Search

The EUOS/SLAS challenge aimed to facilitate the development of reliable algorithms to predict the aqueous solubility of small molecules using experimental data from 100 K compounds. In total, hundred teams took part in the challenge to predict low, medium and highly soluble compounds as measured by the nephelometry assay. This article describes the winning model, which was developed using the publicly available Online CHEmical database and Modeling environment (OCHEM) available on the website https://ochem.eu/article/27. We describe in detail the assumptions and steps used to select methods, descriptors and strategy which contributed to the winning solution. In particular we show that consensus based on 28 models calculated using descriptor-based and representation learning methods allowed us to obtain the best score, which was higher than those based on individual approaches or consensus models developed using each individual approach. A combination of diverse models allowed us to decrease both bias and variance of individual models and to calculate the highest score. The model based on Transformer CNN contributed the best individual score thus highlighting the power of Natural Language Processing (NLP) methods. The inclusion of information about aleatoric uncertainty would be important to better understand and use the challenge data by the contestants.

Plasma protein binding (PPB) is closely related to pharmacokinetics, pharmacodynamics and drug toxicity. Existing models for predicting PPB often suffer from low prediction accuracy and poor interpretability, especially for high PPB compounds, and are most often not experimentally validated. Here, we carried out a strict data curation protocol, and applied consensus modeling to obtain a model with a coefficient of determination of 0.90 and 0.91 on the training set and the test set, respectively. This model (available on the OCHEM platform https://ochem.eu/article/29) was further retrospectively validated for a set of 63 poly-fluorinated molecules and prospectively validated for a set of 25 highly diverse compounds, and its performance for both these sets was superior to that of the other previously reported models. Furthermore, we identified the physicochemical and structural characteristics of high and low PPB molecules for further structural optimization. Finally, we provide practical and detailed recommendations for structural optimization to decrease PPB binding of lead compounds., (Copyright © 2024. Published by Elsevier B.V.)

Bioconcentration factors (BCFs) are markers of chemical substance accumulation in organisms, and they play a significant role in determining the environmental risk of various chemicals. Experiments to obtain BCFs are expensive and time-consuming; therefore, it is better to estimate BCF early in the chemical development process. The current research aims to evaluate the ecotoxicity potential of 122 pharmaceuticals and identify possible important structural attributes using BCF as the determining feature against a group of fish species. We have calculated the theoretical 2D descriptors from the OCHEM platform and SiRMS descriptor calculating software. The regression-based quantitative structure–property relationship (QSPR) modeling was used to identify the chemical features responsible for acute fish bioconcentration. Multiple models with the “intelligent consensus” algorithm were employed for the regression-based approach improving the predictive ability of the models. To ensure the robustness and interpretability of the developed models, rigorous validation was performed employing various statistical internal and external validation metrics. From the developed models, it can be specified that the presence of large lipophilic and electronegative moieties greatly enhances the bioaccumulative potential of pharmaceuticals, whereas the hydrophilic characteristics have shown a negative impact on BCF. Furthermore, the developed models were employed to screen the DrugBank database (https://go.drugbank.com/) for assessing the BCF properties of the entire database. The evidence acquired from the modeled descriptors might be used for aquatic risk assessment in the future, with the added benefit of providing an early caution of their probable negative impact on aquatic ecosystems for regulatory purposes.

A previously developed model to predict antibacterial activity of ionic liquids against a resistant A. baumannii strain was used to assess activity of phosphonium ionic liquids. Their antioxidant potential was additionally evaluated with newly developed models, which were based on public data. The accuracy of the models was rigorously evaluated using cross-validation as well as test set prediction. Six alkyl triphenylphosphonium and alkyl tributylphosphonium bromides with the C8, C10, and C12 alkyl chain length were synthesized and tested in vitro. Experimental studies confirmed their activity against A. baumannii as well as showed pronounced antioxidant properties. These results suggest that phosphonium ionic liquids could be promising lead structures against A. baumannii. [ABSTRACT FROM AUTHOR]

