ALADDIN: an integrated tool for computer-assisted molecular design and pharmacophore recognition from geometric, steric, and substructure searching of three-dimensional molecular structures

Machine learning-based classification models for non-covalent Bruton's tyrosine kinase inhibitors: predictive ability and interpretability

In this study, we built classification models using machine learning techniques to predict the bioactivity of non-covalent inhibitors of Bruton's tyrosine kinase (BTK) and to provide interpretable and transparent explanations for these predictions. To achieve this, we gathered data on BTK inhibitors from the Reaxys and ChEMBL databases, removing compounds with covalent bonds and duplicates to obtain a dataset of 3895 inhibitors of non-covalent. These inhibitors were characterized using MACCS fingerprints and Morgan fingerprints, and four traditional machine learning algorithms (decision trees (DT), random forests (RF), support vector machines (SVM), and extreme gradient boosting (XGBoost)) were used to build 16 classification models. In addition, four deep learning models were developed using deep neural networks (DNN). The best model, Model D_4, which was built using XGBoost and MACCS fingerprints, achieved an accuracy of 94.1% and a Matthews correlation coefficient (MCC) of 0.75 on the test set. To provide interpretable explanations, we employed the SHAP method to decompose the predicted values into the contributions of each feature. We also used K-means dimensionality reduction and hierarchical clustering to visualize the clustering effects of molecular structures of the inhibitors. The results of this study were validated using crystal structures, and we found that the interaction between the BTK amino acid residue and the important features of clustered scaffold was consistent with the known properties of the complex crystal structures. Overall, our models demonstrated high predictive ability and a qualitative model can be converted to a quantitative model to some extent by SHAP, making them valuable for guiding the design of new BTK inhibitors with desired activity.

A Data Resource for Sulfuric Acid Reactivity of Organic Chemicals

We describe a dataset of the quantitative reactivity of organic chemicals with concentrated sulfuric acid. As well as being a key industrial chemical, sulfuric acid is of environmental and planetary importance. In the absence of measured reaction kinetics, the reaction rate of a chemical with sulfuric acid can be estimated from the reaction rate of structurally related chemicals. To allow an approximate prediction, we have collected 589 sets of kinetic data on the reaction of organic chemicals with sulfuric acid from 262 literature sources and used a functional group-based approach to build a model of how the functional groups would react in any sulfuric acid concentration from 60–100%, and between −20 °C and 100 °C. The data set provides the original reference data and kinetic measurements, parameters, intermediate computation steps, and a set of first-order rate constants for the functional groups across the range of conditions −20 °C–100 °C and 60–100% sulfuric acid. The dataset will be useful for a range of studies in chemistry and atmospheric sciences where the reaction rate of a chemical with sulfuric acid is needed but has not been measured.

Molecular docking and structure-based drug design strategies

Pharmaceutical research has successfully incorporated a wealth of molecular modeling methods, within a variety of drug discovery programs, to study complex biological and chemical systems. The integration of computational and experimental strategies has been of great value in the identification and development of novel promising compounds. Broadly used in modern drug design, molecular docking methods explore the ligand conformations adopted within the binding sites of macromolecular targets. This approach also estimates the ligand-receptor binding free energy by evaluating critical phenomena involved in the intermolecular recognition process. Today, as a variety of docking algorithms are available, an understanding of the advantages and limitations of each method is of fundamental importance in the development of effective strategies and the generation of relevant results. The purpose of this review is to examine current molecular docking strategies used in drug discovery and medicinal chemistry, exploring the advances in the field and the role played by the integration of structure- and ligand-based methods.

Virtual screening strategies: recent advances in the identification and design of anti-cancer agents

Virtual screening (VS) is a well-established technique, which is now routinely employed in computer aided drug designing process. VS can be broadly classified into two categories, i.e., ligand-based and structure-based approach. In recent years, VS has emerged as a time saving and cost effective technique, capable of screening millions of compounds in a user friendly manner. In the area of cancer drug design, VS methods have been widely used and helped in identifying novel molecules as potential anti-cancer agents. Both ligand-based VS (LBVS) structure-based VS (SBVS) methods have been highly useful in the identification of a number of potential anti-cancer agents exhibiting activities in nanomolar range. In tune with the rapid progress in the enhancement of computational power, VS has witnessed significant change in terms of speed and hit rate and in future it is expected that VS will be a preferential alternative to high throughput screening (HTS). This review, discusses recent trends and contribution of VS in the area of anti-cancer drug discovery.

