Molecular Similarity Measures Research Articles

Matching Polynomial-Based Similarity Matrices and Descriptors for Isomers of Fullerenes

I have computed the matching polynomials of a number of isomers of fullerenes of various sizes with the objective of developing molecular descriptors and similarity measures for isomers of fullerenes on the basis of their matching polynomials. Two novel matching polynomial-based topological descriptors are developed, and they are demonstrated to have the discriminating power to contrast a number of closely related isomers of fullerenes. The number of ways to place up to seven disjoint dimers on fullerene isomers are shown to be identical, as they are not structure-dependent. Moreover, similarity matrices that provide quantitative similarity measures among a given set of isomers of fullerenes are developed from their matching polynomials and are shown to provide robust quantitative measures of similarity.

Molecular Similarity for Drug Discovery, Target Prediction and Chemical Space Visualization

Similar drug molecules often have similar properties and activities. Therefore, quantifying molecular similarity is central to drug discovery and optimization. Here I review computational methods using molecular similarity measures developed in my group within the interdisciplinary network NCCR TransCure investigating the physiology, structural biology and pharmacology of ion channels and membrane transporters. We designed a 3D molecular shape and pharmacophore comparison algorithm to optimize weak and unselective inhibitors by scaffold hopping and discovered potent and selective inhibitors of the ion channels TRPV6 and TRPM4, of endocannabinoid membrane transport, and of the divalent metal transporters DMT1 and ZIP8. We predicted off-target effects by combining molecular similarity searches from different molecular fingerprints against target annotated compounds from the ChEMBL database. Finally, we created interactive chemical space maps reflecting molecular similarities to facilitate the selection of screening compounds and the analysis of screening results. These different tools are available online at https://gdb.unibe.ch/tools/.

Control of chirality, bond flexing and anharmonicity in an electric field

AbstractWe located “hidden” S‐character chirality in formally achiral glycine using a vector‐based interpretation of the total electronic charge density distribution. We induced the formation of stereoisomers in glycine by the application of an electric field. Control of chirality was indicated from the proportionate response to a non‐structurally distorting electric field. The bond‐flexing was determined to be a measure of bond strain, which could be a factor of three lower or higher, depending on the direction of the electric field, than in the absence of the electric field. The bond‐anharmonicity was found to be approximately independent of the electric field. We also compared the formally achiral glycine with the chiral molecules alanine and lactic acid, quantifying the preferences for the S and R stereoisomers. The proportional response of the chiral discrimination to the magnitude and direction of the applied electric field indicated use of the chirality discrimination as a molecular similarity measure.

Neural network activation similarity: a new measure to assist decision making in chemical toxicology.

Deep learning neural networks, constructed for the prediction of chemical binding at 79 pharmacologically important human biological targets, show extremely high performance on test data (accuracy 92.2 ± 4.2%, MCC 0.814 ± 0.093 and ROC-AUC 0.96 ± 0.04). A new molecular similarity measure, Neural Network Activation Similarity, has been developed, based on signal propagation through the network. This is complementary to standard Tanimoto similarity, and the combined use increases confidence in the computer's prediction of activity for new chemicals by providing a greater understanding of the underlying justification. The in silico prediction of these human molecular initiating events is central to the future of chemical safety risk assessment and improves the efficiency of safety decision making.

Ligand-Based Virtual Screening Using Graph Edit Distance as Molecular Similarity Measure.

Extended reduced graphs provide summary representations of chemical structures using pharmacophore-type node descriptions to encode the relevant molecular properties. Commonly used similarity measures using reduced graphs convert these graphs into 2D vectors like fingerprints, before chemical comparisons are made. This study investigates the effectiveness of a graph-only driven molecular comparison by using extended reduced graphs along with graph edit distance methods for molecular similarity calculation as a tool for ligand-based virtual screening applications, which estimate the bioactivity of a chemical on the basis of the bioactivity of similar compounds. The results proved to be very stable and the graph editing distance method performed better than other methods previously used on reduced graphs. This is exemplified with six publicly available data sets: DUD-E, MUV, GLL&GDD, CAPST, NRLiSt BDB, and ULS-UDS. The screening and statistical tools available on the ligand-based virtual screening benchmarking platform and the RDKit were also used. In the experiments, our method performed better than other molecular similarity methods which use array representations in most cases. Overall, it is shown that extended reduced graphs along with graph edit distance is a combination of methods that has numerous applications and can identify bioactivity similarities in a structurally diverse group of molecules.

