Investigating the influence of data splitting on the predictive ability of QSAR/QSPR models

Machine learning-based QSAR and LB-PaCS-MD guided design of SARS-CoV-2 main protease inhibitors

The global outbreak of the COVID-19 pandemic caused by the SARS-CoV-2 virus had led to profound respiratory health implications. This study focused on designing organoselenium-based inhibitors targeting the SARS-CoV-2 main protease (Mpro). The ligand-binding pathway sampling method based on parallel cascade selection molecular dynamics (LB-PaCS-MD) simulations was employed to elucidate plausible paths and conformations of ebselen, a synthetic organoselenium drug, within the Mpro catalytic site. Ebselen effectively engaged the active site, adopting proximity to H41 and interacting through the benzoisoselenazole ring in a π-π T-shaped arrangement, with an additional π-sulfur interaction with C145. In addition, the ligand-based drug design using the QSAR with GFA-MLR, RF, and ANN models were employed for biological activity prediction. The QSAR-ANN model showed robust statistical performance, with an r2training exceeding 0.98 and an RMSEtest of 0.21, indicating its suitability for predicting biological activities. Integration the ANN model with the LB-PaCS-MD insights enabled the rational design of novel compounds anchored in the ebselen core structure, identifying promising candidates with favorable predicted IC50 values. The designed compounds exhibited suitable drug-like characteristics and adopted an active conformation similar to ebselen, inhibiting Mpro function. These findings represent a synergistic approach merging ligand and structure-based drug design; with the potential to guide experimental synthesis and enzyme assay testing.

Py-CoMFA, docking, and molecular dynamics simulations of Leishmania (L.) amazonensis arginase inhibitors

Leishmaniasis is a disease caused by a protozoan of the genus Leishmania, affecting millions of people, mainly in tropical countries, due to poor social conditions and low economic development. First-line chemotherapeutic agents involve highly toxic pentavalent antimonials, while treatment failure is mainly due to the emergence of drug-resistant strains. Leishmania arginase (ARG) enzyme is vital in pathogenicity and contributes to a higher infection rate, thus representing a potential drug target. This study helps in designing ARG inhibitors for the treatment of leishmaniasis. Py-CoMFA (3D-QSAR) models were constructed using 34 inhibitors from different chemical classes against ARG from L. (L.) amazonensis (LaARG). The 3D-QSAR predictions showed an excellent correlation between experimental and calculated pIC50 values. The molecular docking study identified the favorable hydrophobicity contribution of phenyl and cyclohexyl groups as substituents in the enzyme allosteric site. Molecular dynamics simulations of selected protein-ligand complexes were conducted to understand derivatives' interaction modes and affinity in both active and allosteric sites. Two cinnamide compounds, 7g and 7k, were identified, with similar structures to the reference 4h allosteric site inhibitor. These compounds can guide the development of more effective arginase inhibitors as potential antileishmanial drugs.

Machine learning driven web-based app platform for the discovery of monoamine oxidase B inhibitors

Monoamine oxidases (MAOs), specifically MAO-A and MAO-B, play important roles in the breakdown of monoamine neurotransmitters. Therefore, MAO inhibitors are crucial for treating various neurodegenerative disorders, including Parkinson's disease (PD), Alzheimer's disease (AD), and amyotrophic lateral sclerosis (ALS). In this study, we developed a novel cheminformatics pipeline by generating three diverse molecular feature-based machine learning-assisted quantitative structural activity relationship (ML-QSAR) models concerning MAO-B inhibition. PubChem fingerprints, substructure fingerprints, and one-dimensional (1D) and two-dimensional (2D) molecular descriptors were implemented to unravel the structural insights responsible for decoding the origin of MAO-B inhibition in 249 non-reductant molecules. Based on a random forest ML algorithm, the final PubChem fingerprint, substructure fingerprint, and 1D and 2D molecular descriptor prediction models demonstrated significant robustness, with correlation coefficients of 0.9863, 0.9796, and 0.9852, respectively. The significant features of each predictive model responsible for MAO-B inhibition were extracted using a comprehensive variance importance plot (VIP) and correlation matrix analysis. The final predictive models were further developed as a web application, MAO-B-pred ( https://mao-b-pred.streamlit.app/ ), to allow users to predict the bioactivity of molecules against MAO-B. Molecular docking and dynamics studies were conducted to gain insight into the atomic-level molecular interactions between the ligand-receptor complexes. These findings were compared with the structural features obtained from the ML-QSAR models, which supported the mechanistic understanding of the binding phenomena. The presented models have the potential to serve as tools for identifying crucial molecular characteristics for the rational design of MAO-B target inhibitors, which may be used to develop effective drugs for neurodegenerative disorders.

MASSA Algorithm: an automated rational sampling of training and test subsets for QSAR modeling

QSAR models capable of predicting biological, toxicity, and pharmacokinetic properties were widely used to search lead bioactive molecules in chemical databases. The dataset's preparation to build these models has a strong influence on the quality of the generated models, and sampling requires that the original dataset be divided into training (for model training) and test (for statistical evaluation) sets. This sampling can be done randomly or rationally, but the rational division is superior. In this paper, we present MASSA, a Python tool that can be used to automatically sample datasets by exploring the biological, physicochemical, and structural spaces of molecules using PCA, HCA, and K-modes. The proposed algorithm is very useful when the variables used for QSAR are not available or to construct multiple QSAR models with the same training and test sets, producing models with lower variability and better values for validation metrics. These results were obtained even when the descriptors used in the QSAR/QSPR were different from those used in the separation of training and test sets, indicating that this tool can be used to build models for more than one QSAR/QSPR technique. Finally, this tool also generates useful graphical representations that can provide insights into the data.