AlphaScreen is one of the most widely used assay technologies in drug discovery due to its versatility, dynamic range and sensitivity. However, a presence of false positives and frequent hitters contributes to difficulties with an interpretation of measured HTS data. Although filters do exist to identify frequent hitters for AlphaScreen, they are frequently based on privileged scaffolds. The development of such filters is time consuming and requires deep domain knowledge. Recently, machine learning and artificial intelligence methods are emerging as important tools to advance drug discovery and chemoinformatics, including their application to identification of frequent hitters in screening assays. However, the relative performance and complementarity of the Machine Learning and scaffold‐based techniques has not yet been comprehensively compared. In this study, we analysed filters based on the privileged scaffolds with filters built using machine learning. Our results demonstrate that machine‐learning methods provide more accurate filters for identification of frequent hitters in AlphaScreen assays than scaffold‐based methods and can be easily redeveloped once new data are measured. We present highly accurate models to identify frequent hitters in AlphaScreen assays. [ABSTRACT FROM AUTHOR]

The problem of designing new antiviral drugs against Human Cytomegalovirus (HCMV) was addressed using the Online Chemical Modeling Environment (OCHEM). Data on compound antiviral activity to human organisms were collected from the literature and uploaded in the OCHEM database. The predictive ability of the regression models was tested through cross-validation, giving coefficient of determination q2 = 0.71–0.76. The validation of the models using an external test set proved that the models can be used to predict the activity of newly designed compounds with reasonable accuracy within the applicability domain (q2 = 0.70–0.74). The models were applied to screen a virtual chemical library of imidazole derivatives, which was designed to have antiviral activity. The six most promising compounds were identified, synthesized and their antiviral activities against HCMV were evaluated in vitro. However, only two of them showed some activity against the HCMV AD169 strain. [ABSTRACT FROM AUTHOR]

A previously developed model to predict antibacterial activity of ionic liquids against a resistant A. baumannii strain was used to assess activity of phosphonium ionic liquids. Their antioxidant potential was additionally evaluated with newly developed models, which were based on public data. The accuracy of the models was rigorously evaluated using cross-validation as well as test set prediction. Six alkyl triphenylphosphonium and alkyl tributylphosphonium bromides with the C8, C10, and C12 alkyl chain length were synthesized and tested in vitro. Experimental studies confirmed their activity against A. baumannii as well as showed pronounced antioxidant properties. These results suggest that phosphonium ionic liquids could be promising lead structures against A. baumannii.

QSAR analysis of a set of previously synthesized phosphonium ionic liquids (PILs) tested against Gram‐negative multidrug‐resistant clinical isolate Acinetobacter baumannii was done using the Online Chemical Modeling Environment (OCHEM). To overcome the problem of overfitting due to descriptor selection, fivefold cross‐validation with variable selection in each step of the model development was applied. The predictive ability of the classification models was tested by cross‐validation, giving balanced accuracies (BA) of 76%–82%. The validation of the models using an external test set proved that the models can be used to predict the activity of newly designed compounds with a reasonable accuracy within the applicability domain (BA = 83%–89%). The models were applied to screen a virtual chemical library with expected activity of compounds against MDR Acinetobacter baumannii. The eighteen most promising compounds were identified, synthesized, and tested. Biological testing of compounds was performed using the disk diffusion method in Mueller‐Hinton agar. All tested molecules demonstrated high anti‐A. baumannii activity and different toxicity levels. The developed classification SAR models are freely available online at http://ochem.eu/article/113921 and could be used by scientists for design of new more effective antibiotics. [ABSTRACT FROM AUTHOR]

The EUOS/SLAS challenge aimed to facilitate the development of reliable algorithms to predict the aqueous solubility of small molecules using experimental data from 100 K compounds. In total, hundred teams took part in the challenge to predict low, medium and highly soluble compounds as measured by the nephelometry assay. This article describes the winning model, which was developed using the publicly available Online CHEmical database and Modeling environment (OCHEM) available on the website https://ochem.eu/article/27. We describe in detail the assumptions and steps used to select methods, descriptors and strategy which contributed to the winning solution. In particular we show that consensus based on 28 models calculated using descriptor-based and representation learning methods allowed us to obtain the best score, which was higher than those based on individual approaches or consensus models developed using each individual approach. A combination of diverse models allowed us to decrease both bias and variance of individual models and to calculate the highest score. The model based on Transformer CNN contributed the best individual score thus highlighting the power of Natural Language Processing (NLP) methods. The inclusion of information about aleatoric uncertainty would be important to better understand and use the challenge data by the contestants., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported., (Copyright © 2024. Published by Elsevier Inc.)

