Abraham Parameters Research Articles

Machine Learning Models to Predict Early Breakthrough of Recalcitrant Organic Micropollutants in Granular Activated Carbon Adsorbers.

Granular activated carbon (GAC) adsorption is frequently used to remove recalcitrant organic micropollutants (MPs) from water. The overarching aim of this research was to develop machine learning (ML) models to predict GAC performance from adsorbent, adsorbate, and background water matrix properties. For model calibration, MP breakthrough curves were compiled and analyzed to determine the bed volumes of water that can be treated until MP breakthrough reaches ten percent of the influent MP concentration (BV10). Over 400 data points were split into training, validation, and testing sets. Seventeen variables describing MP, background water matrix, and GAC properties were explored in ML models to predict log10-transformed BV10 values. Using the ML models on the testing set, predicted BV10 values exhibited mean absolute errors of ∼0.12 log units and were highly correlated with experimentally determined values (R2 ≥ 0.88). The top three drivers influencing BV10 predictions were the air-hexadecane partition coefficient and hydrogen bond acidity (Abraham parameters L and A) of the MPs and the dissolved organic carbon concentration of the GAC influent water. The model can be used to rapidly estimate the GAC bed life, select effective GAC products for a given treatment scenario, and explore the suitability of GAC treatment for remediating emerging MPs.

Predicting Critical Properties and Acentric Factors of Fluids Using Multitask Machine Learning.

Knowledge of critical properties, such as critical temperature, pressure, density, as well as acentric factor, is essential to calculate thermo-physical properties of chemical compounds. Experiments to determine critical properties and acentric factors are expensive and time intensive; therefore, we developed a machine learning (ML) model that can predict these molecular properties given the SMILES representation of a chemical species. We explored directed message passing neural network (D-MPNN) and graph attention network as ML architecture choices. Additionally, we investigated featurization with additional atomic and molecular features, multitask training, and pretraining using estimated data to optimize model performance. Our final model utilizes a D-MPNN layer to learn the molecular representation and is supplemented by Abraham parameters. A multitask training scheme was used to train a single model to predict all the critical properties and acentric factors along with boiling point, melting point, enthalpy of vaporization, and enthalpy of fusion. The model was evaluated on both random and scaffold splits where it shows state-of-the-art accuracies. The extensive data set of critical properties and acentric factors contains 1144 chemical compounds and is made available in the public domain together with the source code that can be used for further exploration.

A new computational method for drug solubility prediction in methanol + water mixtures

General solubility prediction models were developed using both classic least square and a novel method, in order to predict the solubility of the solutes in methanol + water binary solvent system. The novel approach to the regression analysis was investigated using an error minimization method. This aim was achieved by using a user-defined loss function regression, instead of the classic least square regression approach. To examine the results of the novel methodology, previous solubility data of 41 solutes were used for comparison. Both least square and novel methods were applied to the Jouyban-Acree model, Jouyban-Acree model in combination with Abraham parameters, and the modified Wilson model. The generally trained versions of the mentioned models produced more accurate predictions using the novel method than the least square method that has been confirmed by t-test analyses. The Jouyban-Acree model was the most accurate model among other generally trained models. Finally, the results were validated using a cross-validation analysis which produced the acceptable prediction accuracy of 24.6% mean percentage deviation (MPD) for the new methodology against 32.1% of the least square method. Also a new arithmetically transformed version of aforementioned models was introduced in this study to make the calculations easier to execute.

Target site model: Predicting mode of action and aquatic organism acute toxicity using Abraham parameters and feature-weighted k-nearest neighbors classification.