Application of chemometrics tools for removal of crystal violet and methylene blue in binary solution by eco-friendly magnetic adsorbent modified on Heracleum persicum waste

Synthetic dyes can be hazardous to the ecosystem, even at low concentrations in the effluent. In this research, the Heracleum persicum stems-Fe3O4 (MHPS) adsorbent performance for the removal of crystal violet (CV) and methylene blue (MB) from binary aqueous solutions was investigated in a batch method under the influence of different parameters. In addition, predictive models for the adsorption process were developed using machine learning techniques such as artificial neural networks and random forests. ANN and RF models achieved high R² values of 0.9501 and 0.9797 for CV, 0.9471, and 09,834 for MB, respectively, and obtained low MSE values of 0.07107 and 0.03405 for CV, 0.09933, and 0.02908 for MB. The proposed adsorbent is cheap and eco-friendly and, on the other hand, is easily collected by the magnetic field. The adsorbent was characterized by applying FESEM-EDX, FESEM, BET, and FTIR. Various isotherm and kinetics models for the simultaneous adsorption of CV and MB were investigated in aqueous solutions. The adsorption isotherm and kinetics studies explain that the extended Langmuir model and pseudo-second-order models are best suited for CV and MB in the binary solution. The exothermic adsorption was achieved in the temperature range of 5-45 °C.

Evaluating the utility of a high throughput thiol-containing fluorescent probe to screen for reactivity: A case study with the Tox21 library

High-throughput screening (HTS) assays for bioactivity in the Tox21 program aim to evaluate an array of different biological targets and pathways, but a significant barrier to interpretation of these data is the lack of high-throughput screening (HTS) assays intended to identify non-specific reactive chemicals. This is an important aspect for prioritising chemicals to test in specific assays, identifying promiscuous chemicals based on their reactivity, as well as addressing hazards such as skin sensitisation which are not necessarily initiated by a receptor-mediated effect but act through a non-specific mechanism. Herein, a fluorescence-based HTS assay that allows the identification of thiol-reactive compounds was used to screen 7,872 unique chemicals in the Tox21 10K chemical library. Active chemicals were compared with profiling outcomes using structural alerts encoding electrophilic information. Random Forest classification models based on chemical fingerprints were developed to predict assay outcomes and evaluated through 10-fold stratified cross validation (CV). The mean CV Balanced Accuracy of the validation set was 0.648. The model developed shows promise as a tool to screen untested chemicals for their potential electrophilic reactivity based solely on chemical structural features.

Quantum-Mechanical Approach to Predicting the Carcinogenic Potency of N-Nitroso Impurities in Pharmaceuticals

N-Nitroso contaminants in medicinal products are of concern due to their high carcinogenic potency; however, not all these compounds are created equal, and some are relatively benign chemicals. Understanding the structure-activity relationships (SARs) that drive hazards in one molecule versus another is key to both protecting human health and alleviating costly and sometimes inaccurate animal testing. Here, we report on an extension of the CADRE (computer-aided discovery and REdesign) platform, which is used broadly by the pharmaceutical and personal care industries to assess environmental and human health endpoints, to predict the carcinogenic potency of N-nitroso compounds. The model distinguishes compounds in three potency categories with 77% accuracy in external testing, which surpasses the reproducibility of rodent cancer bioassays and constraints imposed by limited (high-quality) data. The robustness of predictions for more complex pharmaceuticals is maximized by capturing key SARs using quantum mechanics, that is, by hinging the model on the underlying chemistry versus chemicals in the training set. To this end, the present approach can be leveraged in a quantitative hazard assessment and to offer qualitative guidance using electronic structure comparisons between well-studied analogues and unknown contaminants.

How the Structure of Per- and Polyfluoroalkyl Substances (PFAS) Influences Their Binding Potency to the Peroxisome Proliferator-Activated and Thyroid Hormone Receptors-An In Silico Screening Study

In this study, we investigated PFAS (per- and polyfluoroalkyl substances) binding potencies to nuclear hormone receptors (NHRs): peroxisome proliferator-activated receptors (PPARs) α, β, and γ and thyroid hormone receptors (TRs) α and β. We have simulated the docking scores of 43 perfluoroalkyl compounds and based on these data developed QSAR (Quantitative Structure-Activity Relationship) models for predicting the binding probability to five receptors. In the next step, we implemented the developed QSAR models for the screening approach of a large group of compounds (4464) from the NORMAN Database. The in silico analyses indicated that the probability of PFAS binding to the receptors depends on the chain length, the number of fluorine atoms, and the number of branches in the molecule. According to the findings, the considered PFAS group bind to the PPARα, β, and γ only with low or moderate probability, while in the case of TR α and β it is similar except that those chemicals with longer chains show a moderately high probability of binding.

Quantitative structure-activity relationship and machine learning studies of 2-thiazolylhydrazone derivatives with anti-Cryptococcus neoformans activity

Cryptococcus neoformans is a fungus responsible for infections in humans with a significant number of cases in immunosuppressed patients, mainly in underdeveloped countries. In this context, the thiazolylhydrazones are a promising class of compounds with activity against C. neoformans. The understanding of the structure-activity relationship of these derivatives could lead to the design of robust compounds that could be promising drug candidates for fungal infections. Specifically, modern techniques such as 4D-QSAR and machine learning methods were employed in this work to generate two QSAR models (one 2D and one 4D) with high predictive power (r² for the test set equals to 0.934 and 0.831, respectively), and one random forest classification model was reported with Matthews correlation coefficient equals to 1 and 0.62 for internal and external validations, respectively. The physicochemical interpretation of selected models, indicated the importance of aliphatic substituents at the hydrazone moiety to antifungal activity, corroborating experimental data.Communicated by Ramaswamy H. Sarma.