The problem of designing new antitubercular drugs against multiple drug‐resistant tuberculosis (MDR‐TB) was addressed using advanced machine learning methods. As there are only few published measurements against MDR‐TB, we collected a large literature data set and developed models against the non‐resistant H37Rv strain. The predictive accuracy of these models had a coefficient of determination q2 = .7–.8 (regression models) and balanced accuracies of about 80% (classification models) with cross‐validation and independent test sets. The models were applied to screen a virtual chemical library, which was designed to have MDR‐TB activity. The seven most promising compounds were identified, synthesized and tested. All of them showed activity against the H37Rv strain, and three molecules demonstrated activity against the MDR‐TB strain. The docking analysis indicated that the discovered molecules could bind enoyl reductase, InhA, which is required in mycobacterial cell wall development. The models are freely available online ( http://ochem.eu/article/103868) and can be used to predict potential anti‐TB activity of new chemicals. [ABSTRACT FROM AUTHOR]

Аbstract This paper describes Quantitative Structure-Activity Relationships (QSAR) studies, molecular docking and in vitro antibacterial activity of several potent imidazolium-based ionic liquids (ILs) against S. aureus ATCC 25923 and its clinical isolate. Small set of 131 ILs was collected from the literature and uploaded in the OCHEM database. QSAR methodologies used Associative Neural Networks and Random Forests (WEKA-RF) methods. The predictive ability of the models was tested through cross-validation, giving cross-validated coefficients q 2  = 0.82–0.87 for regression models and overall prediction accuracies of 80–82.1% for classification models. The proposed QSAR models are freely available online on OCHEM server at https://ochem.eu/article/107364 and can be used for estimation of antibacterial activity of new imidazolium-based ILs. A series of synthesized 1,3-dialkylimidazolium ILs with predicted activity were evaluated in vitro . The high activity of 7 ILs against S. aureus strain and its clinical isolate was measured and thereafter analyzed by the molecular docking to prokaryotic homologue of a eukaryotic tubulin FtsZ. [ABSTRACT FROM AUTHOR]

Inorganic nanomaterials have become one of the new areas of modern knowledge and technology and have already found an increasing number of applications. However, some nanoparticles show toxicity to living organisms, and can potentially have a negative influence on environmental ecosystems. While toxicity can be determined experimentally, such studies are time consuming and costly. Computational toxicology can provide an alternative approach and there is a need to develop methods to reliably assess Quantitative Structure–Property Relationships for nanomaterials (nano-QSPRs). Importantly, development of such models requires careful collection and curation of data. This article overviews freely available nano-QSPR models, which were developed using the Online Chemical Modeling Environment (OCHEM). Multiple data on toxicity of nanoparticles to different living organisms were collected from the literature and uploaded in the OCHEM database. The main characteristics of nanoparticles such as chemical composition of nanoparticles, average particle size, shape, surface charge and information about the biological test species were used as descriptors for developing QSPR models. QSPR methodologies used Random Forests (WEKA-RF), k-Nearest Neighbors and Associative Neural Networks. The predictive ability of the models was tested through cross-validation, giving cross-validated coefficients q 2 = 0.58–0.80 for regression models and balanced accuracies of 65–88% for classification models. These results matched the predictions for the test sets used to develop the models. The proposed nano-QSPR models and uploaded data are freely available online at http://ochem.eu/article/103451 and can be used for estimation of toxicity of new and emerging nanoparticles at the early stages of nanomaterial development. [ABSTRACT FROM AUTHOR]

