Search

AbstractBioactive peptides (BPs) are short amino acid sequences that that are known to exhibit physiological characteristics such as antioxidant, antimicrobial, antihypertensive and antidiabetic properties, suggesting that they could be exploited as functional foods in the nutraceutical industry. These BPs can be derived from a variety of food sources, including milk, meat, marine, and plant proteins. In the past decade, various methods including in silico, in vitro, and in vivo techniques have been explored to unravel underlying mechanisms of BPs. To forecast interactions between peptides and their targets, in silico methods such as BIOPEP, molecular docking and Quantitative Structure-Activity Relationship modeling have been employed. Additionally, in vitro research has examined how BPs affect enzyme activities, protein expressions, and cell cultures. In vivo studies on the contrary have appraised the impact of BPs on animal models and human subjects. Hence, in the light of recent literature, this review examines the multifaceted aspects of BPs production from milk, meat, marine, and plant proteins and their potential bioactivities. We envisage that the various concepts discussed will contribute to a better understanding of the food derived BP production, which could pave a way for their potential applications in the nutraceutical industry. [ABSTRACT FROM AUTHOR]

A myriad of tobacco-associated chemicals may have possibilities to developmental/reproductive axis and endocrine-disruption impacts. Mostly they breach the biotransformation of cholesterol in mitochondria by interfering with steroidogenic pathway genes, prompting to adverse effects in steroid biosynthesis; however, studies are scanty. The quantitative structure–activity relationship (QSAR) modeling and comparative docking strategies were used to understand structural features of dataset compounds that influence developmental/reproductive toxicity and estrogen and androgen receptor–binding abilities, and to predict binding levels of toxicants with steroidogenic acute regulatory protein (StAR) and cholesterol side-chain cleavage enzyme (CYP11A1) active sites. Developed QSAR models presented good robustness and predictive ability that were determined from the applicability domain and, clustering and classification of chemicals by performing self-organizing maps. Accordingly, the exorbitant amount of polycyclic aromatic hydrocarbons (PAHs) and a limited number of other chemicals including N-nitrosamines and nicotine was represented as potential developmental/reproductive toxicants as well as estrogen and androgen receptor binders. From the docking analysis, hydrogen bonding, nonpolar, atomic π-stacking, and π-cation interactions were found between PAHs (bay and fjord structural pockets) and functional hotspot residues of StAR and CYP11A1, which strengthened the subtle structural changes at domains. These govern barrier effects to cholesterol binding and/or locking cholesterol to complicate its ejection from the Ω1 loop of StAR, and further mitigates steroid biosynthesis through cholesterol by CYP11A1; therefore, they are presumably considered as block-cluster mechanisms. These outcomes are significant to be hopeful to estimate developmental/reproductive toxicity and endocrine-disruption activities of other environmental pollutants, and could be useful for further assessment to discover binding mechanisms of PAHs with other steroidogenesis pathway genes. [ABSTRACT FROM AUTHOR]

New models for ACE2 receptor binding, based on QSAR and docking algorithms were developed, using XRD structural data and ChEMBL 26 database hits as training sets. The selectivity of the potential ACE2-binding ligands towards Neprilysin (NEP) and ACE was evaluated. The Enamine screening collection (3.2 million compounds) was virtually screened according to the above models, in order to find possible ACE2-chemical probes, useful for the study of SARS-CoV2-induced neurological disorders. An enzymology inhibition assay for ACE2 was optimized, and the combined diversified set of predicted selective ACE2-binding molecules from QSAR modeling, docking, and ultrafast docking was screened in vitro. The in vitro hits included two novel chemotypes suitable for further optimization. [ABSTRACT FROM AUTHOR]

[Display omitted] • QSAR model based on 5z-scale was established to predict ACE-I pentapeptides. • Two novel ACE-I peptides were obtained by QSAR model and in silico screening. • PLITT showed the strongest ACE-I activity. • Hydrogen bonding interactions played a critical role in ACE inhibition. • The inhibition mode of PLITT was a mixture of non-competition and competition. Food-derived angiotensin-converting enzyme-inhibitory (ACE-I) peptides have attracted extensive attention. Herein, the ACE-I peptides from Scomber japonicus muscle hydrolysates were screened, and their mechanisms of action and inhibition stability were explored. The quantitative structure–activity relationship (QSAR) model based on 5z-scale metrics was developed to rapidly screen for ACE-I peptides. Two novel potential ACE-I peptides (LTPFT, PLITT) were predicted through this model coupled with in silico screening, of which PLITT had the highest activity (IC 50 : 48.73 ± 7.59 μM). PLITT inhibited ACE activity with a mixture of non-competitive and competitive mechanisms, and this inhibition mainly contributed to the hydrogen bonding based on molecular docking study. PLITT is stable under high temperatures, pH, glucose, and NaCl. The zinc ions (Zn2+) and copper ions (Cu2+) enhanced ACE-I activity. The study suggests that the QSAR model is effective in rapidly screening for ACE-I inhibitors, and PLITT can be supplemented in foods to lower blood pressure. [ABSTRACT FROM AUTHOR]