A database of 1480 chemicals with 47 associated modes of action compiled from the literature encompasses a wide range of chemical classes (alkanes, polycyclic aromatic hydrocarbons, pesticides, and polar compounds) and includes toxicity data for 79 different aquatic genera. The data were split into a calibration group and a validation group (80/20) to apply k-nearest neighbors (k-NN) methodology to predict the toxic mode of action for the compound. Other approaches were tested (support vector machines and linear discriminant analysis) as well as variations in the k-NN technique (distance weighting, feature weighting). Best-prediction results were found with k = 3, in a voting platform with optimized feature weighting. Using the predicted mode of action, the appropriate polyparameter target site model for that mode of action is applied to calculate the 50% lethal concentration (LC50). Predicted LC50s for the validation database resulted in a root-mean squared error (RMSE) of 0.752. This can be compared to an RMSE of 0.655 for the same validation set using the reference mode of action labels. The complete database resulted in an RMSE of 0.793 for reference mode of action labels. This confirms that the classification model has sufficient accuracy for predicting the mode of action and for determining toxicity using the target site model. Environ Toxicol Chem 2019;38:375-386. © 2018 SETAC.

Modeling Sorptive Fractionation of Organic Matter at the Mineral‐Water Interface

Core IdeasOM fractionation is studied for six minerals and more than 180 OM moieties.Mineral surface variability is explained by the pp‐LFER system parameters.OM functionality variability is quantified by the Abraham parameters.Factors governing fractionation are polarity, H‐bonding, dispersion, and cavitation.A link is developed between sorption and OM moiety molecular characteristics.Sorptive fractionation of organic matter (OM) at mineral–water interfaces alters the composition of OM and therefore affects the fate, transport, and bioavailability of nutrients and contaminants in terrestrial and aquatic environments. It is essential to identify the fractions of OM that are selectively sorbed on minerals and to elucidate the mechanisms of sorptive fractionation. In this study, the partition between mineral and water for a large set of organic molecules covering the most relevant OM functional groups and building blocks were predicted using the mechanistic Polyparameter Linear Free Energy Relationships (pp‐LFERs). The results were used to infer the sorptive fractionation tendency of OM moieties. It is found that for a OM moiety belonging to the chemical classes of alkanes, alkenes, or benzenes, its sorption to bentonite is the strongest, followed by kaolinite, talc or hematite, and Al2O3or quartz, relative to water. For a moiety with either double bond, or –OH, or –COOH, or R–COO–R′, or R–O–R′, or R–COH, or R–CO–R′, or –NH2, its partition to water is enhanced as compared to the respective alkane moiety. Opposite is found for thiols and sulfides. Meanwhile, a correlation is developed, that is, as the double bond equivalence decreases and the number of C atoms increases, the degree of sorption increases. Overall, the key factors controlling the sorptive fractionation are polarity, H‐bonding, van de Waals dispersion, and the energy for cavity formation. This work provides insights into the OM sorption at the mineral and water interface at the molecular level.

Viscosity prediction of ionic liquid + molecular solvent mixtures at various temperatures

To evaluate the ability of a combined form of the Jouyban-Acree model with the Abraham parameters, the viscosity data for 84 sets of ionic liquid + molecular solvent binary mixtures at various temperatures were predicted. Percent average relative deviations (%ARD) of the predicted and experimental values were calculated as an accuracy criterion and were obtained as 19.1 and 20.1 by considering the theoretical and experimental Abraham parameters, respectively. The collected data were also correlated using the Jouyban-Acree model, which is resulted in the overall %ARDs of 16.8 and 18.1 for back-calculated data by regarding the theoretical and experimental Abraham parameters, respectively. Moreover, to update the previously proposed global model, the earlier data were merged with the collected data from recent publications and total data sets were used to retrain the global Jouyban-Acree model by regarding the theoretical and experimental Abraham parameters. Comparing %ARD of the proposed model using correlated newly data and total (new and earlier) data sets show no statistically significant difference while there is a meaningful difference between calculations by considering the theoretical and experimental Abraham parameters. Moreover, to compare the capability of the Jouyban-Acree model for correlating viscosities of binary liquid mixtures, the overall %ARD were determined and compared to three selected models; Wilson, Katti and Chaudhary, and Eyring-NRTL. The order of the amount of obtained %ARD are Jouyban-Acree < Wilson < Katti and Chaudhary < Eyring-NRTL which is indicative of the suitability of Jouyban-Acree for correlation and prediction of the viscosity data for the binary liquid mixtures.