Exposing the Limitations of Molecular Machine Learning with Activity Cliffs

Machine learning has become a crucial tool in drug discovery and chemistry at large, e.g., to predict molecular properties, such as bioactivity, with high accuracy. However, activity cliffs─pairs of molecules that are highly similar in their structure but exhibit large differences in potency─have received limited attention for their effect on model performance. Not only are these edge cases informative for molecule discovery and optimization but also models that are well equipped to accurately predict the potency of activity cliffs have increased potential for prospective applications. Our work aims to fill the current knowledge gap on best-practice machine learning methods in the presence of activity cliffs. We benchmarked a total of 24 machine and deep learning approaches on curated bioactivity data from 30 macromolecular targets for their performance on activity cliff compounds. While all methods struggled in the presence of activity cliffs, machine learning approaches based on molecular descriptors outperformed more complex deep learning methods. Our findings highlight large case-by-case differences in performance, advocating for (a) the inclusion of dedicated "activity-cliff-centered" metrics during model development and evaluation and (b) the development of novel algorithms to better predict the properties of activity cliffs. To this end, the methods, metrics, and results of this study have been encapsulated into an open-access benchmarking platform named MoleculeACE (Activity Cliff Estimation, available on GitHub at: https://github.com/molML/MoleculeACE). MoleculeACE is designed to steer the community toward addressing the pressing but overlooked limitation of molecular machine learning models posed by activity cliffs.

Combined 3D-QSAR, molecular docking and dynamics simulations studies to model and design TTK inhibitors

Tyrosine threonine kinase (TTK) is the key component of the spindle assembly checkpoint (SAC) that ensures correct attachment of chromosomes to the mitotic spindle and thereby their precise segregation into daughter cells by phosphorylating specific substrate proteins. The overexpression of TTK has been associated with various human malignancies, including breast, colorectal and thyroid carcinomas. TTK has been validated as a target for drug development, and several TTK inhibitors have been discovered. In this study, ligand and structure-based alignment as well as various partial charge models were used to perform 3D-QSAR modelling on 1H-Pyrrolo[3,2-c] pyridine core containing reported inhibitors of TTK protein using the comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA) approaches to design better active compounds. Different statistical methods i.e., correlation coefficient of non-cross validation (r2), correlation coefficient of leave-one-out cross-validation (q2), Fisher's test (F) and bootstrapping were used to validate the developed models. Out of several charge models and alignment-based approaches, Merck Molecular Force Field (MMFF94) charges using structure-based alignment yielded highly predictive CoMFA (q2 = 0.583, Predr2 = 0.751) and CoMSIA (q2 = 0.690, Predr2 = 0.767) models. The models exhibited that electrostatic, steric, HBA, HBD, and hydrophobic fields play a key role in structure activity relationship of these compounds. Using the contour maps information of the best predictive model, new compounds were designed and docked at the TTK active site to predict their plausible binding modes. The structural stability of the TTK complexes with new compounds was confirmed using MD simulations. The simulation studies revealed that all compounds formed stable complexes. Similarly, MM/PBSA method based free energy calculations showed that these compounds bind with reasonably good affinity to the TTK protein. Overall molecular modelling results suggest that newly designed compounds can act as lead compounds for the optimization of TTK inhibitors.

A regression-based QSAR-model to predict acute toxicity of aromatic chemicals in tadpoles of the Japanese brown frog (Rana japonica): Calibration, validation, and future developments to support risk assessment of chemicals in amphibians

Amphibian populations are undergoing a global decline worldwide. Such decline has been attributed to their unique physiology, ecology, and exposure to multiple stressors including chemicals, temperature, and biological hazards such as fungi of the Batrachochytrium genus, viruses such as Ranavirus, and habitat reduction. There are limited toxicity data for chemicals available for amphibians and few quantitative structure-activity relationship (QSAR) models have been developed and are publicly available. Such QSARs provide important tools to assess the toxicity of chemicals particularly in a data poor context. QSARs provide important tools to assess the toxicity of chemicals particularly when no toxicological data are available. This manuscript provides a description and validation of a regression-based QSAR model to predict, in a quantitative manner, acute lethal toxicity of aromatic chemicals in tadpoles of the Japanese brown frog (Rana japonica). QSAR models for acute median lethal molar concentrations (LC50-12 h) of waterborne chemicals using the Monte Carlo method were developed. The statistical characteristics of the QSARs were described as average values obtained from five random distributions into training and validation sets. Predictions from the model gave satisfactory results for the overall training set (R² = 0.72 and RMSE = 0.33) and were even more robust for the validation set (R² = 0.96 and RMSE = 0.11). Further development of QSAR models in amphibians, particularly for other life stages and species, are discussed.

QSAR analysis on a large and diverse set of potent phosphoinositide 3-kinase gamma (PI3Kγ) inhibitors using MLR and ANN methods

Phosphorylation of PI3Kγ as a member of lipid kinases-enzymes, plays a crucial role in regulating immune cells through the generation of intracellular signals. Deregulation of this pathway is involved in several tumors. In this research, diverse sets of potent and selective isoform-specific PI3Kγ inhibitors whose drug-likeness was confirmed based on Lipinski’s rule of five were used in the modeling process. Genetic algorithm (GA)-based multivariate analysis was employed on the half-maximal inhibitory concentration (IC₅₀) of them. In this way, multiple linear regression (MLR) and artificial neural network (ANN) algorithm, were used to QSAR models construction on 245 compounds with a wide range of pIC₅₀ (5.23–9.32). The stability and robustness of the models have been evaluated by external and internal validation methods (R² 0.623–0.642, RMSE 0.464–0.473, F 40.114, Q²_LOO 0.600, and R²_y-random 0.011). External verification using a wide variety of structures out of the training and test sets show that ANN is superior to MLR. The descriptors entered into the model are in good agreement with the X-ray structures of target-ligand complexes; so the model is interpretable. Finally, Williams plot-based analysis was applied to simultaneously compare the inhibitory activity and structural similarity of training, test and validation sets.

Design of Dual COX-2 and 5-LOX Inhibitors with Iron-Chelating Properties Using Structure-Based and Ligand-Based Methods

Background:

Inflammation is a critical component of many disease progressions, such as malignancy, cardiovascular and rheumatic diseases. The inhibition of inflammatory mediators synthesis by modulation of cyclooxygenase (COX) and lipoxygenase (LOX) pathways provides challenging strategy for development of more effective drugs.

Objective:

The aim of this study was to design dual COX-2 and 5-LOX inhibitors with iron-chelating properties using a combination of ligand-based (three-dimensional quantitative structure-activity relationship (3D-QSAR)) and structure-based (molecular docking) methods.