Human African trypanosomiasis (HAT), a fatal infection caused by Trypanosoma brucei (T. brucei) is considered as a neglected disease in the tropical areas, and newer agents with unique mechanism of action are urgently needed. In this work, 65 Imidazopyridine analogues from known literature were selected for building statistically robust genetic algorithm (GA) based QSAR models. Furthermore, values for the various cross-validation properties supported its statistical robustness (model 1, R tr 2 = 0. 8 5 3 7 , RMSE cv = 0. 3 5 8 6 , MAE cv = 0. 2 6 4 2 , CCC cv = 0. 8 9 5 3 , and Q LMO 2 = 0. 7 8 8 7). Our in silico ADMET analysis revealed that a designed molecule, S10 may act as potent lead (T. brucei, pEC 5 0 (μ M), predicted = 8. 2 0 0) with better pharmacokinetics, no carcinogenicity, class III acute oral toxicity, minimal OCT1 and OCT2 inhibitions and no eye corrosion profiles. Our molecular docking analysis (on 42 drug targets) for a dataset and designed molecules demonstrated higher binding affinity of Imidazopyridine analogues with T. brucei farnesyl diphosphate synthase (TbFPPS) (PDB id: 2I19). This observation was further supported by 100 ns molecular dynamics analysis retaining better stability of complex. Thus, imidazopyridine analogues would provide a promising scaffold for the development of anti-HAT agents. Human African trypanosomiasis (HAT), a fatal infection caused by Trypanosoma brucei, is considered as a neglected disease in the tropical areas, and newer agents with unique mechanism of action are urgently needed. In this present work, 65 Imidazopyridine analogues from known literature were selected for building statistically robust genetic algorithm (GA) based QSAR models. Furthermore, values for the various cross-validation properties supported its statistical robustness. Our in-silico ADMET analysis revealed that a designed molecule, S10 may act as potent lead (T. brucei, pEC 50 (μM), predicted = 8.200) with better pharmacokinetics, no carcinogenicity, class III acute oral toxicity, minimal OCT1 and OCT2 inhibitions, and no eye corrosion profiles. [ABSTRACT FROM AUTHOR]

With the consolidation of deep learning in drug discovery, several novel algorithms for learning molecular representations have been proposed. Despite the interest of the community in developing new methods for learning molecular embeddings and their theoretical benefits, comparing molecular embeddings with each other and with traditional representations is not straightforward, which in turn hinders the process of choosing a suitable representation for Quantitative Structure–Activity Relationship (QSAR) modeling. A reason behind this issue is the difficulty of conducting a fair and thorough comparison of the different existing embedding approaches, which requires numerous experiments on various datasets and training scenarios. To close this gap, we reviewed the literature on methods for molecular embeddings and reproduced three unsupervised and two supervised molecular embedding techniques recently proposed in the literature. We compared these five methods concerning their performance in QSAR scenarios using different classification and regression datasets. We also compared these representations to traditional molecular representations, namely molecular descriptors and fingerprints. As opposed to the expected outcome, our experimental setup consisting of over |$25 000$| trained models and statistical tests revealed that the predictive performance using molecular embeddings did not significantly surpass that of traditional representations. Although supervised embeddings yielded competitive results compared with those using traditional molecular representations, unsupervised embeddings tended to perform worse than traditional representations. Our results highlight the need for conducting a careful comparison and analysis of the different embedding techniques prior to using them in drug design tasks and motivate a discussion about the potential of molecular embeddings in computer-aided drug design. [ABSTRACT FROM AUTHOR]