How Predictable Are Human Stratum Corneum Lipid/Water Partition Coefficients? Assessment and Useful Correlations for Dermal Absorption

Partition coefficients between human stratum corneum lipids and water (Ksclip/w) are collected or deduced from a variety of sources in a manner that approximately doubles the available data compared to the current state-of-the-art model (Hansen et al., Adv Drug Deliv Rev. 2013;65(2):251-264). An additional datum for water itself in porcine SC that considerably extends the molecular size and lipophilicity range of the data set is considered. The data are analyzed in terms of an extended linear free energy relationship involving octanol/water partition coefficients, Abraham solvation parameters, and a secondary, power law molecular weight dependence. The optimum fit to log Ksclip/w for the full data set reduces the standard error of prediction from 0.50 for a Hansen-like model to 0.39; corresponding multiplicative errors in Ksclip/w are reduced from a factor of 3.1 to one of 2.5. The difference in performance is driven by the water datum, which requires a more complex dependence on molecular size than that afforded by Abraham parameters. In the absence of the water value, the Hansen-like model, which does not include a dependence on molecular size, is essentially optimum. A comparison is presented to fluid-phase phospholipid–water systems, which have a demonstrably different structure–property relationship.

Quantum Chemically Estimated Abraham Solute Parameters Using Multiple Solvent-Water Partition Coefficients and Molecular Polarizability.

Polyparameter Linear Free Energy Relationships (pp-LFERs), also called Linear Solvation Energy Relationships (LSERs), are used to predict many environmentally significant properties of chemicals. A method is presented for computing the necessary chemical parameters, the Abraham parameters (AP), used by many pp-LFERs. It employs quantum chemical calculations and uses only the chemical's molecular structure. The method computes the Abraham E parameter using density functional theory computed molecular polarizability and the Clausius-Mossotti equation relating the index refraction to the molecular polarizability, estimates the Abraham V as the COSMO calculated molecular volume, and computes the remaining AP S, A, and B jointly with a multiple linear regression using sixty-five solvent-water partition coefficients computed using the quantum mechanical COSMO-SAC solvation model. These solute parameters, referred to as Quantum Chemically estimated Abraham Parameters (QCAP), are further adjusted by fitting to experimentally based APs using QCAP parameters as the independent variables so that they are compatible with existing Abraham pp-LFERs. QCAP and adjusted QCAP for 1827 neutral chemicals are included. For 24 solvent-water systems including octanol-water, predicted log solvent-water partition coefficients using adjusted QCAP have the smallest root-mean-square errors (RMSEs, 0.314-0.602) compared to predictions made using APs estimated using the molecular fragment based method ABSOLV (0.45-0.716). For munition and munition-like compounds, adjusted QCAP has much lower RMSE (0.860) than does ABSOLV (4.45) which essentially fails for these compounds.

Predicting solvent-water partitioning of charged organic species using quantum-chemically estimated Abraham pp-LFER solute parameters

Methods for obtaining accurate predictions of solvent-water partitioning for neutral organic chemicals (e.g., Kow) are well-established. However, methods that provide comparable accuracy are not available for predicting the solvent-water partitioning of ionic species. Previous methods for addressing charge contributions to solvent-water partitioning rely on charged solute descriptors which are obtained from regressions to neutral species descriptors as well as charged descriptors which are specific to unique charge-functionalities and structural moieties. This paper presents a method for obtaining Abraham poly-parameter linear free energy relationship (pp-LFER) descriptors using quantum chemical calculations and molecular structure, only. The method utilizes a large number of solvent-water systems to overcome large errors in individual quantum chemical computations of ionic solvent-water partition coefficients. The result is a single set of quantum-chemically estimated Abraham solute parameters (QCAP) which are solvent-independent, and can be used to predict the solvent-water partitioning of ionic species.Predictions of solvent-water partition coefficients for ionic species using quantum-chemically estimated Abraham parameters (QCAPs) are shown to provide improved accuracy compared over both existing Absolv-estimated Abraham solute parameters (AAP) as well as direct a priori quantum chemical (QC) calculations for partitioning of anionic solutes in 4 organic solvent-water systems (RMS = 0.740, 2.48 and 0.426 for the Absolv, QC and QCAP methods, respectively). For quaternary amine cations in the octanol-water system the RMS errors of the solvent-water partition coefficients were larger and similar between the two Abraham models (RMSE = 0.997 and 1.16, for the AAP and QCAP methods, respectively). Both methods showed significant improvement over direct QC calculations (RMSE = 2.82).