Molecular fingerprint-derived similarity measures for toxicological read-across: Recommendations for optimal use

Computational approaches are increasingly used to predict toxicity due, in part, to pressures to find alternatives to animal testing. Read-across is the “new paradigm” which aims to predict toxicity by identifying similar, data rich, source compounds. This assumes that similar molecules tend to exhibit similar activities i.e. molecular similarity is integral to read-across. Various of molecular fingerprints and similarity measures may be used to calculate molecular similarity. This study investigated the value and concordance of the Tanimoto similarity values calculated using six widely used fingerprints within six toxicological datasets. There was considerable variability in the similarity values calculated from the various molecular fingerprints for diverse compounds, although they were reasonably concordant for homologous series acting via a common mechanism. The results suggest generic fingerprint-derived similarities are likely to be optimally predictive for local datasets, i.e. following sub-categorisation. Thus, for read-across, generic fingerprint-derived similarities are likely to be most predictive after chemicals are placed into categories (or groups), then similarity is calculated within those categories, rather than for a whole chemically diverse dataset.

SimCAL: a flexible tool to compute biochemical reaction similarity

BackgroundComputation of reaction similarity is a pre-requisite for several bioinformatics applications including enzyme identification for specific biochemical reactions, enzyme classification and mining for specific inhibitors. Reaction similarity is often assessed at either two levels: (i) comparison across all the constituent substrates and products of a reaction, reaction level similarity, (ii) comparison at the transformation center with various degrees of neighborhood, transformation level similarity. Existing reaction similarity computation tools are designed for specific applications and use different features and similarity measures. A single system integrating these diverse features enables comparison of the impact of different molecular properties on similarity score computation.ResultsTo address these requirements, we present SimCAL, an integrated system to calculate reaction similarity with novel features and capability to perform comparative assessment. SimCAL provides reaction similarity computation at both whole reaction level and transformation level. Novel physicochemical features such as stereochemistry, mass, volume and charge are included in computing reaction fingerprint. Users can choose from four different fingerprint types and nine molecular similarity measures. Further, a comparative assessment of these features is also enabled. The performance of SimCAL is assessed on 3,688,122 reaction pairs with Enzyme Commission (EC) number from MetaCyc and achieved an area under the curve (AUC) of > 0.9. In addition, SimCAL results showed strong correlation with state-of-the-art EC-BLAST and molecular signature based reaction similarity methods.ConclusionsSimCAL is developed in java and is available as a standalone tool, with intuitive, user-friendly graphical interface and also as a console application. With its customizable feature selection and similarity calculations, it is expected to cater a wide audience interested in studying and analyzing biochemical reactions and metabolic networks.

Understanding the comparative molecular field analysis (CoMFA) in terms of molecular quantum similarity and DFT-based reactivity descriptors.

The three-dimensional quantitative structure-activity relationship (3D QSAR) models have many applications, although the inherent complexity to understand the results coming from 3D-QSAR arises the necessity of new insights in the interpretation of them. Hence, the quantum similarity field as well as reactivity descriptors based on the density functional theory were used in this work as a consistent approach to better understand the 3D-QSAR studies in drug design. For this purpose, the quantification of steric and electrostatic effects on a series of bicycle [4.1.0] heptane derivatives as melanin-concentrating hormone receptor 1 antagonists were performed on the basis of molecular quantum similarity measures. The maximum similarity superposition and the topo-geometrical superposition algorithms were used as molecular alignment methods to deal with the problem of relative molecular orientation in quantum similarity. In addition, a chemical reactivity analysis using global and local descriptors such as chemical hardness, softness, electrophilicity, and Fukui functions, was developed. Overall, our results suggest that the application of this methodology in drug design can be useful when the receptor is known or even unknown.