Methods:

The 3D-QSAR analysis was applied on a literature dataset consisting of 28 dual COX-2 and 5- LOX inhibitors in Pentacle software. The quality of developed COX-2 and 5-LOX 3D-QSAR models were evaluated by internal and external validation methods. The molecular docking analysis was performed in GOLD software, while selected ADMET properties were predicted in ADMET predictor software.

Results:

According to the molecular docking studies, the class of sulfohydroxamic acid analogues, previously designed by 3D-QSAR, were clustered as potential dual COX-2 and 5-LOX inhibitors with ironchelating properties. Based on the 3D-QSAR and molecular docking, 1j, 1g and 1l were selected as the most promising dual COX-2 and 5-LOX inhibitors. According to the in silico ADMET predictions, all compounds had ADMET_Risk score less than 7 and CYP_Risk score lower than 2.5. Designed compounds were not estimated as hERG inhibitors and 1j had improved intrinsic solubility (8.704) in comparison to the dataset compounds (0.411-7.946).

Conclusion:

By combining 3D-QSAR and molecular docking, three compounds (1j, 1g and 1l) were selected as the most promising designed dual COX-2 and 5-LOX inhibitors, for which good activity, as well as favourable ADMET properties and toxicity, are expected.

Application of random forest for modeling batch and continuous fixed-bed removal of crystal violet from aqueous solutions using Gypsophila aretioides stem-based biosorbent

In this work, the Gypsophila aretioides (GYP-A) stem is used as a biosorbent to remove crystal violet (CV) by the static and dynamic systems from aqueous solutions; the biosorbent is interesting in green chemistry and, on the other hand, cheaper than activated carbon and does not have the limitation of industrialization. The effects of different operating parameters such as pH(3-9), biosorbent dosage(0.4-1.8 mg/L), and initial concentration of CV(100-250 mg/L) and time for the batch method and the bed height, inlet CV concentration(75-250 mg/L), and flow rate(3-8) on the breakthrough curves for the continuous method is investigated. The result of CV adsorption onto GYP-A using the batch method indicates that the model fits Freundlich > Temkin > Langmuir > R-D, and R² equal 0.9953, 0.9847, 0.9161, 0.7909 were obtained for isotherm model, respectively. A pseudo-second-order model (R² = 0.9995-0.9997) is recommended to describe the adsorption kinetics. The Thomas and Yoon-Nelson models were analyzed to study the adsorption kinetics. The random forest model shows an excellent ability to predict the parameters involved in the CV adsorption process with appropriate accuracy and useable for large data, robust against noise; it can be very effective in selecting important variables.

The Feasibility of Two Handheld Spectrometers for Meat Speciation Combined with Chemometric Methods and Its Application for Halal Certification

Handheld visible-near-infrared (Vis-NIR) and near-infrared (NIR) spectroscopy can be cost-effective, rapid, non-destructive and transportable techniques for identifying meat species and may be valuable for enforcement authorities, retail and consumers. In this study, a handheld Vis-NIR (400-1000 nm) and a handheld NIR (900-1700 nm) spectrometer were applied to discriminate halal meat species from pork (halal certification), as well as speciation of intact and ground lamb, beef, chicken and pork (160 meat samples). Several types of class modeling multivariate approaches were applied. The presented one-class classification (OCC) approach, especially with the Vis-NIR sensor (95-100% correct classification rate), was found to be suitable for the application of halal from non-halal meat-species discrimination. In a discriminant approach, using the Vis-NIR data and support vector machine (SVM) classification, the four meat species tested could be classified with accuracies of 93.4% and 94.7% for ground and intact meat, respectively, while with partial least-squares discriminant analysis (PLS-DA), classification accuracies were 87.4% (ground) and 88.6% (intact). Using the NIR sensor, total accuracies of the SVM models were 88.2% and 81.5% for ground and intact meats, respectively, and PLS-DA classification accuracies were 88.3% (ground) and 80% (intact). We conclude that the Vis-NIR sensor was most successful in the halal certification (OCC approaches) and speciation (discriminant approaches) for both intact and ground meat using SVM.

Validation strategies for target prediction methods

Computational methods for target prediction, based on molecular similarity and network-based approaches, machine learning, docking and others, have evolved as valuable and powerful tools to aid the challenging task of mode of action identification for bioactive small molecules such as drugs and drug-like compounds. Critical to discerning the scope and limitations of a target prediction method is understanding how its performance was evaluated and reported. Ideally, large-scale prospective experiments are conducted to validate the performance of a model; however, this expensive and time-consuming endeavor is often not feasible. Therefore, to estimate the predictive power of a method, statistical validation based on retrospective knowledge is commonly used. There are multiple statistical validation techniques that vary in rigor. In this review we discuss the validation strategies employed, highlighting the usefulness and constraints of the validation schemes and metrics that are employed to measure and describe performance. We address the limitations of measuring only generalized performance, given that the underlying bioactivity and structural data are biased towards certain small-molecule scaffolds and target families, and suggest additional aspects of performance to consider in order to produce more detailed and realistic estimates of predictive power. Finally, we describe the validation strategies that were employed by some of the most thoroughly validated and accessible target prediction methods.

Sample-size dependence of validation parameters in linear regression models and in QSAR

The dependence of statistical validation parameters was investigated on the size of the sample taken in fit of multivariate linear curves. We observed that R² and related internal parameters were misleading as they overestimated the goodness-of-fit of models at small sample size. Cross-validation metrics showed correct trends. It was possible to scale the leave-one-out and the leave-many-out results close to identical by correcting the degrees of freedom of the models. y and x-randomized validation parameters were calculated and the methods provided close to identical results. We suggest to use the simplest methods in both cases. The external parameters followed correct trends with respect to the sample size, but their sensitivity differed. We plotted the Roy-Ojha metrics in 2D and we coloured them with respect to other external parameters to provide an easy classification of models. The rank correlations were calculated between the performance parameters. Up to a sample size, goodness-of-fit and robustness were distinguishable, but above a certain sample size, the parameters were redundant. The external-internal pairs were weakly correlated. Our data show that all the three aspects of validation are necessary at small sample sizes, but the internal check of robustness is not informative above a given sample size.

