Group Contribution and Machine Learning Approaches to Predict Abraham Solute Parameters, Solvation Free Energy, and Solvation Enthalpy.

We present a group contribution method (SoluteGC) and a machine learning model (SoluteML) to predict the Abraham solute parameters, as well as a machine learning model (DirectML) to predict solvation free energy and enthalpy at 298 K. The proposed group contribution method uses atom-centered functional groups with corrections for ring and polycyclic strain while the machine learning models adopt a directed message passing neural network. The solute parameters predicted from SoluteGC and SoluteML are used to calculate solvation energy and enthalpy via linear free energy relationships. Extensive data sets containing 8366 solute parameters, 20,253 solvation free energies, and 6322 solvation enthalpies are compiled in this work to train the models. The three models are each evaluated on the same test sets using both random and substructure-based solute splits for solvation energy and enthalpy predictions. The results show that the DirectML model is superior to the SoluteML and SoluteGC models for both predictions and can provide accuracy comparable to that of advanced quantum chemistry methods. Yet, even though the DirectML model performs better in general, all three models are useful for various purposes. Uncertain predicted values can be identified by comparing the three models, and when the 3 models are combined together, they can provide even more accurate predictions than any one of them individually. Finally, we present our compiled solute parameter, solvation energy, and solvation enthalpy databases (SoluteDB, dGsolvDBx, dHsolvDB) and provide public access to our final prediction models through a simple web-based tool, software packages, and source code.

Properties of the tert-butyl halide solvolysis transition states.

We have obtained properties (or descriptors) of the transition states in the solvolysis of tert-butyl chloride, bromide and iodide. We show that all three transition states, in both protic and in aprotic solvents, are highly dipolar and are strong hydrogen bond acids and strong hydrogen bond bases, except for the tert-butyl iodide transition state in aprotic solvents, which has a rather low hydrogen bond acidity. Thus, the transition states are stabilized by solvents that are hydrogen bond bases (nucleophiles) and are hydrogen bond acids (electrophiles). We show also that the partition of the transition states between water and solvents is aided by both nucleophilic and electrophilic solvents and conclude that the rate of solvolysis of the three halides is increased by both nucleophilic and electrophilic solvents.

Modeling Aquatic Toxicity through Chromatographic Systems.

Environmental risk assessment requires information about the toxicity of the growing number of chemical products coming from different origins that can contaminate water and become toxicants to aquatic species or other living beings via the trophic chain. Direct toxicity measurements using sensitive aquatic species can be carried out but they may become expensive and ethically questionable. Literature refers to the use of chromatographic measurements that correlate to the toxic effect of a compound over a specific aquatic species as an alternative to get toxicity information. In this work, we have studied the similarity in the response of the toxicity to different species and we have selected eight representative aquatic species (including tadpoles, fish, water fleas, protozoan, and bacteria) with known nonspecific toxicity to chemical substances. Next, we have selected four chromatographic systems offering good perspectives for surrogation of the eight selected aquatic systems, and thus prediction of toxicity from the chromatographic measurement. Then toxicity has been correlated to the chromatographic retention factor. Satisfactory correlation results have been obtained to emulate toxicity in five of the selected aquatic species through some of the chromatographic systems. Other aquatic species with similar characteristics to these five representative ones could also be emulated by using the same chromatographic systems. The final aim of this study is to model chemical products toxicity to aquatic species by means of chromatographic systems to reduce in vivo testing.

Dose-Response Functions for the Olfactory, Nasal Trigeminal, and Ocular Trigeminal Detectability of Airborne Chemicals by Humans.