Human ether-a-go-go related gene (hERG) K+ channel plays an important role in cardiac action potential. Blockage of hERG channel may result in long QT syndrome (LQTS), even cause sudden cardiac death. Many drugs have been withdrawn from the market because of the serious hERG-related cardiotoxicity. Therefore, it is quite essential to estimate the chemical blockage of hERG in the early stage of drug discovery. In this study, a diverse set of 3721 compounds with hERG inhibition data was assembled from literature. Then, we make full use of the Online Chemical Modeling Environment (OCHEM), which supplies rich machine learning methods and descriptor sets, to build a series of classification models for hERG blockage. We also generated two consensus models based on the top-performing individual models. The consensus models performed much better than the individual models both on 5-fold cross validation and external validation. Especially, consensus model II yielded the prediction accuracy of 89.5 % and MCC of 0.670 on external validation. This result indicated that the predictive power of consensus model II should be stronger than most of the previously reported models. The 17 top-performing individual models and the consensus models and the data sets used for model development are available at https://ochem.eu/article/103592. [ABSTRACT FROM AUTHOR]

To develop novel antimicrobial agents a series of 2(4)-hydrazone derivatives of quinoline were designed, synthesized and tested. QSAR models of the antibacterial activity of quinoline derivatives were developed by the OCHEM web platform using different machine learning methods. A virtual set of quinoline derivatives was verified with a previously published classification model of anti-E. coli activity and screened using the regression model of anti-S. aureus activity. Selected and synthesized 2(4)-hydrazone derivatives of quinoline exhibited antibacterial activity against the standard and antibiotic-resistant S. aureus and E. coli strains in the range from 15 to 30 mm by the diameter of growth inhibition zones. Molecular docking showed the complex formation of the studied compounds into the catalytic domain of dihydrofolate reductase with an estimated binding affinity from -8.4 to -9.4 kcal/mol., (© 2023 Wiley-VHCA AG, Zurich, Switzerland.)

Bioconcentration factors (BCFs) are markers of chemical substance accumulation in organisms, and they play a significant role in determining the environmental risk of various chemicals. Experiments to obtain BCFs are expensive and time-consuming; therefore, it is better to estimate BCF early in the chemical development process. The current research aims to evaluate the ecotoxicity potential of 122 pharmaceuticals and identify possible important structural attributes using BCF as the determining feature against a group of fish species. We have calculated the theoretical 2D descriptors from the OCHEM platform and SiRMS descriptor calculating software. The regression-based quantitative structure–property relationship (QSPR) modeling was used to identify the chemical features responsible for acute fish bioconcentration. Multiple models with the "intelligent consensus" algorithm were employed for the regression-based approach improving the predictive ability of the models. To ensure the robustness and interpretability of the developed models, rigorous validation was performed employing various statistical internal and external validation metrics. From the developed models, it can be specified that the presence of large lipophilic and electronegative moieties greatly enhances the bioaccumulative potential of pharmaceuticals, whereas the hydrophilic characteristics have shown a negative impact on BCF. Furthermore, the developed models were employed to screen the DrugBank database (https://go.drugbank.com/) for assessing the BCF properties of the entire database. The evidence acquired from the modeled descriptors might be used for aquatic risk assessment in the future, with the added benefit of providing an early caution of their probable negative impact on aquatic ecosystems for regulatory purposes. [ABSTRACT FROM AUTHOR]

Structural variation and different bioactivity of ionic liquids (ILs) make them highly promising for the development of novel biocides. Application of computational methods to the evaluation of potential antibacterial activity of chemical compounds is a useful, time- and cost-saving tool replacing numerous experimental syntheses. In the present study, quantitative structure–activity relationship (QSAR) modeling is applied to develop models (based on more than 800 data points) aiming to predict the minimal inhibitory concentration (MIC) of ILs against three types of human pathogens – Staphylococcus aureus , Escherichia coli and Pseudomonas aeruginosa. The random forest model with the AlvaDesc descriptors in general demonstrates the best performance for all the three types of bacteria and is suggested as a final model. To interpret the final model and determine the most significant descriptors, a SHapley Additive exPlanation (SHAP) method was applied. Six amino acid ILs, which were synthesized for the first time, and five halogenide ionic liquids purchased, all based on 1-alkyl-3methylimidozolium cations with different alkyl chain lengths, C 10 , C 12 and C 14 , are tested in vitro and used to validate the developed QSAR models. The data sets and developed model are available free of charge at http://ochem.eu/article/147386. [Display omitted] • QSAR model to predict the MIC values of ILs against three bacteria was developed. • We compiled the largest dataset (>800) on the MIC values of ILs against 3 bacteria. • IL properties which mostly governed the MIC value were determined with SHAP method. • The final model was validated with 11 ILs with 1-alkyl-3-methylimidazolium cations. [ABSTRACT FROM AUTHOR]