Growing attention has been focussed towards a considerable amount of experimental and theoretical study of anti toxoplasmosis compounds having inhibitory activities against toxoplasmosis. Toxoplasmosis is a dangerous disease in both urban as well as rural areas. It is caused by Toxoplasma gondii. The life cycle involves feline species such as both domestic and wild cats, and other felids such as lions, etc. There are various targets for developing anti toxoplasmosis agents. One of the most promising targets is dihydrofolate reductase (DHFR). 1,3,5-Triazine compounds have been reported to inhibit the T. gondii. But there is hardly any formulation of quantitative structure-activity relationship (QSAR) involving 1,3,5-triazine inhibitors to date. Therefore, it is our target in the present study to develop QSAR models based on the computed theoretical molecular descriptors to scale the essential features responsible for the DHFR inhibition. These screened features of the selected compounds will help to design the potent congeneric series. [ABSTRACT FROM AUTHOR]

Abstract A myriad of phytochemicals may have potential to lead toxicity and endocrine disruption effects by interfering with nuclear hormone receptors. In this examination, the toxicity and estrogen receptor−binding abilities of a set of 2826 phytochemicals were evaluated. The endpoints mutagenicity, carcinogenicity (both CAESAR and ISS models), developmental toxicity, skin sensitization and estrogen receptor relative binding affinity (ER_RBA) were studied using the VEGA QSAR modeling package. Alongside the predictions, models were providing possible information for applicability domains and most similar compounds as similarity sets from their training sets. This information was subjected to perform the clustering and classification of chemicals using Self−Organizing Maps. The identified clusters and their respective indicators were considered as potential hotspot structures for the specified data set analysis. Molecular screening interpretations of models were exhibited accurate predictions. Moreover, the indication sets were defined significant clusters and cluster indicators with probable prediction labels (precision). Accordingly, developed QSAR models showed good predictive abilities and robustness, which observed from applicability domains, representation spaces, clustering and classification schemes. Furthermore, the designed new path could be useful as a valuable approach to determine toxicity levels of phytochemicals and other environmental pollutants and protect the human health. Highlights • Six toxic endpoints were elucidated for a set of 2826 phytochemicals. • Prediction probabilities of data set were defined by applicability domain analysis. • The specific cluster indicators were determined by using the Self-Organizing Maps. • Cluster indicators were represented as novel hotspot structures for clusters. [ABSTRACT FROM AUTHOR]

The enormous production of fruit waste and the generation of countless organic micropollutants are serious environmental problems. To solve the problems, the biowastes, i.e., orange, mandarin, and banana peels, were used as biosorbents to remove the organic pollutants. In this application, the difficult challenge is knowing the degree of adsorption affinity of biomass for each type of micropollutant. However, since there are numerous micropollutants, it requires enormous material consumption and labor to physically estimate the adsorbability of biomass. To address this limitation, quantitative structure-adsorption relationship (QSAR) models for the adsorption assessment were established. In this process, the surface properties of each adsorbent were measured with instrumental analyzers, their adsorption affinity values for several organic micropollutants were determined through isotherm experiments, and QSAR models for each adsorbent were developed. The results showed that the tested adsorbents had significant adsorption affinity for cationic and neutral micropollutants, while the anionic one had low adsorption. As a result of the modeling, it was found that the adsorption could be predicted for a modeling set with an R2 of 0.90–0.915, and the models were validated via the prediction of a test set that was not included in the modeling set. Also, using the models, the adsorption mechanisms were identified. It is speculated that these developed models can be used to rapidly estimate adsorption affinity values for other micropollutants. • Fruit-based biowaste can be recycled for the removal of organic micropollutants. • Adsorption of cationic micropollutants by the biosorbents is notable. • QSAR models for predicting the adsorption of fruit-based biosorbents for micropollutants. • Fruit-based biosorbents mostly absorb micropollutants due to their hydrophobicity. • The adsorption characteristics of the investigated fruit-based biosorbents are comparable. [ABSTRACT FROM AUTHOR]

Ligand based virtual screening (LBVS) approaches could be broadly divided into those relying on chemical similarity searches and those employing Quantitative Structure-Activity Relationship (QSAR) models. We have compared the predictive power of these approaches using some datasets of compounds tested against several G-Protein Coupled Receptors (GPCRs). The k-Nearest Neighbors ( kNN) QSAR models were built for known ligands of each GPCR target independently, with a fraction of tested ligands for each target set aside as a validation set. The prediction accuracies of QSAR models for making active/inactive calls for compounds in both training and validation sets were compared to those achieved by the Prediction of Activity Spectra for Substances' (PASS) and the Similarity Ensemble Approach (SEA) tools both available online. Models developed with the kNN QSAR method showed the highest predictive power for almost all tested GPCR datasets. The PASS software, which incorporates multiple end-point specific QSAR models demonstrated a moderate predictive power, while SEA, a chemical similarity based approach, had the lowest prediction power. Our studies suggest that when sufficient amount of data is available to develop and rigorously validate QSAR models such models should be chosen as the preferred virtual screening tool in ligand-based computational drug discovery as compared to chemical similarity based approaches. [ABSTRACT FROM AUTHOR]