QNAR modeling of cytotoxicity of mixing nano-TiO2 and heavy metals

The Quantitative Structure-Activity Relationship (QSAR) has been used to investigate organic mixtures but QSAR in the nanomaterial field (QNAR) is still new. Toxicity is a result of the interaction of many substances. QNAR research focuses on a single nanomaterial in the long-term. It is difficult to find an appropriate descriptor to build a model due to the complexity of the mixture. Here, we attempt to build a QNAR model to predict cell viability for HK-2 cells exposed to a mixture containing nano-TiO₂ and heavy metals. HK-2 cells were exposed to four groups of mixtures containing heavy-metals and nanomaterials and CCK8 was added to obtain the number of living cells. At the same time, ROS was investigated to study this mechanism. Each descriptor of the components and mixtures were obtained using the formula D_mix= [Formula: see text] respectively. We used the Multiple Partial Least Squares Regression (PLS) and Random Forest Regression (RF) to build a QNAR model. Both models reliably predict and assess viability of HK-2 cells exposed to the mixture. The RF model showed greater stability and higher precision in toxicity predictability and can be applied to environmental nano-toxicology.

Application 2D Descriptors and Artificial Neural Networks for Beta-Glucosidase Inhibitors Screening

Beta-glucosidase inhibitors play important medical and biological roles. In this study, simple two-variable artificial neural network (ANN) classification models were developed for beta-glucosidase inhibitors screening. All bioassay data were obtained from the ChEMBL database. The classifiers were generated using 2D molecular descriptors and the data miner tool available in the STATISTICA package (STATISTICA Automated Neural Networks, SANN). In order to evaluate the models' accuracy and select the best classifiers among automatically generated SANNs, the Matthews correlation coefficient (MCC) was used. The application of the combination of maxHBint3 and SpMax8_Bhs descriptors leads to the highest predicting abilities of SANNs, as evidenced by the averaged test set prediction results (MCC = 0.748) calculated for ten different dataset splits. Additionally, the models were analyzed employing receiver operating characteristics (ROC) and cumulative gain charts. The thirteen final classifiers obtained as a result of the model development procedure were applied for a natural compounds collection available in the BIOFACQUIM database. As a result of this beta-glucosidase inhibitors screening, eight compounds were univocally classified as active by all SANNs.

Molecular Docking and QSAR Studies as Computational Tools Exploring the Rescue Ability of F508del CFTR Correctors

Cystic fibrosis (CF) is the autosomal recessive disorder most recurrent in Caucasian populations. Different mutations involving the cystic fibrosis transmembrane regulator protein (CFTR) gene, which encodes the CFTR channel, are involved in CF. A number of life-prolonging therapies have been conceived and deeply investigated to combat this disease. Among them, the administration of the so-called CFTR modulators, such as correctors and potentiators, have led to quite beneficial effects. Recently, based on QSAR (quantitative structure activity relationship) studies, we reported the rational design and synthesis of compound 2, an aminoarylthiazole-VX-809 hybrid derivative exhibiting promising F508del-CFTR corrector ability. Herein, we explored the docking mode of the prototype VX-809 as well as of the aforementioned correctors in order to derive useful guidelines for the rational design of further analogues. In addition, we refined our previous QSAR analysis taking into account our first series of in-house hybrids. This allowed us to optimize the QSAR model based on the chemical structure and the potency profile of hybrids as F508del-CFTR correctors, identifying novel molecular descriptors explaining the SAR of the dataset. This study is expected to speed up the discovery process of novel potent CFTR modulators.

Computational strategies towards developing novel antimelanogenic agents

AIMS

Extrinsic ageing or photoageing relates to the onset of age-linked phenotypes such as skin hyperpigmentation due to UV exposure. UV induced upregulated production of tyrosinase enzyme, which catalyses the vital biochemical reactions of melanin synthesis is responsible for the inception of skin hyperpigmentation. We aimed to generate a validated QSAR model with a dataset consisting of 69 thio-semicarbazone derivatives to elucidate the physicochemical properties of compounds essential for tyrosinase inhibition and to identify novel lead molecules with enhanced tyrosinase inhibitory activity and bioavailability.

MAIN METHODS

Lead optimization and insilico approaches were employed in this research work. QSAR model was generated and validated by exploiting Multiple Linear Regression method. Prioritization of lead-like compounds was accomplished by performing multi parameter optimization depleting molecular docking, bioavailability assessments and toxicity prediction for 69 compounds Derivatives of best lead compound were retrieved from chemical spaces.

KEY FINDINGS

Molecular descriptors explicated the significance of chemical properties essential for chelation of copper ions present in the active site of tyrosinase protein target. Further, derivatives which comprise of electron donating groups in their chemical structure were predicted and analysed for tyrosinase inhibitory activity by employing insilico methodologies including chemical space exploration.

SIGNIFICANCE

Our research work resulted in the generation of a validated QSAR model with higher degree of external predictive ability and significance to tyrosinase inhibitory activity. We propose 11 novel derivative compounds with enhanced tyrosinase inhibitory activity and bioavailability.

Construction of Quantitative Structure Activity Relationship (QSAR) Models to Predict Potency of Structurally Diversed Janus Kinase 2 Inhibitors

Janus kinase 2 (JAK2) inhibitors represent a promising therapeutic class of anticancer agents against many myeloproliferative disorders. Bioactivity data on pIC50 of 2229 JAK2 inhibitors were employed in the construction of quantitative structure-activity relationship (QSAR) models. The models were built from 100 data splits using decision tree (DT), support vector machine (SVM), deep neural network (DNN) and random forest (RF). The predictive power of RF models were assessed via 10-fold cross validation, which afforded excellent predictive performance with R2 and RMSE of 0.74 ± 0.05 and 0.63 ± 0.05, respectively. Moreover, test set has excellent performance of R2 (0.75 ± 0.03) and RMSE (0.62 ± 0.04). In addition, Y-scrambling was utilized to evaluate the possibility of chance correlation of the predictive model. A thorough analysis of the substructure fingerprint count was conducted to provide insights on the inhibitory properties of JAK2 inhibitors. Molecular cluster analysis revealed that pyrazine scaffolds have nanomolar potency against JAK2.