[Display omitted] • QSPR modeling of glass transition, melting and decomposition temperatures of ionic liquids was performed. • Rigorous component validation protocol provided better agreement of statistical parameters for the training and test sets. • Explanation of the models using statistical analysis of functional groups and Molecular Matched Pairs was provided. • Public freely-accessible models to predict phase transition and decomposition temperatures of ILs were contributed. The working temperature range of Ionic Liquids (IL) is determined by their liquid state range, where the IL's melting point/glass transition and decomposition temperatures define the lower and upper limits of the range, respectively. Computational prediction of the structure of new ILs with required properties, e.g. which can exist in a liquid state at room temperature and are stable up to high temperatures, is a much less time-consuming and less expensive approach than stepwise synthesis and experimental examination of all presupposed ILs. Therefore, in the present work the quantitative structure–property relationship (QSPR) models were developed to predict the glass transition temperature (T g), melting point (T m), and decomposition temperatures (T d) of ILs. We showed that a use of component validation protocol provided a better agreement of statistical parameters for the training and test sets. The performance of various modeling algorithms and descriptor sets was discussed and compared and advantages of descriptor-less as well as multi-task modeling were shown. An explanation of the models using statistical analysis of functional groups and Molecular Matched Pairs were provided for the mixtures for the first time. The experimental data and models, which are the first publicly available models for prediction of transition and decomposition temperatures of ILs, are publicly available online at http://ochem.eu/article/140250. [ABSTRACT FROM AUTHOR]

[Display omitted] • A machine learning model for molar absorption coefficient prediction was created. • The model was trained on data more 20,000 unique dye molecules. • The model is freely available as https://ochem.eu/article/145413. In this article, we provide a convenient tool for all researchers to predict the value of the molar absorption coefficient for a wide number of dyes without any computer costs. The new model is based on RFR method (ALogPS, OEstate + Fragmentor + QNPR) and is able to predict the molar absorption coefficient with an accuracy (5-fold cross-validation RMSE) of 0.26 log unit. This accuracy was achieved due to the fact that the model was trained on data for more than 20,000 unique dye molecules. To our knowledge, this is the first model for predicting the molar absorption coefficient trained on such a large and diverse set of dyes. The model is available at https://ochem.eu/article/145413. We hope that the new model will allow researchers to predict dyes with practically significant spectral characteristics and verify existing experimental data. [ABSTRACT FROM AUTHOR]

In this work, two porphyrins bearing [3,6-di-tert-butyl-carbazol-9-yl-benzoyloxy)]- (1) and [3,6-bis(3′,6′-di(tert -butyl)-9′H-carbazol)-9H-carbazolbenzoyloxy]phenyl (2) groups and their zinc (1Zn, 2Zn)/cobalt (1Co, 2Co) complexes were synthesized and studied by experimental and theoretical methods. The spectral parameters (UV–vis absorption/femtosecond transient absorption/fluorescence, IR, 1H NMR, mass spectra) of the compounds were observed. Their structure was also examined by the DFT method. The comparative study of the UV–vis spectra by the DFT/TDDFT calculation, and by the prediction of the Soret band maximum using machine learning methods, namely the consensus models based on the data of over 10000 porphyrin free bases and their complexes with metals was performed. The absorption maximum wavelength (Soret band) of porphyrins predicted with machine learning methods showed better agreement with the experimental data compared to the DFT/TDDFT calculation. The final consensus model is freely available at https://ochem.eu/article/145340 and can be used by the other researchers to obtain new functionalized porphyrins with desired optical properties. • New meso-carbazole substituted porphyrin complexes were synthesized. • They were characterized by UV–visible, infrared, fluorescence, transient absorption spectroscopy, and mass spectrometry. • Prediction of the Soret band maximum by DFT calculations and machine learning methods. [ABSTRACT FROM AUTHOR]