CoRILISA: a local similarity based receptor dependent QSAR method.

Molecular similarity methods have played a crucial role in the success of structure-based and computer-assisted drug design. However, with the exception of CoMSIA, the current approaches for estimating molecular similarity yield a global picture thereby providing limited information about the local spatial molecular features responsible for the variation of activity with the 3D structure. Application of molecular similarity measures, each related to the functional "pieces" of a ligand-receptor complex, is advantageous over a composite molecular similarity alone and will provide more insights to rationally interpret the activity based on the receptor and ligand structural features. Building on the ideas of our previously published methodologies-CoRIA and LISA, we present here a local molecular similarity based receptor dependent QSAR method termed CoRILISA which is a hybrid of the two approaches. The method improves on previous techniques by inclusion of receptor attributes for the calculation and comparison of similarity between molecules. For validation studies, the CoRILISA methodology was applied on three large and diverse data sets-glycogen phosphorylase b (GPb), human immunodeficiency virus-1 protease (HIV PR), and cyclin dependent kinase 2 (CDK2) inhibitors. The statistics of the CoRILISA models were benchmarked against the standard CoRIA approach and with other published approaches. The CoRILISA models were found to be significantly better, especially in terms of the predictivity for the test set. CoRILISA is able to identify the thermodynamic properties associated with residues that define the active site and modulate the variation in the activity of the molecules. It is a useful tool in the fragment-based drug discovery approach for ligand activity prediction.

Molecular Similarity Searching Method Based on Adaptive IR Technique

An age-old question remains as an open research problem in the field of chemoinformatics, which is how much could the proposed approach enhance the effectiveness of lead-discovery programmes? Answering that question is a target of any new virtual screening approach. The current research tries to contribute in this direction by improving the performance of molecular similarity searching process. In this paper, Okapi similarity measure, which is effective and widely used in text retrieval, is adapted to perform the role of molecular similarity measure in 2D fingerprints. The adapted similarity measure calculates the molecular similarity between a reference structure and a database structure. The experimental results showed that the proposed method performs well compared to Tanimoto coefficient.

COSMOsim3D: 3D-Similarity and Alignment Based on COSMO Polarization Charge Densities

COSMO σ-surfaces resulting from quantum chemical calculations of molecules in a simulated conductor, and their histograms, the so-called σ-profiles, are widely proven to provide a very suitable and almost complete basis for the description of molecular interactions in condensed systems. The COSMOsim method therefore introduced a global measure of molecular similarity on the basis of similarity of σ-profiles, but it had the disadvantage of neglecting the 3D distribution of molecular polarities, which is crucially determining all ligand-receptor binding. This disadvantage is now overcome by COSMOsim3D, which is a logical and physically sound extension of the COSMOsim method, which uses local σ-profiles on a spatial grid. This new method is used to measure intermolecular similarities on the basis of the 3D representation of the surface polarization charge densities σ of the target and the probe molecule. The probe molecule is translated and rotated in space in order to maximize the sum of local σ-profile similarities between target and probe. This sum, the COSMOsim3D similarity, is a powerful descriptor of ligand similarity and allows for a good discrimination between bioisosters and random pairs. Validation experiments using about 600 pharmacological activity classes in the MDDR database are given. Furthermore, COSMOsim3D represents a unique and very robust method for a field-based ligand-ligand alignment.

Analysis and Study of Molecule Data Sets Using Snowflake Diagrams of Weighted Maximum Common Subgraph Trees

Isomorphism measures based on the maximum common subgraph (MCS) calculation are widely used in computational chemistry for classifying, screening, and predicting properties and biological activity within chemical databases. The development of a weighted hierarchical structure based on the MCS is described in this paper. Furthermore, a 2D representation model is proposed as the proper tool for the study and preliminary analysis of molecule data sets. The development process of the weighted MCS tree is open to the use of different approaches. By taking into account different molecular descriptors, similarity and distance measures in the weighted MCS tree, the relationships between the molecular property or the activity, and the variables considered for the building and display of the weighted MCS tree can be observed. Besides that, the representation model based on snowflake diagrams allows to display of those relationships as well as shows any existing degeneration, in order to detect any possible outlier that could be obtained during the development of predictive models and to extract new variables that can be used in the building of quantitative structure-activity relationship models.