Solubility of naproxen in 2-propanol + water mixtures at various temperatures

The experimental solubility of naproxen in 2-propanol+water mixtures at different temperatures (298.2, 303.2, 308.2 and 313.2K) was reported. The solubility was correlated and/or predicted using three numerical methods; i.e., the combined van't Hoff equation and Jouyban–Acree model (method I), the extended version of the Jouyban–Acree model with Abraham parameters (method II), and the minimum number of data points (N=10) were used to train the combined model. The accuracies of the calculated solubilities were evaluated by computing the mean percentage deviation (MPD). The obtained MPDs for investigated numerical analyses varied between 13 and 30%.

Solubility of atenolol in ethanol+ water mixtures at various temperatures

The experimental solubility of atenolol in ethanol + water mixtures at different temperatures (298.2, 303.2, 308.2 and 313.2 K) was reported. The solubility was calculated with five numerical methods; i.e., the Jouyban-Acree model (method I), its combination with van’t Hoff equation (method II), the extended version of the Jouyban-Acree model with Abraham parameters (method III), the minimum number of data points (N) were used to train the Jouyban-Acree model (N=11) and its combination with van’t Hoff equation (N=22), then obtained parameters of the models were used to calculate the solubilites at other temperatures (methods IV and V). The accuracies of the calculated solubilites were evaluated by computing mean percentage deviation (MPD). The obtained MPDs (± standard deviation) for methods I, II, III, IV and V were 5.6±7.1, 5.1±4.6, 34.1±28.0, 10.0±9.6 and 6.6±4.8%, respectively.

A reductionist mechanistic model for bioconcentration of neutral and weakly polar organic compounds in fish

The bioconcentration factor (BCF) of neutral and weakly polar organic chemicals in fish is modeled using independently calibrated models of chemical partitioning (freely dissolved fraction of chemical in the aqueous phase [φsys ] and wet-weight fish-water partition coefficient [KFW ]), respiratory exchange (respiratory update rate constant [k1 ], and respiratory elimination rate constant [k2 = k1 /KFW ]), and biotransformation (whole-body biotransformation rate constant [kM ]) as BCF = φsys KFW /(1 + kM /k2 ). Existing k1 models tend to overestimate for chemicals with log KOW < 3.5, which constituted 30% to 50% of the examined chemicals. A revised k1 model covering a wider log KOW range (0-8.5) is presented k1 = (5.46 × 10(-6) MW + 0.261/KOW )(-1) , where MW is the molecular weight. The biotransformation rate constant kM is modeled using biota internal partitioning and Abraham parameters as reactivity descriptors. The reductionist model was tested using 3 different BCF data sets (US Environmental Protection Agency's Estimation Programs Interface [EPI], n = 548; Hertfordshire, n = 210; Arnot-Gobas, n = 1855) and compared with the following 3 state-of-the-art models: 1) the EPI Suite BCFBAF module, 2) the European Commision's Computer Assisted Evaluation of industrial chemical Substances According to Regulations (CAESAR), and 3) the EPI/Arnot mechanistic kinetic model. The reductionist model performed comparably with the alternative models (root mean square errors [RMSEs] = 0.72-0.77), with only 5 fitting parameters and no training against experimental BCFs. Respiratory elimination and biotransformation dominate the total depuration (i.e., [k2 + kM ]/kT ≥ 0.8) for approximately 98% of the data entries, thus validating the reductionist approximation. Mechanistic models provide greater insights into bioaccumulation and are more sensitive to biological variation. All three BCF data sets and relevant properties and checkpoint values necessary for reproducing predictions of the reductionist model have been documented. The present study shows that a streamlined mechanistic model of BCF is possible for assessment purposes.