Background QSAR is an established and powerful method for cheap in silico assessment of physicochemical properties and biological activities of chemical compounds. However, QSAR models are rather complex mathematical constructs that cannot easily be interpreted. Medicinal chemists would benefit from practical guidance regarding which molecules to synthesize. Another possible approach is analysis of pairs of very similar molecules, so-called matched molecular pairs (MMPs). Such an approach allows identification of molecular transformations that affect particular activities (e.g. toxicity). In contrast to QSAR, chemical interpretation of these transformations is straightforward. Furthermore, such transformations can give medicinal chemists useful hints for the hit-to-lead optimization process. Results The current study suggests a combination of QSAR and MMP approaches by finding MMP transformations based on QSAR predictions for large chemical datasets. The study shows that such an approach, referred to as prediction-driven MMP analysis, is a useful tool for medicinal chemists, allowing identification of large numbers of "interesting" transformations that can be used to drive the molecular optimization process. All the methodological developments have been implemented as software products available online as part of OCHEM (http://ochem.eu/). Conclusions The prediction-driven MMPs methodology was exemplified by two use cases: modelling of aquatic toxicity and CYP3A4 inhibition. This approach helped us to interpret QSAR models and allowed identification of a number of "significant" molecular transformations that affect the desired properties. This can facilitate drug design as a part of molecular optimization process. [ABSTRACT FROM AUTHOR]

• A large-scale comparison of numerous QSPR models for six temperature-dependent properties of ionic liquids was performed. • None of the machine learning methods and molecular representations lead to the best models in all cases. • Only systematic enumeration reveals the best combinations of machine learning methods and molecular representations. • Transformer neural networks tend to provide better predictions of the properties of ionic liquids than other methods. • The properties of ionic liquids at different temperatures can be predicted using multi-task learning. The great importance of the ability to quantitatively predict the properties of ionic liquids (ILs) using quantitative structure–property relationships (QSPR) models necessitates the understanding of which modern machine learning (ML) methods in combination with which types of molecular representations are preferable to use for this purpose. To address this problem, a large-scale benchmarking study of QSPR models built by combining three traditional ML methods and neural networks with seven different architectures with five types of molecular representations (in the form of either numerical molecular descriptors or SMILES text strings) to predict six important physical properties of ILs (density, electrical conductance, melting point, refractive index, surface tension, and viscosity) was carried out. The datasets include from 407 to 1204 diverse ILs composed of various organic and inorganic ions. QSPR models for predicting the properties of ILs at eight different temperatures were built using multi-task learning. The best combinations of ML methods and molecular representations were identified for each of the properties. A unified ranking system was introduced to rank and prioritize different ML methods and molecular representations. It was shown in this study that on average: (i) nonlinear ML methods perform much better than linear ones, (ii) neural networks perform better than traditional ML methods, (iii) Transformers, which are actively used in natural language processing (NLP), perform better than other types of neural networks due to the advanced ability to analyze chemical structures of ILs encoded into SMILES text strings. A special "component-wise" cross-validation scheme was applied to assess how much the predictive performance deteriorates for the ILs composed of cations and anions that are not present in the dataset. [ABSTRACT FROM AUTHOR]

[Display omitted] • QSPR model to accurately predict the BODIPYs absorption wavelength was developed; • Largest database of 6000-plus fluorescent dyes was analyzed; • Excellent results for retrospective and prospective validation of BODIPYs; • The model is freely available as https://ochem.eu/article/134921. A possibility to accurately predict the absorption maximum wavelength of BODIPYs was investigated. We found that previously reported models had a low accuracy (40–57 nm) to predict BODIPYs due to the limited dataset sizes and/or number of BODIPYs (few hundreds). New models developed in this study were based on data of 6000-plus fluorescent dyes (including 4000-plus BODIPYs) and the deep neural network architecture. The high prediction accuracy (five-fold cross-validation room mean squared error (RMSE) of 18.4 nm) was obtained using a consensus model, which was more accurate than individual models. This model provided the excellent accuracy (RMSE of 8 nm) for molecules previously synthesized in our laboratory as well as for prospective validation of three new BODIPYs. We found that solvent properties did not significantly influence the model accuracy since only few BODIPYs exhibited solvatochromism. The analysis of large prediction errors suggested that compounds able to have intermolecular interactions with solvent or salts were likely to be incorrectly predicted. The consensus model is freely available at https://ochem.eu/article/134921 and can help the other researchers to accelerate design of new dyes with desired properties. [ABSTRACT FROM AUTHOR]