In the paper first the two main learning strategies of the artificial neural networks (ANNs), the error-back propagation (EBP) and Kohonen self-organizing maps (SOM) are briefly described. Next, two nonstandard network layouts of the ANNs (bottle-neck and pyramidal decision tree) one for each of both learning strategies are suggested. In the last part, the use of counter-propagation (CP) ANN for handling chemical structures in QSAR modeling, classification and clustering is discussed. These concepts are of particular interest for the computer-aided drug research and in computer-aided toxicology. In the case study the results of fish toxicity research are described. [ABSTRACT FROM AUTHOR]

• Novel functional soy chhurpi cheese was produced using L. delbrueckii strains. • Soy chhurpi produced using L. delbrueckii WS4 had highest yield and functionality. • Simulated in vitro GI digestion enhanced ACE-inhibitory and antioxidant effect. • Bioactive peptides with potential ACE-inhibitory and antioxidant activity detected. • Molecular docking and QSAR models predicted ACE-inhibitory peptides in soy chhurpi. A novel soy chhurpi product was developed by fermentation of soymilk using proteolytic Lactobacillus delbrueckii strains isolated from traditional chhurpi production of Sikkim Himalaya. Soymilk fermentation by L. delbrueckii WS4 was associated with the hydrolysis of globulin proteins, with observed antioxidant, and ACE-inhibitory activity which further increased upon simulated in vitro gastrointestinal digestion. Peptidomics analysis of soy chhurpi and its gastrointestinal digest resulted in the identification of bioactive peptides with ACE-inhibitory and antioxidant properties. In silico antihypertensive property prediction followed by molecular docking study demonstrated strong binding affinity of selected peptides with ACE. The glycinin-derived peptide, SVIKPPTDE escaped gastrointestinal digestion and demonstrated strong non-bond interactions with ACE catalytic residues. QSAR models predicted an ACE-inhibitory IC 50 of 21.29 µM for SVIKPPTDE. This is the first report on the production of novel functional soy chhurpi cheese using defined starter strains and the identification of bioactive peptides in undigested and gastrointestinal digested soy chhurpi. [ABSTRACT FROM AUTHOR]

Antioxidative peptides derived from food proteins have been screened and characterized in in vitro assays, and several quantitative structure-activity relationship (QSAR) models have been established to understand the structure-activity relationship of the selected antioxidant peptides. However, a systematic evaluation of each fragment that could be produced by a certain protein is still lacking. In this study, all the tripeptides that could possibly derive from bovine beta-lactoglobulin based on its amino acid sequence have been synthesized, and their antioxidant activities were measured using a ferric-reducing antioxidant power assay. Three of the peptides (i.e., LTC, CQC, and GTW) exhibited higher activities than that of glutathione. A QSAR model was also computed using multiple linear regression with divided physicochemical property scores descriptors, and the validity of the model was confirmed by cross-validation and Y-scrambling. The result of the QSAR modeling indicated that the electronic and hydrogen-bonding properties of the amino acids in the tripeptide sequences, as well as the steric properties of the amino acid residues at the C- and N-termini, played an important role in the antioxidant activities of the tripeptides. The antioxidant activities of the tripeptides were generally higher with Cys and Trp amino acid residues in the sequence. The structural characteristics we found may contribute to the further research of antioxidative mechanism. [ABSTRACT FROM AUTHOR]

The structure-acute-toxicity relationship of organic acids and their derivatives was studied. The effect of the structure of 88 organic compounds on the acute toxicity upon i.p. administration in mice was studied using 2D simplex representation of the molecular structure and partial least-squares methods. Satisfactory quantitative structure-activity relationship (QSAR) models were constructed and analyzed to reveal the structural fragments that enhance toxicity. [ABSTRACT FROM AUTHOR]