Whether the Validation of the Predictive Potential of Toxicity Models is a Solved Task?

Different kinds of biological activities are defined by complex biochemical interactions, which are termed as a "mathematical function" not only of the molecular structure but also for some additional circumstances, such as physicochemical conditions, interactions via energy and information effects between a substance and organisms, organs, cells. These circumstances lead to the great complexity of prediction for biochemical endpoints, since all "details" of corresponding phenomena are practically unavailable for the accurate registration and analysis. Researchers have not a possibility to carry out and analyse all possible ways of the biochemical interactions, which define toxicological or therapeutically attractive effects via direct experiment. Consequently, a compromise, i.e. the development of predictive models of the above phenomena, becomes necessary. However, the estimation of the predictive potential of these models remains a task that is solved only partially. This mini-review presents a collection of attempts to be used for the above-mentioned task, two special statistical indices are proposed, which may be a measure of the predictive potential of models. These indices are (i) Index of Ideality of Correlation; and (ii) Correlation Contradiction Index.

Artificial neural network modelling for organic and total nitrogen removal of aerobic granulation under steady-state condition

Aerobic granulation is a recent technology with high level of complexity and sensitivity to environmental and operational conditions. Artificial neural networks (ANNs), computational tools capable of describing complex non-linear systems, are the best fit to simulate aerobic granular bioreactors. In this study, two feedforward backpropagation ANN models were developed to predict chemical oxygen demand (Model I) and total nitrogen removal efficiencies (Model II) of aerobic granulation technology under steady-state condition. Fundamentals of ANN models and the steps to create them were briefly reviewed. The models were respectively fed with 205 and 136 data points collected from laboratory-, pilot-, and full-scale studies on aerobic granulation technology reported in the literature. Initially, 60%, 20%, and 20%, and 80%, 10%, and 10% of the points in the corresponding datasets were randomly chosen and used for training, testing, and validation of Model I, and Model II, respectively. Overall coefficient of determination (R²) value and mean squared error (MSE) of the two models were initially 0.49 and 15.5, and 0.37 and 408, respectively. To improve the model performance, two data division methods were used. While one method is generic and potentially applicable to other fields, the other can only be applied to modelling the performance of aerobic granular reactors. R² value and MSE were improved to 0.90 and 2.54, and 0.81 and 121.56, respectively, after applying the new data division methods. The results demonstrated that ANN-based models were capable simulation approach to predict a complicated process like aerobic granulation.

The Acid Strength of the Lewis-Brønsted Superacids - A QSPR Study

The acidity of Lewis-Brønsted superacids can be derived from the theoretical calculations as the Gibbs free energy of the deprotonation reaction (ΔG_acid ), which describes the tendency of a studied compound to donate a proton. This paper presents the first Quantitative Structure - Property Relationship (QSPR) model that correlates the ΔG_acid of superacid (HF/MeX₃ formula (X=F, Cl, Br)) with their structure. Developed model is well fitted, roubustness, has good predictive abilities, fulfills all OECD recommendation for good model. Obtained results provide the insight into the relation of structural features of superacids, which are responsible for their acid strength - the structures characterized by strong F-Me dative bond (with relatively large vibrational frequency), small positive partial atomic charge on Me central atom, possibly large polarity exhibit large acid strength. Such assumption can be used in the future as valuable information in the process of the designing new, stronger, more effective superacids.

DFT-Based QSAR Modelling of Inhibitory Activity of Coumarins and Sulfocoumarins on Carbonic Anhydrase (CA) Isoforms (CA I and CA II)

OBJECTIVE

We present three robust, validated and statistically significant quantitative structure-activity relationship (QSAR) models, which deal with the calculated molecular descriptors and experimental inhibition constant (Ki) of 42 coumarin and sulfocoumarin derivatives measured against CA I and II isoforms.

METHODS

The compounds were subjected to DFT calculations in order to obtain quantum chemical molecular descriptors. Multiple linear regression algorithms were applied to construct QSAR models. Separation of the compounds into training and test sets was accomplished using Kennard-Stone algorithm. Leverage approach was applied to determine Applicability Domain (AD) of the obtained models.

RESULTS

Three models were developed. The first model, CAI_model1 comprises 30/11 training/test compounds with the statistical parameters of R2=0.85, Q2=0.77, F=27.57, R2 (test) =0.72. The second one, CAII_model2 comprises 30/12 training/test compounds with the statistical parameters of R2=0.86, Q2=0.78, F=30.27, R2 (test) =0.85. The final model, ΔpKi_model3 consists of 25/3 training/ test compounds with the statistical parameters of R2=0.78, Q2=0.62, F=13.80 and R2(test) =0.99.

CONCLUSION

Interpretation of reactivity-related descriptors such as HOMO-1 and LUMO energies and visual inspection of their maps of orbital electron density leads to a conclusion that the binding free energy of the entire binding process may be modulated by the kinetics of the hydrolyzing step of coumarins.

On Splitting Training and Validation Set: A Comparative Study of Cross-Validation, Bootstrap and Systematic Sampling for Estimating the Generalization Performance of Supervised Learning

Model validation is the most important part of building a supervised model. For building a model with good generalization performance one must have a sensible data splitting strategy, and this is crucial for model validation. In this study, we conducted a comparative study on various reported data splitting methods. The MixSim model was employed to generate nine simulated datasets with different probabilities of mis-classification and variable sample sizes. Then partial least squares for discriminant analysis and support vector machines for classification were applied to these datasets. Data splitting methods tested included variants of cross-validation, bootstrapping, bootstrapped Latin partition, Kennard-Stone algorithm (K-S) and sample set partitioning based on joint X–Y distances algorithm (SPXY). These methods were employed to split the data into training and validation sets. The estimated generalization performances from the validation sets were then compared with the ones obtained from the blind test sets which were generated from the same distribution but were unseen by the training/validation procedure used in model construction. The results showed that the size of the data is the deciding factor for the qualities of the generalization performance estimated from the validation set. We found that there was a significant gap between the performance estimated from the validation set and the one from the test set for the all the data splitting methods employed on small datasets. Such disparity decreased when more samples were available for training/validation, and this is because the models were then moving towards approximations of the central limit theory for the simulated datasets used. We also found that having too many or too few samples in the training set had a negative effect on the estimated model performance, suggesting that it is necessary to have a good balance between the sizes of training set and validation set to have a reliable estimation of model performance. We also found that systematic sampling method such as K-S and SPXY generally had very poor estimation of the model performance, most likely due to the fact that they are designed to take the most representative samples first and thus left a rather poorly representative sample set for model performance estimation.