We gathered from the literature 47 odor and 37 trigeminal (nasal and ocular) chemesthetic psychometric (i.e., detectability or dose-response) functions from a group of 41 chemicals. Vapors delivered were quantified by analytical methods. All functions were very well fitted by the sigmoid (logistic) equation: y = 1 / (1 + e({-(x-C)/D})), where parameter C quantifies the detection threshold concentration and parameter D the steepness of the function. Odor and chemesthetic functions showed no concentration overlap: olfactory functions grew along the parts per billion (ppb by volume) range or lower, whereas trigeminal functions grew along the part per million (ppm by volume) range. Although, on average, odor detectability rose from chance detection to perfect detection within 2 orders of magnitude in concentration, chemesthetic detectability did it within one. For 16 compounds having at least 1 odor and 1 chemesthetic function, the average gap between the 2 functions was 4.6 orders of magnitude in concentration. A quantitative structure-activity relationship (QSAR) using 5 chemical descriptors that had previously described stand-alone odor and chemesthetic threshold values, also holds promise to describe, and eventually predict, olfactory and chemesthetic detectability functions, albeit functions from additional compounds are needed to strengthen the QSAR.

Comment on "structural determinants of drug partitioning in surrogates of phosphatidylcholine bilayer strata".

The Abraham model correlation for describing the partitioning behavior of solutes between water and hexadecane was redetermined using the measured partition coefficient data and solute descriptors derived from experimental data. The newly derived correlation provides a much better mathematical description of the observed partition coefficient data than the correlation given in the published paper.

Is there an intramolecular hydrogen bond in 2-halophenols? A theoretical and spectroscopic investigation.

The Abraham solute hydrogen bond acidity parameter A can be derived both from physical methods, A(Gen) and NMR experiments, A(NMR) and results for a large number of hydroxylic solutes show that the two methods agreed very well. However for halophenols the values of A(NMR) were not consistent with the A(Gen) values. The values of A(NMR) suggest that there is no intra-molecular hydrogen bonding in any of the 2-halophenols. In contrast the values of A(Gen) indicate that there is no intra-molecular hydrogen bonding in 2-fluorophenol, but weak intra-molecular hydrogen bonding in 2-chloro, 2-bromo, and 2-iodo-phenol. In view of this uncertainty in the presence or absence of intra-molecular H-bonds in the 2-halophenols, a detailed investigation of the methods used in the literature is presented together with a novel NMR method to determine the ratio of cis and trans forms in these compounds. The experimental data is complemented by a detailed theoretical analysis of the structures and bonding in these molecules to assess the presence or absence of an intra-molecular H-bond. We conclude that there is weak hydrogen bonding in 2-chloro, 2-bromo and 2-iodophenol but very little in 2-fluorophenol.

Correction: Is there an intramolecular hydrogen bond in 2-halophenols? A theoretical and spectroscopic investigation.

Correction for 'Is there an intramolecular hydrogen bond in 2-halophenols? A theoretical and spectroscopic investigation' by Michael H. Abraham et al., Phys. Chem. Chem. Phys., 2015, DOI: 10.1039/c5cp04061b.

An NMR method for the quantitative assessment of intramolecular hydrogen bonding; application to physicochemical, environmental, and biochemical properties.

(1)H NMR chemical shifts have been obtained in the solvents deuterochloroform and dimethyl sulfoxide. The difference in the chemical shifts of an OH or NH group in these two solvents, Δδ = δ(DMSO) - δ(CDCl3), can be converted into the hydrogen bond acidity, A, of the group using the equation A = 0.0065 + 0.133Δδ. The NMR A value, ANMR, can be used as a quantitative assessment of intramolecular hydrogen bonding. We list values of Δδ and ANMR for 55 compounds containing an OH group and 60 compounds with an NH group. For the hydroxy compounds, if A > 0.5 then the OH group is not part of an intramolecular hydrogen bond, but if A < 0.1 then the OH group forms part of an intramolecular hydrogen bond. For NH compounds, if A > 0.16 the NH group is not part of an intramolecular hydrogen bond, and if A < 0.05 the NH group is part of an intramolecular hydrogen bond. No comparison compounds are needed, and the method is extremely simple. We further show how it is possible to relate intramolecular hydrogen bonding to the actual effect on values of a number of physicochemical, environmental, and biochemical properties.

An algorithm for 353 odor detection thresholds in humans.