Setting the record straight: the origin of the pharmacophore concept

For over a century since the early 1900s, Paul Ehrlich was credited with originating the concept of pharmacophores. This was challenged by John Van Drie in 2007 due to the fact that Ehrlich did not use the word "pharmacophore" in his writings. Van Drie claimed that the attribution of the pharmacophore concept to Ehrlich was due to an erroneous citation made by Ariëns in a 1966 paper, and instead he claimed, Lemont B. Kier developed the pharmacophore concept (in the modern sense, as defined by the IUPAC) during 1967-1971. There are two separate issues that may have triggered this conflict. The first one is the shift in the meaning of pharmacophore from "chemical groups" to patterns of "abstract features" of a molecule that are responsible for a biological effect. Indeed, the original use of the term is different than the current definition proposed by the IUPAC. The term was redefined in 1960 by Schueler, and this modification formed the basis of IUPAC's modern definition. The second issue is the origin of the "concept" of pharmacophore. While Ehrlich's contemporaries have consistently attributed the origin of the concept to him, the issue is further complicated by the fact that Ehrlich did not use the term pharmacophore in his papers. He, instead, referred to the features of a molecule that are responsible for biological effects as toxophores, while his contemporaries were using the term pharmacophore for the same features. In this paper, we resolve any doubts about the origins of the pharmacophore concept. Our research points to Paul Ehrlich's 1898 paper for originating the concept, which identifies peripheral chemical groups in molecules responsible for binding that leads to the subsequent biological effect, and to Schueler's 1960 book that extends the concept to the modern definition where spatial patterns of abstract features of a molecule define the pharmacophore and are ultimately responsible for the biological effect.

Biasing conformational ensembles towards bioactive-like conformers for ligand-based drug design

IMPORTANCE OF THE FIELD

In silico or virtual screening has become a common practice in contemporary computer-aided drug discovery efforts and currently constitutes a reasonably mature paradigm. Application of ligand-based approaches to virtual screening requires the ability to identify the bioactive conformers of drug-like compounds as these conformers are expected to elicit the biological activity. However, given the complexity of the energy potential surfaces of such ligands and in particular those exhibiting some degree of flexibility and the limitation of contemporary energy functions, this is not an easy task.

AREAS COVERED IN THIS REVIEW

The current contribution provides an in-depth review of recent developments in the field of generating conformational ensembles of drug-like compounds with a particular emphasis of focusing such ensembles on bioactive conformers using both energy and structural criteria. The literature reviewed in this manuscript roughly covers the last decade.

WHAT THE READER WILL GAIN

Readers of this review will gain an appreciation for the complexity of identifying bioactive conformers of drug-like compounds and an exposure to the different computational methods which were developed in order to tackle this problem as well as to the remaining challenges in this field.

TAKE HOME MESSAGE

The identification of ensembles of bioactive conformers of drug-like compounds is far from being a solved problem. Recent research has advanced the field to the point where bioactive conformers could be readily identified from within conformational ensembles generated by contemporary computational tools. However, as such conformers are inevitably accompanied by many other non-relevant conformations, a focusing mechanism is required. New methods in this field are showing promise but more work is clearly needed. New research lines are proposed which are believed to enhance the performances and with it the usefulness of 3D ligand-based methods in drug discovery and development.

The Discovery of Novel Selective D1 Dopaminergic Agonists: A-68930, A-77636, A-86929, and ABT-413

The novel selective D1 dopaminergic full agonists A-68930, A-77636 were discovered by the synthesis of molecules to probe the bioactive conformation of the partial agonist SKF-38393, by the use of this information to add D1 affinity and selectivity to a screening hit, and by traditional medicinal chemistry exploration of structure-activity relationships. The subsequent design of A-86929 and ABT-413 capitalized on these results, recently disclosed agonists, and traditional medicinal chemistry.