Quantitative structure-activity relationship (QSAR) models have gained popularity in the pharmaceutical industry due to their potential to substantially decrease drug development costs by reducing expensive laboratory and clinical tests. QSAR modeling consists of two fundamental steps, namely, descriptor discovery and model building. Descriptor discovery methods are either based on chemical domain knowledge or purely data-driven. The former, chemoinformatics-based, and the latter, substructures-based, methods for QSAR modeling, have been developed quite independently. As a consequence, evaluations involving both types of descriptor discovery method are rarely seen. In this study, a comparative analysis of chemoinformatics-based and substructure-based approaches is presented. Two chemoinformatics-based approaches; ECFI and SELMA, are compared to five approaches for substructure discovery; CP, graphSig, MFI, MoFa and SUBDUE, using 18 QSAR datasets. The empirical investigation shows that one of the chemo-informatics-based approaches, ECFI, results in significantly more accurate models compared to all other methods, when used on their own. Results from combining descriptor sets are also presented, showing that the addition of ECFI descriptors to any other descriptor set leads to improved predictive performance for that set, while the use of ECFI descriptors in many cases also can be improved by adding descriptors generated by the other methods. [ABSTRACT FROM AUTHOR]

Abstract: Exposure to chemicals occurs often as mixtures. Presented in this paper is information on alkoxyethanols and the impact they might have on human health in combination with some commonly found aliphatic and aromatic compounds. Our studies to evaluate the joint toxicity of these chemicals among themselves and in combination with other chemicals reveal a variety of possible outcomes depending on the exposure scenario. The interactions are predominantly based on metabolic pathways and are common among several solvents and organic compounds. Quantitative structure activity relationship (QSAR) analysis can be used with high confidence to identify chemicals that will interact to influence overall joint toxicity. Potential human exposure to a combination of alkoxyethanol, toluene and substituted benzenes may increase reproductive and developmental disease conditions. Inheritable gene alterations result in changes in the enzyme function in different subpopulations causing variations in quantity and/or quality of particular isoenzymes. These changes are responsible for differential metabolism of chemicals in species, genders, and life stages and are often the basis of a population’s susceptibility. Unique genotypes introduced as a function of migration can alter the genetic makeup of any given population. Hence special consideration should be given to susceptible populations while conducting chemical health risk assessments. [Copyright &y& Elsevier]

Abstract: Validated predictive QSAR modeling was done on some N-aryl-oxazolidinone-5-carboxamides for higher anti-HIV protease activities. Stepwise regression developed significant models showing importance of atom based descriptors like RTSA indices, Wang–Ford charges and different whole molecular descriptors. The true predictabilities of QSAR models were justified by challenging these against an external dataset. A representative high active compound was predicted by this modeling. It showed that internal and external validations may lead to the same conclusion. [Copyright &y& Elsevier]

Abstract: The applicability of the newly developed RM1 and PM6 methods implemented in the semiempirical quantum chemistry mopac2009™ software package in modeling free radical scavenging activity of flavonoids was examined. Bond dissociation enthalpy (BDE) of OH groups could be calculated much faster than with DFT method but with similar quality. Despite the known shortcomings of the Trolox equivalent antioxidant capacity (TEAC) assay, we show that taking into account the hydrogen atom transfer (HAT) mechanism of free radical scavenging of flavonoids encoded by minimal BDE values (BDEmin) and the number of OH groups (nOH), as well as experimental data, reasonable QSAR models could be developed. For TEAC values of 38 flavonoids measured by the ABTS free radical, a model based on BDEmin and nOH was developed, having very good statistical parameters (r =0.983, r cv =0.976). The applicability of this model to three different data sets of flavonoids and reliability of TEAC values measured in distinct laboratories were discussed. Finally, a reasonably good model of experimental vitamin C equivalent antioxidant capacity (VCEAC) of 36 flavonoids was obtained (r =0.954, r cv =0.947), involving BDEmin and nOH as descriptors. Additionally, all presented models have comparable fit and cross-validated statistical parameters, as well as significant regression coefficients. [Copyright &y& Elsevier]

In this article, we present an automatic model generation process for building QSAR models using Gaussian Processes, a powerful machine learning modeling method. We describe the stages of the process that ensure models are built and validated within a rigorous framework: descriptor calculation, splitting data into training, validation and test sets, descriptor filtering, application of modeling techniques and selection of the best model. We apply this automatic process to data sets of blood–brain barrier penetration and aqueous solubility and compare the resulting automatically generated models with ‘manually’ built models using external test sets. The results demonstrate the effectiveness of the automatic model generation process for two types of data sets commonly encountered in building ADME QSAR models, a small set of in vivo data and a large set of physico-chemical data. [ABSTRACT FROM AUTHOR]