• The traditional validation of ionic liquids models based on random training/test split may overestimate accuracy of models. • A rigorous mixtures validation protocol provides better agreement of predicted and calculated performances. • Natural Language Processing (NLP) Transformer-CNN model overperforms traditional descriptor-based machine learning approach. • Public freely-accessible models to predict melting point of ionic liquids are contributed. The melting point (MP) of an ionic liquid (IL) is one of the key physical properties as it determines the lower limit of the IL working temperature range. In this work, we analysed the recently published studies to predict MP of ILs. While we were able to reproduce the statistical parameters reported by the authors, we found that the performance of the models with new test set data was much lower than the reported statistical values. The discrepancy was due to the validation protocol (random split of the initial set into training/test subsets) that did not allow correct estimation of how contributions of individual ions affect the model performance. Using a more rigorous validation protocol we reached good agreement between the training and test set statistical parameters. We strongly suggest using this protocol for proper validation of models for other properties of ILs to avoid reporting overoptimistic statistical parameters. We also showed that the Transformer Convolutional Neural Network, which was based on the representation of molecules as text (SMILES), proposed a model with significantly higher prediction accuracy as compared to those developed using descriptors that were used in the previous studies. The RMSE of this model is 44 °C and the model is applicable to any type of ILs. The data and developed models are publicly available online at http://ochem.eu/article/135195. [ABSTRACT FROM AUTHOR]

Online Chemical Modeling Environment (OCHEM) was used for QSAR analysis of a set of ionic liquids (ILs) tested against multi-drug resistant (MDR) clinical isolate Acinetobacter baumannii and Staphylococcus aureus strains. The predictive accuracy of regression models has coefficient of determination q2 = 0.66 − 0.79 with cross-validation and independent test sets. The models were used to screen a virtual chemical library of ILs, which was designed with targeted activity against MDR Acinetobacter baumannii and Staphylococcus aureus strains. Seven most promising ILs were selected, synthesized, and tested. Three ILs showed high activity against both these MDR clinical isolates. [ABSTRACT FROM AUTHOR]

• Antiurease activity of newly designed and synthesized thiazole-2-imine derivatives. • Docking studies were performed for all compounds and interaction modes with enzyme active sites were determined. • Machine learning method was performed to choose the most active compounds. Machine learning is one of the methods used in the design of new molecules with different biological properties and has become a trend in recent years. Since there are many published studies on urease enzyme inhibition, we accumulated a huge literatüre data set and improved a model for antiurease activity. The balanced accuracy of the selected compounds (classification models) were about 78% and the predictive accuracy of them possessed a coefficient of determination q 2 = 0.2-0.7 (regression models) with cross-validation and independent test sets. Thanks to the chemical library created with the machine learning method, a comparison of the predictive and experimental results of the compounds that previously synthesized by us and investigated urease inhibition was made. Compounds observed to be experimentally active were found to be active with the machine learning method. The models are freely avaible online (http://ochem.eu) and can be used to predict potantial antiurease activity of novel compounds. The activity potentials of compounds 4a-d were further evaluated via molecular docking studies with AutoDock4 and AutoDock Vina softwares. Urease enzyme Image, graphical abstract [ABSTRACT FROM AUTHOR]

A matched molecular pair (MMP) analysis was used to examine the change in melting point (MP) between pairs of similar molecules in a set of ∼275k compounds. We found many cases in which the change in MP (ΔMP) of compounds correlates with changes in functional groups. In line with the results of a previous study, correlations between ΔMP and simple molecular descriptors, such as the number of hydrogen bond donors, were identified. In using a larger dataset, covering a wider chemical space and range of melting points, we observed that this method remains stable and scales well with larger datasets. This MMP-based method could find use as a simple privacy-preserving technique to analyze large proprietary databases and share findings between participating research groups., (© 2017 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA.)