Boosted feature selectors: a case study on prediction P-gp inhibitors and substrates

Feature selection is commonly used as a preprocessing step to machine learning for improving learning performance, lowering computational complexity and facilitating model interpretation. This paper proposes the application of boosting feature selection to improve the classification performance of standard feature selection algorithms evaluated for the prediction of P-gp inhibitors and substrates. Two well-known classification algorithms, decision trees and support vector machines, were used to classify the chemical compounds. The experimental results showed better performance for boosting feature selection with respect to the standard feature selection algorithms while maintaining the capability for feature reduction.

Probing the origin of estrogen receptor alpha inhibition via large-scale QSAR study

Estrogen is an important component for the sustenance of normal physiological functions of the mammary glands, particularly for growth and differentiation. Approximately, two-thirds of breast cancers are positive for estrogen receptor (ERs), which is a predisposing factor for the growth of breast cancer cells. As such, ERα represents a lucrative therapeutic target for breast cancer that has attracted wide interest in the search for inhibitory agents. However, the conventional laboratory processes are cost- and time-consuming. Thus, it is highly desirable to develop alternative methods such as quantitative structure-activity relationship (QSAR) models for predicting ER-mediated endocrine agitation as to simplify their prioritization for future screening. In this study, we compiled and curated a large, non-redundant data set of 1231 compounds with ERα inhibitory activity (pIC50). Using comprehensive validation tests, it was clearly observed that the model utilizing the substructure count as descriptors, performed well considering two objectives: using less descriptors for model development and achieving high predictive performance (R Tr 2 = 0.94, Q CV 2 = 0.73, and Q Ext 2 = 0.73). It is anticipated that our proposed QSAR model may become a useful high-throughput tool for identifying novel inhibitors against ERα.

Multiple target-based pharmacophore design from active site structures

Health care systems have benefitted from rational drug discovery processes like vHTS, virtual high throughput screening pharmacophores and quantitative structure-activity relationships, and many challenges have been explored using such techniques: decisions on specificity and selectivity are made after screening millions of molecules for multiple targets. Recent challenges in drug research emphasize the design of drugs that bind with more than one target of interest (multi-target) and do not bind with undesirable targets. This work attempts to use a three-dimensional interaction profile of the active site of a class of proteins, identify selective positions for the binding of functional groups, called features, and develop ensembles of multi-targeted pharmacophores that retain specificity and selectivity. The goal of this study is to develop multi-target pharmacophores by computational methods using protein structures alone to guide the discovery of novel inhibitors of plasmepsins, displaying selectivity over their human homologs, cathepsin D and pepsin. The development of such novel tools is attempted using a combination of different approaches such as the molecular interaction field, clique graph and inductive logic programming to identify and compare specific and selective complementary features. The identification of selective combinations of features has resulted in the design of multi-featured specific and selective pharmacophores that are validated using antimalarial compounds in ChEMBL that are known for their anti-plasmepsin II activity. This novel method is computationally less intensive and is applicable to any known class of target structures for finding specific and selective binders simultaneously.

PVLOO-Based Training Set Selection Improves the External Predictability of QSAR/QSPR Models

In QSAR/QSPR modeling, the indispensable way to validate the predictability of a model is to perform its statistical external validation. It is common that a division algorithm should be used to select training sets from chemical compound libraries or collections prior to external validations. In this study, a division method based on the posterior variance of leave-one-out cross-validation (PV_LOO) of the Gaussian process (GP) has been developed with the goal of producing more predictive models. Four structurally diverse data sets of good quality are collected from the literature and then redeveloped and validated on the basis of training set selection methods, namely, four kinds of PV_LOO-based training set selection methods with three types of covariance functions (squared exponential, rational quadratic, and neural network covariance functions), the Kennard-Stone algorithm, and random division. The root mean squared error (RMSE) of external validation reported for each model serves as a basis for the final comparison. The results of this study indicate that the training sets with higher values of PV_LOO have statistically better external predictability than the training sets generated from other division methods discussed here. These findings could be explained by proposing that the PV_LOO value of GP could indicate the mechanism diversity of a specific compound in QSAR/QSPR data sets.

How the structure of ionic liquid affects its toxicity to Vibrio fischeri?

In the present work, we have proposed a statistical model predicting the toxicity of ionic liquids (ILs) to Vibrio fischeri bacteria using the Quantitative Structure-Activity Relationships (QSAR) method. The model was developed with Multiple Linear Regression (MLR) technique, using the Gutman molecular topological index (GMTI), the lopping centric information index (LOC) and the number of oxygen atoms. Presented model is characterized by the good fit to the experimental data (R(2) = 0.78), high robustness (Q(2)CV = 0.72) and good predictive ability (Q(2)EXT = 0.75). This approach, with using very simple descriptors, helps to initially evaluate the toxicity of newly designed ionic liquids. The studied toxicity of ionic liquids depends mainly on their cations' structure: larger, more branched cations with long alkyl chains are more toxic than the smaller, linear ones. The presence of polar functional groups in the cation's structure reduces the toxic properties of ionic liquids. The structure of the anion has little effect on the toxicity of the studied ionic liquids. Obtained results will provide insight into the toxicity mechanisms and useful information for assessing the potential ecological risk of ionic liquids.