ChemInform Abstract: Molecular Quantum Similarity Measures Tuned 3D QSAR: An Antitumoral Family Validation Study.

In this work, a new methodology to construct a tuned QSAR model is presented, which is based on a convex set formalism. The present procedure continues previous 3D QSAR studies, performed using molecular quantum similarity measures (MQSM). With this new computational tool, the efficiency of MQSM applied to QSAR analysis is significantly improved. A reliable QSAR model is obtained using convex linear combinations of different kinds of MQSM, corresponding to different quantum-mechanical operators related to the quantum similarity integral. The active compounds studied here, as a case study, are a set of antitumor agents, the camptothecin molecule and analogues, and the property evaluated is the topoisomerase-I inhibition activity. Before performing a tuned QSAR analysis with this particular molecular set, a simple QSAR study for all the different possible types of MQSM is carried out. In addition, another application of MQSM is presented, to determine which method can be used to optimize molecular structure...

Graph Kernels for Molecular Similarity.

Molecular similarity measures are important for many cheminformatics applications like ligand-based virtual screening and quantitative structure-property relationships. Graph kernels are formal similarity measures defined directly on graphs, such as the (annotated) molecular structure graph. Graph kernels are positive semi-definite functions, i.e., they correspond to inner products. This property makes them suitable for use with kernel-based machine learning algorithms such as support vector machines and Gaussian processes. We review the major types of kernels between graphs (based on random walks, subgraphs, and optimal assignments, respectively), and discuss their advantages, limitations, and successful applications in cheminformatics.

Filtering and Counting of Extended Connectivity Fingerprint Features Maximizes Compound Recall and the Structural Diversity of Hits

Extended connectivity fingerprints produce variable numbers of structural features for molecules and quantitative comparison of feature ensembles is typically carried out as a measure of molecular similarity. As an alternative way to utilize the information content of extended connectivity fingerprint features, we have introduced a compound class-directed feature filtering technique. In combination with a simple feature counting protocol, feature filtering significantly improves the performance of extended connectivity fingerprint similarity searching compared with state-of-the-art fingerprint search methods. Subsets of extended connectivity fingerprint features that are unique to active compounds are found to be responsible for high compound recall. Moreover, feature filtering and counting is shown to result in significantly higher scaffold hopping potential than data fusion or fingerprint averaging methods. Extended connectivity fingerprint feature filtering and counting represents one of the simplest similarity search methods introduced to date, yet it produces top compound recall and maximizes the scaffold diversity of hits, which is a longstanding goal of similarity searching.

Quantitative assessment of the effect of basis set superposition error on the electron density of molecular complexes by means of quantum molecular similarity measures

AbstractThe Chemical Hamiltonian Approach (CHA) method is applied to obtain Basis Set Superposition Error (BSSE)‐free molecular orbitals at the Hartree–Fock (HF) and Density Functional Theory (DFT) levels of theory. To assess qualitatively the effect of the BSSE on the first‐order electron density, we had previously applied Bader's analysis of the intermolecular critical points located on the electron density, as well as density difference maps for several hydrogen bonded complexes. In this work, Quantum Molecular Similarity Measures are probed as an alternative avenue to properly quantify the electronic relaxation due to the BSSE removal by means of distance indices between the uncorrected and corrected charge densities. It is shown that BSSE contamination is more important at the DFT level of theory, and in some cases, changes on the topology of the electron density are observed upon BSSE correction. Inclusion of diffuse functions have been found to dramatically decrease the BSSE effect in both geometry and electron density. The CHA method represents a good compromise to obtain accurate results with small basis sets. © 2009 Wiley Periodicals, Inc. Int J Quantum Chem, 2009

A Similarity Search Using Molecular Topological Graphs

A molecular similarity measure has been developed using molecular topological graphs and atomic partial charges. Two kinds of topological graphs were used. One is the ordinary adjacency matrix and the other is a matrix which represents the minimum path length between two atoms of the molecule. The ordinary adjacency matrix is suitable to compare the local structures of molecules such as functional groups, and the other matrix is suitable to compare the global structures of molecules. The combination of these two matrices gave a similarity measure. This method was applied to in silico drug screening, and the results showed that it was effective as a similarity measure.