Some Trends in Chem(o)informatics

This introductory chapter gives a brief overview of the history of cheminformatics, and then summarizes some recent trends in computing, cultures, open systems, chemical structure representation, docking, de novo design, fragment-based drug design, molecular similarity, quantitative structure-activity relationships (QSAR), metabolite prediction, the use of phamacophores in drug discovery, data reduction and visualization, and text mining. The aim is to set the scene for the more detailed exposition of these topics in the later chapters.

AutoGrow: a novel algorithm for protein inhibitor design

Due in part to the increasing availability of crystallographic protein structures as well as rapid improvements in computing power, the past few decades have seen an explosion in the field of computer-based rational drug design. Several algorithms have been developed to identify or generate potential ligands in silico by optimizing the ligand-receptor hydrogen bond, electrostatic, and hydrophobic interactions. We here present AutoGrow, a novel computer-aided drug design algorithm that combines the strengths of both fragment-based growing and docking algorithms. To validate AutoGrow, we recreate three crystallographically resolved ligands from their constituent fragments.

Pharmacophore-based discovery of ligands for drug transporters

The ability to identify ligands for drug transporters is an important step in drug discovery and development. It can both improve accurate profiling of lead pharmacokinetic properties and assist in the discovery of new chemical entities targeting transporters. In silico approaches, especially pharmacophore-based database screening methods have great potential in improving the throughput of current transporter ligand identification assays, leading to a higher hit rate by focusing in vitro testing to the most promising hits. In this review, the potential of different in silico methods in transporter ligand identification studies are compared and summarized with an emphasis on pharmacophore modeling. Various implementations of pharmacophore model generation, database compilation and flexible screening algorithms are also introduced. Recent successful utilization of database searching with pharmacophores to identify novel ligands for the pharmaceutically significant transporters hPepT1, P-gp, BCRP, MRP1 and DAT are reviewed and the challenges encountered with current approaches are discussed.

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.

Hydrophobic Molecular Similarity from MST Fractional Contributions to the Octanol/water Partition Coefficient

The use of a recently proposed hydrophobic similarity index for the alignment of molecules and the prediction of their differences in biological activity is described. The hydrophobic similarity index exploits atomic contributions to the octanol/water transfer free energy, which are evaluated by means of the fractional partitioning scheme developed within the framework of the Miertus-Scrocco-Tomasi continuum model. Those contributions are used to define global and local measures of hydrophobic similarity. The suitability of this computational strategy is examined for two series of compounds (ACAT inhibitors and 5-HT3 receptor agonists), which are aligned to maximize the global hydrophobic similarity using a Monte Carlo-simulated protocol. Indeed, the concept of local hydrophobic similarity is used to explore structure-activity relationships in a series of COX-2 inhibitors. Inspection of the 3D distribution of hydrophobic/hydrophilic contributions in the aligned molecules is valuable to identify regions of very similar hydrophobicity, which can define pharmacophoric recognition patterns. Moreover, low similar regions permit to identify structural elements that modulate the differences in activity between molecules. Finally, the quantitative relationships found between the pharmacological activity and the hydrophobic similarity index points out that not only the global hydrophobicity, but its 3D distribution, is important to gain insight into the activity of molecules.

Two‐ and three‐dimensional QSAR of carrier‐mediated transport of β‐lactam antibiotics in Caco‐2 cells

In this study, we investigated whether such a topological descriptor-based approach is suitable for predicting the carrier-mediated transport of 20 beta-lactam antibiotics that are substrates of peptide transporters. To select the molecular descriptors that can effectively predict a targeted property in QSAR analysis, the genetic algorithm-combined partial least squares approach was used. The feasibility of the two-dimensional (2D)-QSAR approach was compared with that of comparative molecular field analysis (CoMFA). The logarithm of the uptake values of 20 beta-lactam antibiotics in Caco-2 cells obtained from the literature ranged from -1.15 to 1.09 (nmol/cm2/2 h). When preliminary leave-one-out cross-validated partial least squares analyses implemented in the SYBYL/CoMFA program were conducted, the r2pred was 0.759 and the standard error of prediction (s) was 0.373. However, the 2D-QSAR approach based on Molconn-Z descriptors gave a better predictability (r2pred = 0.923, s = 0.211), where 14 descriptors were selected and the optimal number of principal components was 4. Considering that the 2D-topological descriptors are less computationally intensive and practically completely automated, the simple 2D-QSAR model is also of great importance in drug discovery settings.