Multilayer perceptron neural network-based approach for modeling phycocyanin pigment concentrations: case study from lower Charles River buoy, USA

This paper proposes multilayer perceptron neural network (MLPNN) to predict phycocyanin (PC) pigment using water quality variables as predictor. In the proposed model, four water quality variables that are water temperature, dissolved oxygen, pH, and specific conductance were selected as the inputs for the MLPNN model, and the PC as the output. To demonstrate the capability and the usefulness of the MLPNN model, a total of 15,849 data measured at 15-min (15 min) intervals of time are used for the development of the model. The data are collected at the lower Charles River buoy, and available from the US Environmental Protection Agency (USEPA). For comparison purposes, a multiple linear regression (MLR) model that was frequently used for predicting water quality variables in previous studies is also built. The performances of the models are evaluated using a set of widely used statistical indices. The performance of the MLPNN and MLR models is compared with the measured data. The obtained results show that (i) the all proposed MLPNN models are more accurate than the MLR models and (ii) the results obtained are very promising and encouraging for the development of phycocyanin-predictive models.

Probing the origins of human acetylcholinesterase inhibition via QSAR modeling and molecular docking

Alzheimer's disease (AD) is a chronic neurodegenerative disease which leads to the gradual loss of neuronal cells. Several hypotheses for AD exists (e.g., cholinergic, amyloid, tau hypotheses, etc.). As per the cholinergic hypothesis, the deficiency of choline is responsible for AD; therefore, the inhibition of AChE is a lucrative therapeutic strategy for the treatment of AD. Acetylcholinesterase (AChE) is an enzyme that catalyzes the breakdown of the neurotransmitter acetylcholine that is essential for cognition and memory. A large non-redundant data set of 2,570 compounds with reported IC50 values against AChE was obtained from ChEMBL and employed in quantitative structure-activity relationship (QSAR) study so as to gain insights on their origin of bioactivity. AChE inhibitors were described by a set of 12 fingerprint descriptors and predictive models were constructed from 100 different data splits using random forest. Generated models afforded R (2), [Formula: see text] and [Formula: see text] values in ranges of 0.66-0.93, 0.55-0.79 and 0.56-0.81 for the training set, 10-fold cross-validated set and external set, respectively. The best model built using the substructure count was selected according to the OECD guidelines and it afforded R (2), [Formula: see text] and [Formula: see text] values of 0.92 ± 0.01, 0.78 ± 0.06 and 0.78 ± 0.05, respectively. Furthermore, Y-scrambling was applied to evaluate the possibility of chance correlation of the predictive model. Subsequently, a thorough analysis of the substructure fingerprint count was conducted to provide informative insights on the inhibitory activity of AChE inhibitors. Moreover, Kennard-Stone sampling of the actives were applied to select 30 diverse compounds for further molecular docking studies in order to gain structural insights on the origin of AChE inhibition. Site-moiety mapping of compounds from the diversity set revealed three binding anchors encompassing both hydrogen bonding and van der Waals interaction. Molecular docking revealed that compounds 13, 5 and 28 exhibited the lowest binding energies of -12.2, -12.0 and -12.0 kcal/mol, respectively, against human AChE, which is modulated by hydrogen bonding, π-π stacking and hydrophobic interaction inside the binding pocket. These information may be used as guidelines for the design of novel and robust AChE inhibitors.

Linear and nonlinear models for predicting fish bioconcentration factors for pesticides

This work is devoted to the applications of the multiple linear regression (MLR), multilayer perceptron neural network (MLP NN) and projection pursuit regression (PPR) to quantitative structure-property relationship analysis of bioconcentration factors (BCFs) of pesticides tested on Bluegill (Lepomis macrochirus). Molecular descriptors of a total of 107 pesticides were calculated with the DRAGON Software and selected by inverse enhanced replacement method. Based on the selected DRAGON descriptors, a linear model was built by MLR, nonlinear models were developed using MLP NN and PPR. The robustness of the obtained models was assessed by cross-validation and external validation using test set. Outliers were also examined and deleted to improve predictive power. Comparative results revealed that PPR achieved the most accurate predictions. This study offers useful models and information for BCF prediction, risk assessment, and pesticide formulation.

Estimation of melting points of large set of persistent organic pollutants utilizing QSPR approach

The presence of polyhalogenated persistent organic pollutants (POPs), such as Cl/Br-substituted benzenes, biphenyls, diphenyl ethers, and naphthalenes has been identified in all environmental compartments. The exposure to these compounds can pose potential risk not only for ecological systems, but also for human health. Therefore, efficient tools for comprehensive environmental risk assessment for POPs are required. Among the factors vital for environmental transport and fate processes is melting point of a compound. In this study, we estimated the melting points of a large group (1419 compounds) of chloro- and bromo- derivatives of dibenzo-p-dioxins, dibenzofurans, biphenyls, naphthalenes, diphenylethers, and benzenes by utilizing quantitative structure-property relationship (QSPR) techniques. The compounds were classified by applying structure-based clustering methods followed by GA-PLS modeling. In addition, random forest method has been applied to develop more general models. Factors responsible for melting point behavior and predictive ability of each method were discussed.

Advantages and limitations of classic and 3D QSAR approaches in nano-QSAR studies based on biological activity of fullerene derivatives

In this contribution, the advantages and limitations of two computational techniques that can be used for the investigation of nanoparticles activity and toxicity: classic nano-QSAR (Quantitative Structure-Activity Relationships employed for nanomaterials) and 3D nano-QSAR (three-dimensional Quantitative Structure-Activity Relationships, such us Comparative Molecular Field Analysis, CoMFA/Comparative Molecular Similarity Indices Analysis, CoMSIA analysis employed for nanomaterials) have been briefly summarized. Both approaches were compared according to the selected criteria, including: efficiency, type of experimental data, class of nanomaterials, time required for calculations and computational cost, difficulties in the interpretation. Taking into account the advantages and limitations of each method, we provide the recommendations for nano-QSAR modellers and QSAR model users to be able to determine a proper and efficient methodology to investigate biological activity of nanoparticles in order to describe the underlying interactions in the most reliable and useful manner.