Mathematical Representation of Viscosity of Ionic Liquid + Molecular Solvent Mixtures at Various Temperatures Using the Jouyban–Acree Model

The viscosity of room temperature ionic liquid + molecular solvents is correlated using the Jouyban–Acree model, and the accuracy of the model is evaluated using average relative deviations (ARD) of the calculated and experimental values. The overall ARD of 6.9 % was obtained for the proposed model where the overall ARD for a similar model from the literature was 22.4 %. The mean difference between ARDs of the models was statistically significant. Combined forms of the Jouyban–Acree model and the Abraham parameters were used for predicting the viscosity of the mixtures at various temperatures. The overall ARD for the investigated mixtures was 15.0 % when the experimental viscosity of the pure solvents was included. The ARD of the in silico version of the model was 20.7 %.

Application of Abraham solvation parameters for surface tension prediction of mono-solvents and solvent mixtures at various temperatures

A combined form of the Jouyban–Acree model and the Abraham solvation parameters was used to predict the surface tension of binary and ternary solvent mixtures at various temperatures. The new equations have been tested by comparisons with the experimental results for 364 binary and 79 ternary mixtures reported in the literature. After the training procedures, the models were used for the prediction of the surface tension of mixed solvent systems. The mean relative deviations (MRDs) between predicted and experimental surface tensions are used as an accuracy criterion. The overall MRD for binary mixtures was 6.5% and the corresponding value for ternary mixtures was 11.2%. These versions of the models require the experimental surface tension of the mono-solvents as input data. An attempt has been made here to correlate the surface tension of the mono-solvents at various temperatures using Abraham parameters. The model calculates the surface tension with the MRD of 11.0%. A combined version of the models was presented to predict the surface tension of binary and ternary mixtures without using any experimental input data in which the MRDs of 14.6 and 11.7% were obtained. Results show that the addition of solvation parameters may lead to an improvement in the calculated surface tension and show good agreement between theory and experiment.

Study of Intermolecular Interactions Involved in Capillary GLC with Liquid Crystal Compounds as Stationary Phases

Abraham's equation was employed in order to investigate nonspecific intermolecular interactions involving liquid crystals. Several aromatic azo derivatives, proved to have liquid crystal behavior, were used as stationary phases in capillary gas chromatography. The polarizability, polarity, hydrophobicity and hydrogen-bond donor or acceptor characters were estimated using the Abraham solvation model. Calculations were based on the isothermal retention times for 25 compounds, in the temperature range between 70˚ and 170˚C, both in the heating and the cooling mode. The descriptors R, � , �� , �� , and logL 16 (Abraham parameters) were converted into orthogonal descriptors following Randi� 's procedure. The above-mentioned properties involve the r, s, a, b and l coefficients, respectively, which were determined by means of multilinear regression applied to the considered set. Following the orthogonalization procedure, the polarizability term (rR*) was found to be statistically insignificant in the case of the studied compounds. Plots of these coefficients vs temperature are discussed, in order to allow estimations regarding the retention mechanism and contributions of polar and Van der Waals interactions. The variation of properties (e.g. polarity, hydrogen-bond donor and hydrogen-bond acceptor character) with temperature was assigned to the changes in the molecular geometry of the compound used as stationary phase.

Quantum chemical characterization of Abraham solvation parameters for gas–liquid chromatographic stationary phases

Quantum chemical based investigation is presented on the Abraham solvation parameters for 23 molecular (non-polymeric) GLC stationary phases. PM6 semiempirical calculations combined with conductor-like screening model (COSMO) have been utilized. Comprehensive search for an optimal model was carried out, based on best subset selection from 86 variables considered. A unified quantitative structure–property relationship model has been developed for all five Abraham parameters reported. The selected set of five structure-driven descriptors was subjected to statistical analyses, and was shown to be useful for stationary phase classification.