Parameterization and Conformational Sampling Effects in Pharmacophore Multiplet Searching

Pharmacophore patterns in ligands can be effectively characterized in terms of their constituent pharmacophore multiplets. Bitsets (fingerprints) encoding which particular multiplets are found in a given ligand have been and continue to be used as molecular descriptors in a range of molecular modeling applications, from ligand alignment and diversity analysis to pharmacophore-based flexible searching. Being able to create, store, and manipulate multiplets in compressed form - as bitmaps - has made it possible to integrate them into high-throughput technologies. A number of key parameters affect how well multiplets perform, including the granularity of edge length binning; how different multiplets are weighted in creating hypotheses from multiple ligands; and the number of bits that should be included in a pharmacophore hypothesis. The similarity metric employed for bitmap comparisons also affects search performance, as does the conformational sampling regime used for characterizing flexible molecules. In this report we explore the effect of parameter variation on within- and between-class similarity across seven different pharmacological classes and introduce a new measure of molecular similarity - the asymmetric stochastic cosine - uniquely suited to searching a database for matches to query hypotheses deduced from multiple ligands. Surprisingly, it turns out that the most discriminating bitmaps are obtained using relatively few conformers. The extreme discrimination power seen for single conformers, however, seems to reflect consistent effects of 2D connectivity on the 3D structure obtained. Conformational sampling by systematic search reinforces such circumstantial discrimination and should be avoided. The potential for systematic bias becomes clear when the behavior of otherwise similar conformational ensembles created by local energy minimization or by random sampling is considered. Consolidating information from multiple known actives or establishing single "bioactive" conformations a priori are safer ways to improve discrimination in pharmacophoric multiplet searching.

Brominated phenols as auxin-like molecules

2,6-Dibromophenol (DBP) was reported as an auxin-like molecule using molecular quantum similarity measures. In this study, the auxin activity of this molecule and its chlorinated homologue is further determined using a bacterial biosensor: the auxin-inducible ipdC promoter of Azospirillum brasilense. We were able to demonstrate that DBP can induce gene expression, but to a lesser extent than the auxin indole-3-acetic acid (IAA) and that DBP is not an antagonist for the IAA signalling pathway. To investigate the role of the bromine groups, the molecule 2,6-dichlorophenol (DCP) was also tested for gene expression induction. However, no induction could be observed. In a second part, DBP and other molecules were modelled in the auxin-binding pocket of the plant auxin receptor TIR1 to evaluate theoretical binding energies. Both DBP and its chlorinated homologue DCP are not strong ligands compared to other known auxins such as IAA. The importance of a carboxylated side chain for optimal binding (and probably auxin activity) was demonstrated.

BLASTing small molecules—statistics and extreme statistics of chemical similarity scores

Motivation: Small organic molecules, from nucleotides and amino acids to metabolites and drugs, play a fundamental role in chemistry, biology and medicine. As databases of small molecules continue to grow and become more open, it is important to develop the tools to search them efficiently. In order to develop a BLAST-like tool for small molecules, one must first understand the statistical behavior of molecular similarity scores.Results: We develop a new detailed theory of molecular similarity scores that can be applied to a variety of molecular representations and similarity measures. For concreteness, we focus on the most widely used measure—the Tanimoto measure applied to chem-ical fingerprints. In both the case of empirical fingerprints and fingerprints generated by several stochastic models, we derive accurate approximations for both the distribution and extreme value distribution of similarity scores. These approximation are derived using a ratio of correlated Gaussians approach. The theory enables the calculation of significance scores, such as Z-scores and P-values, and the estimation of the top hits list size. Empirical results obtained using both the random models and real data from the ChemDB database are given to corroborate the theory and show how it can be applied to mine chemical space.Availability: Data and related resources are available through http://cdb.ics.uci.eduContact: pfbaldi@ics.uci.edu