Pruned Receptor Surface Models and Pharmacophores for Three-Dimensional Database Searching

A pharmacophore represents the 3D arrangement of chemical features that are shared by molecules exhibiting activity at a protein receptor. Pharmacophores are routinely used in 3D database searching for identifying potential lead compounds. The lack of shape constraints causes the query to identify compounds that could not fit into the active site. In the absence of structural information, a receptor surface model (RSM) can be used to represent the active site. The RSM consists of a surface that envelops a set of known actives after these have been aligned using their common features. When used for database searching, a RSM is overconstraining as it restricts access to regions that could be occupied by ligands, such as the solvent-protein interface or unexplored pockets. We describe a protocol for developing pruned RSMs using information gleaned from 3D quantitative structure-activity relationship (QSAR) models. We examined the performance of queries that consist of pharmacophores used alone or with pruned or unpruned RSMs by performing searches on six databases containing known actives distributed among inactives. The pruned RSMs yield an average selectivity 1.8 times greater than that for pharmacophore queries, compared to 1.6 times for unpruned RSMs. However, the pruned RSMs retrieve on average 73% of the actives identified using the pharmacophores, compared to 40% for the unpruned RSMs. As such, pruned RSMs represent a useful compromise between the high sensitivity of pharmacophores and the high selectivity of unpruned RSMs.

De Novo Generation of Molecular Structures Using Optimization To Select Graphs on a Given Lattice

A recurrent problem in organic chemistry is the generation of new molecular structures that conform to some predetermined set of structural constraints that are imposed in an endeavor to build certain required properties into the newly generated structure. An example of this is the pharmacophore model, used in medicinal chemistry to guide de novo design or selection of suitable structures from compound databases. We propose here a method that efficiently links up a selected number of required atom positions while at the same time directing the emergent molecular skeleton to avoid forbidden positions. The linkage process takes place on a lattice whose unit step length and overall geometry is designed to match typical architectures of organic molecules. We use an optimization method to select from the many different graphs possible. The approach is demonstrated in an example where crystal structures of the same (in this case rigid) ligand complexed with different proteins are available.

TGSA-Flex: Extending the capabilities of the Topo-Geometrical superposition algorithm to handle flexible molecules

In this work, an extension of the already studied Topo-Geometrical Superposition Approach (TGSA) is presented. TGSA, a general-purpose, fast, automatic, and user-intuitive three-dimensional molecular alignment procedure, was originally designed to superpose rigid molecules simply based on atomic numbers, molecular coordinates, and connectivity. The algorithm is further developed to enable handling rotations around single bonds; in this way, common structural features, which were not properly aligned due to conformational causes, can be brought together, thus improving the molecular similarity picture of the final alignment. The present procedure, implemented in Fortran 90 and named TGSA-Flex, is deeply detailed and tested over four molecular sets: amino acids, nordihydroguaiaretic acid (NDGA) derivatives, HIV-1 protease inhibitors, and 1-[2-hydroxyethoxy)methyl]-6-(phenylthio)thymine (HEPT) derivatives. TGSA-Flex performance is evaluated by means of computational time, number of superposed atoms (also comparing it with respect to the rigid approach), and index of fit between the compared structures.

QSAR analysis of the inhibition of recombinant CYP 3A4 activity by structurally diverse compounds using a genetic algorithm-combined partial least squares method

PURPOSE

To develop a quantitative structure/activity relationship (QSAR) model for predicting drug-CYP 3A4 interactions.

METHOD

The inhibitory effect of 53 structurally diverse drugs on the metabolism of 7-benzyloxy-4-trifluoromethyl coumarin (BFC) by recombinant CYP 3A4 was evaluated using a rapid microtiter plate assay. For each drug, a total of 220 two-dimensional topological indices were calculated using Molconn-Z software. Using a genetic algorithm-based partial least squares (GA-PLS) method, the desired descriptors were automatically selected to maximize the predictability of the IC50 values.

RESULTS

The IC50 values of the drugs tested ranged from 9 nM to 2 mM. Based on the GA-PLS method, five principal components derived from 20 Molconn-Z descriptors were found to be effective for QSAR modeling. Interestingly, these descriptors suggested that the molecular size would be an important factor in determining drug-CYP 3A4 interactions. In the leave-one-out prediction, the rpred and the standard error of prediction (s) were 0.754 and 0.787, respectively. Even in an external validation, the predictions were in good agreement with experimental values (rpred = 0.744, s = 0.769, n = 9).