Prediction of Physicochemical Properties of Organic Compounds from 2D Molecular Structure – Fragment Methods vs. LFER Models

A large number of models is available to predict physicochemical properties directly from the two-dimensional molecular structure. An alternative to conventional fragment methods is given by linear free-energy relationships (LFERs) employing Abraham parameters. The latter have a solid mechanistic background, but a drawback in practice is the limited availability of Abraham parameters for substances of interest. As a consequence, more complex compounds typically require the estimation of Abraham parameters from the chemical structure. Comparative analysis of prediction methods for Henry's law constant and sorption to soil organic matter shows that at present, fragment methods are superior to the LFER approach when employing calculated Abraham parameters. For the subset of typically more simple compounds with experimental Abraham parameters, the respective LFERs are competitive to general-purpose fragment models. The discussion includes analyses of compound subsets without and with hydrogen bond functionalities, and of the impact of structural complexity on the model performance.

Molecular dynamics study of pressure effect on glass formation and the crystallization in liquid CuNi alloy

Abstract The cooling rate and pressure are two critical factors affecting the resulting structure of a liquid–solid transition. The pressure effect on the rapid solidification of liquid CuNi alloys is studied by using molecular dynamic simulation method. To model the interactions between atoms Sutton–Chen type of embedded atom method based on many-body interaction were used. Radial distribution function (RDF), the changes of volume and Wendt–Abraham parameters were calculated to determine temperatures of glass and crystal formation for model alloy system under different pressures. Crystal and amorphous transitions for different pressures were formed at cooling rate of 4 × 10 12 and 4 × 10 13 K/s, respectively. It is observed that the increase of pressure causes in increase of glass and crystal formation temperatures.

Quantitative structure–property relationships for pesticides in biopartitioning micellar chromatography

The retention factor (log k) in the biopartitioning micellar chromatography (BMC) of 79 heterogeneous pesticides was studied by quantitative structure–property relationships (QSPR) method. Heuristic method (HM) and support vector machine (SVM) method were used to build linear and nonlinear models, respectively. Compared the results of these two methods, those obtained by the SVM model are much better. For the test set, a predictive correlation coefficient ( R) of 0.9755 and root-mean-square (RMS) error of 0.1403 were obtained. The proposed QSPR models, both by HM and SVM, contain the same descriptors that agree with the classical Abraham parameters of well-known linear solvation energy relationships (LSER).

Kinetic Solvent Effects on Proton and Hydrogen Atom Transfers from Phenols. Similarities and Differences

Bimolecular rate constants for proton transfer from six phenols to the anthracene radical anion have been determined in up to eight solvents using electrochemical techniques. Effects of hydrogen bonding on measured rate constants were explored over as wide a range of phenolic hydrogen-bond donor (HBD) and solvent hydrogen-bond acceptor (HBA) activities as practical. The phenols' values ranged from 0.261 (2-MeO-phenol) to 0.728 (3,5-Cl(2)-phenol), and the solvents' values from 0.44 (MeCN) to 1.00 (HMPA), where and are Abraham's parameters describing relative HBD and HBA activities (J. Chem. Soc., Perkin Trans. 2 1989, 699; 1990, 521). Rate constants for H-atom transfer (HAT) in HBA solvents, k(S), are extremely well correlated via log k(S) = log k(0) - 8.3 , where k(0) is the rate constant in a non-HBA solvent (Snelgrove et al. J. Am. Chem. Soc. 2001, 123, 469). The same equation describes the general features of proton transfers (k(S) decreases as increases, slopes of plots of log k(S) against increase as increases). However, in some solvents, k(S) values deviate systematically from the least-squares log k(S) versus correlation line (e.g., in THF and MeCN, k(S) is always smaller and larger, respectively, than "expected"). These deviations are attributed to variations in the solvents' anion solvating abilities (THF and MeCN are poor and good anion solvators, respectively). Values of log k(S) for proton transfer, but not for HAT, give better correlations with Taft et al.'s (J. Org. Chem. 1983, 48, 2877) beta scale of solvent HBA activities than with . The beta scale, therefore, does not solely reflect solvents' HBA activities but also contains contributions from anion solvation.