internal validation tests indicated an overall classification accuracy of 75%, whereas predictions applied to an external set of 51 compounds yielded 80% accuracy, with all inhibitors correctly identified. The speed and 2D nature of this model make it appropriate for high-throughput processing of large chemical libraries, and the confidence level provides a continuous scale on which to prioritize compounds.

In silico models for the prediction of dose-dependent human hepatotoxicity.

The liver is extremely vulnerable to the effects of xenobiotics due to its critical role in metabolism. Drug-induced hepatotoxicity may involve any number of different liver injuries, some of which lead to organ failure and, ultimately, patient death. Understandably, liver toxicity is one of the most important dose-limiting considerations in the drug development cycle, yet there remains a serious shortage of methods to predict hepatotoxicity from chemical structure. We discuss our latest findings in this area and present a new, fully general in silico model which is able to predict the occurrence of dose-dependent human hepatotoxicity with greater than 80% accuracy. Utilizing an ensemble recursive partitioning approach, the model classifies compounds as toxic or non-toxic and provides a confidence level to indicate which predictions are most likely to be correct. Only 2D structural information is required and predictions can be made quite rapidly, so this approach is entirely appropriate for data mining applications and for profiling large synthetic and/or virtual libraries.

Quantum mechanical models correlating structure with selectivity: predicting the enantioselectivity of beta-amino alcohol catalysts in aldehyde alkylation.

Quantitative structure selectivity relationship (QSSR) models are described that provide consistently reliable predictions for the asymmetric addition of Et2Zn to PhCHO catalyzed by beta-amino alcohols. Statistically valid two-variable linear regression models that correlate the structures of the chiral catalysts with their enantioselectivities are obtained from three-dimensional physical property grids. The strength of the present method is that statistical models obtained from a small set of experimentally determined selectivities and relatively simple theoretical calculations yield selectivity predictions that are as accurate as those derived from higher-level calculations of transition-structure energies. Only minutes of computing time are required. Simple models are obtained which permit straightforward physical interpretation and generate realistic predictions.

Computation of the physio-chemical properties and data mining of large molecular collections.

Very large data sets of molecules screened against a broad range of targets have become available due to the advent of combinatorial chemistry. This information has led to the realization that ADME (absorption, distribution, metabolism, and excretion) and toxicity issues are important to consider prior to library synthesis. Furthermore, these large data sets provide a unique and important source of information regarding what types of molecular shapes may interact with specific receptor or target classes. Thus, the requirement for rapid and accurate data mining tools became paramount. To address these issues Pharmacopeia, Inc. formed a computational research group, The Center for Informatics and Drug Discovery (CIDD).* In this review we cover the work done by this group to address both in silico ADME modeling and data mining issues faced by Pharmacopeia because of the availability of a large and diverse collection (over 6 million discrete compounds) of drug-like molecules. In particular, in the data mining arena we discuss rapid docking tools and how we employ them, and we describe a novel data mining tool based on a ID representation of a molecule followed by a molecular sequence alignment step. For the ADME area we discuss the development and application of absorption, blood-brain barrier (BBB) and solubility models. Finally, we summarize the impact the tools and approaches might have on the drug discovery process.