CONCLUSIONS

The proposed model, in which two-dimensional topological descriptors were used as molecular descriptors, was able to predict drug-CYP 3A4 interactions with reasonable accuracy.

A web-based 3D-database pharmacophore searching tool for drug discovery

Three-Dimensional (3D) structural database pharmacophore searching has become a very effective approach for discovery of novel lead compounds in drug discovery. Although several commercial programs are available, these commercial programs are primarily used as a stand alone and require a local database. In recent years, the Internet has become the main medium of choice for multiuser application program distribution. Herein, we describe our development of a Web-based 3D-database pharmacophore-searching tool based on the server-client Web architecture. Both rigid and conformationally flexible searching methods are implemented. Our results show that for a typical three-center rigid pharmacophore search, the run time for searching 50 000 compounds is less than three minutes, and for four-center pharmacophore searching, the run time is less than 10 minutes on a desktop computer. For a flexible 3D-pharmacophore search, the run time for searching 50 000 compounds generally takes between one and several hours. The search results are comparable to those obtained using a commercial program. We expect that this Web-based tool will be very useful for scientists who are interested in 3D-database pharmacophore searching via the Internet.

TreeDock: a tool for protein docking based on minimizing van der Waals energies

Predicting protein-protein and protein-ligand docking remains one of the challenging topics of structural biology. The main problems are (i) to reliably estimate the binding free energies of docked states, (ii) to enumerate possible docking orientations at a high resolution, and (iii) to consider mobility of the docking surfaces and structural rearrangements upon interaction. Here we present a novel algorithm, TreeDock, that addresses the enumeration problem in a rigid-body docking search. By representing molecules as multidimensional binary search trees and by exploring a sufficient number of docking orientations such that two chosen atoms, one from each molecule, are always in contact, TreeDock is able to explore all clash-free orientations at very fine resolution in a reasonable amount of time. Due to the speed of the program, many contact pairs can be examined to search partial or complete surface areas. The deterministic systematic search of TreeDock is in contrast to most other docking programs that use stochastic searches such as Monte Carlo or simulated annealing methods. At this point, we have used the Lennard-Jones potential as the only scoring function and show that this can predict the correct docked conformation for a number of protein-protein and protein-ligand complexes. The program is most powerful if some information is known about the location of binding faces from NMR chemical-shift perturbation studies, orientation information from residual dipolar coupling, or mutational screening. The approach has the potential to include docking-site mobility by performing molecular dynamics or other randomization methods of the docking site and docking families to families of structures. The performance of the algorithm is demonstrated by docking three complexes of immunoglobulin superfamily domains, CD2 to CD58, the V(alpha) domain of a T-cell receptor to its V(beta) domain, and a T-cell receptor to a pMHC complex as well as a small molecule inhibitor to a phosphatase.

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.

One-dimensional molecular representations and similarity calculations: methodology and validation

Drug discovery research is increasingly dedicated to biological screening on a massive scale, which seems to imply a basic rejection of many computer-assisted techniques originally designed to add rationality to the early stages of discovery. While ever-faster and more clever 3D methodologies continue to be developed and rejected as alternatives to indiscriminant screening, simpler tools based on 2D structure have carved a stable niche in the high-throughput paradigm of drug discovery. Their staying power is due in no small part to simplicity, ease of use, and demonstrated ability to explain structure-activity data. This observation led us to wonder whether an even simpler view of structure might offer an advantage over existing 2D and 3D methods. Accordingly, we introduce 1D representations of chemical structure, which are generated by collapsing a 3D molecular model or a 2D chemical graph onto a single coordinate of atomic positions. Atoms along this coordinate are differentiated according to elemental type, hybridization, and connectivity. By aligning 1D representations to match up identical atom types, a measure of overall structural similarity is afforded. In extensive structure-activity validation tests, 1D similarities consistently outperform both Daylight 2D fingerprints and Cerius(2) pharmacophore fingerprints, suggesting that this new, simple means of representing and comparing structures may offer a significant advantage over existing tried-and-true methods.