A combined approach of validated QSAR modeling and virtual screening was successfully applied to the discovery of novel tylophrine derivatives as anticancer agents. QSAR models have been initially developed for 52 chemically diverse phenanthrine-based tylophrine derivatives (PBTs) with known experimental EC50 using chemical topological descriptors (calculated with the MolConnZ program) and variable selection k nearest neighbor ( kNN) method. Several validation protocols have been applied to achieve robust QSAR models. The original dataset was divided into multiple training and test sets, and the models were considered acceptable only if the leave-one-out cross-validated R 2 ( q 2) values were greater than 0.5 for the training sets and the correlation coefficient R 2 values were greater than 0.6 for the test sets. Furthermore, the q 2 values for the actual dataset were shown to be significantly higher than those obtained for the same dataset with randomized target properties (Y-randomization test), indicating that models were statistically significant. Ten best models were then employed to mine a commercially available ChemDiv Database (ca. 500 K compounds) resulting in 34 consensus hits with moderate to high predicted activities. Ten structurally diverse hits were experimentally tested and eight were confirmed active with the highest experimental EC50 of 1.8 μM implying an exceptionally high hit rate (80%). The same ten models were further applied to predict EC50 for four new PBTs, and the correlation coefficient ( R 2) between the experimental and predicted EC50 for these compounds plus eight active consensus hits was shown to be as high as 0.57. Our studies suggest that the approach combining validated QSAR modeling and virtual screening could be successfully used as a general tool for the discovery of novel biologically active compounds. [ABSTRACT FROM AUTHOR]

α-Glucosidase is considered a prime drug target for Diabetes Mellitus and its inhibitors are used to delay carbohydrate digestion for the treatment of diabetes mellitus. With the aim to design α-glucosidase inhibitors with novel chemical scaffolds, three folds ligand and structure based virtual screening was applied. Initially linear quantitative structure activity relationship (QSAR) model was developed by a molecular operating environment (MOE) using a training set of thirty-two known inhibitors, which showed good correlation coefficient (r2 = 0.88), low root mean square error (RMSE = 0.23), and cross-validated correlation coefficient r2 (q2 = 0.71 and RMSE = 0.31). The model was validated by predicting the biological activities of the test set which depicted r2 value of 0.82, indicating the robustness of the model. For virtual screening, compounds were retrieved from zinc is not commercial (ZINC) database and screened by molecular docking. The best docked compounds were chosen to assess their pharmacokinetic behavior. Later, the α-glucosidase inhibitory potential of the selected compounds was predicted by their mode of binding interactions. The predicted pharmacokinetic profile, docking scores and protein-ligand interactions revealed that eight compounds preferentially target the catalytic site of α-glucosidase thus exhibit potential α-glucosidase inhibition in silico. The α-glucosidase inhibitory activities of those Hits were predicted by QSAR model, which reflect good inhibitory activities of these compounds. These results serve as a guidelines for the rational drug design and development of potential novel anti-diabetic agents. [ABSTRACT FROM AUTHOR]

The performance of quantitative structure–activity relationship (QSAR) models largely depends on the relevance of the selected molecular representation used as input data matrices. This work presents a thorough comparative analysis of two main categories of molecular representations (vector space and metric space) for fitting robust machine learning models in QSAR problems. For the assessment of these methods, seven different molecular representations that included RDKit descriptors, five different fingerprints types (MACCS, PubChem, FP2-based, Atom Pair, and ECFP4), and a graph matching approach (non-contiguous atom matching structure similarity; NAMS) in both vector space and metric space, were subjected to state-of-art machine learning methods that included different dimensionality reduction methods (feature selection and linear dimensionality reduction). Five distinct QSAR data sets were used for direct assessment and analysis. Results show that, in general, metric-space and vector-space representations are able to produce equivalent models, but there are significant differences between individual approaches. The NAMS-based similarity approach consistently outperformed most fingerprint representations in model quality, closely followed by Atom Pair fingerprints. To further verify these findings, the metric space-based models were fitted to the same data sets with the closest neighbors removed. These latter results further strengthened the above conclusions. The metric space graph-based approach appeared significantly superior to the other representations, albeit at a significant computational cost. [ABSTRACT FROM AUTHOR]