One hundred and ninety three odor detection thresholds, ODTs, obtained by Nagata using the Japanese triangular bag method can be correlated as log (1/ODT) by a linear equation with R(2) = 0.748 and a standard deviation, SD, of 0.830 log units; the latter may be compared with our estimate of 0.66 log units for the self-consistency of Nagata's data. Aldehydes, acids, unsaturated esters, and mercaptans were included in the equation through indicator variables that took into account the higher potency of these compounds. The ODTs obtained by Cometto-Muñiz and Cain, by Cometto-Muñiz and Abraham, and by Hellman and Small could be put on the same scale as those of Nagata to yield a linear equation for 353 ODTs with R(2) = 0.759 and SD = 0.819 log units. The compound descriptors are available for several thousand compounds, and can be calculated from structure, so that further ODT values on the Nagata scale can be predicted for a host of volatile or semivolatile compounds.

The hydrogen bond properties of water from 273 K to 573 K; equations for the prediction of gas-water partition coefficients.

An equation due to Plyasunov and Shock for the calculation of gas to water partition coefficients, as log K(w), at high temperature requires knowledge of experimental values of log K, ΔH the standard enthalpy of hydration at 298 K, ΔCp the heat capacity of hydration at 298 K and b' that is a constant that refers to the variation of ΔCp(w) with temperature. Linear free energy relationships based on Abraham descriptors for solutes have been constructed for all of these four input quantities, so that log K(w), ΔH, ΔCp and b' can be estimated. Known values of the four input quantities, where available, together with estimated values enabled values of log K(w) from 273 to 573 K to be calculated for 555 solutes. At various fixed temperatures LFERs again based on Abraham descriptors were constructed from the obtained log K(w) values for the 555 solutes. These LFERs can then be used to estimate further values of log K(w) between 273 and 573 K for any solute with already determined descriptors, so that knowledge of experimental values of log K, ΔH, ΔCp and b' is no longer necessary. The obtained LFERs show how the solvation properties of water vary with temperature. Both water hydrogen bond acidity and hydrogen bond basicity towards solutes decrease substantially with increase of temperature until at 573 K the hydrogen bond properties of water resemble those for solvents such as nitrobenzene and acetonitrile at 298 K. A characteristic property of water at ambient temperatures is that water will not solvate non-polar solutes, thus leading to large Henry's law constants and small values of K(w). Although this property also decreases with increase of temperature, our results indicate that it still remains to some extent even at 573 K, so that the general solvation properties of water at 573 K do not resemble the solvation properties of organic solvents at 298 K.

Descriptors for some compounds with pharmacological activity; calculation of properties.

Abraham model solute descriptors have been determined for nisoldipine, nizatidine, loratadine, zonisamide, oxaprozin, rebamipide, domperidone, temozolomide, 'florfenicol', florfenicol A, dapsone, chrysin, benorilate, ß-lapachone, and Ipriflavone based on published partition coefficients, molar solubilities and gas chromatographic retention indices. The calculated solute descriptors, combined with our previously published Abraham model correlations, are used to predict several important physicochemical and biological properties, such as air-water, air-blood, air-lung, air-fat, air-skin, water-lipid, water-membrane and water-skin partition coefficients, as well as permeation from water through skin.

Linear free energy relationship models for the retention of partially ionized acid-base compounds in reversed-phase liquid chromatography.

The LFER model of Abraham is applied to the retention of the neutral and ionic forms of 94 solutes in a C18 column and 40% v/v acetonitrile/water mobile phase. The results show that polarizability and cavity formation interactions increase retention, whereas dipole and hydrogen bonding interactions favours partition to the mobile phase and thus, they decrease retention. The coefficients of the ionic descriptors measure the effect of the electrostatic interactions and their contribution to partition of the cation or anion between the two mobile and stationary chromatographic phases. A new LFER model for application to the retention of partially dissociated acids and bases is derived averaging the descriptors of the neutral and ionic forms according to their degrees of ionization in the mobile phase. This new LFER model is satisfactorily compared to other literature modified Abraham models for a set of 498 retention data of partially dissociated acids and bases. All tested models require the calculation of the ionization degrees of the compounds at the measuring pH. Calculation of the ionization degrees in the chromatographic mobile phase (i.e. from pH and pKa in the eluent) give good correlations for all tested models. However, estimation of these ionization degrees from pH - pKa data in pure water gives biased estimations of the retention of the partially ionized solutes.

Odor detection by humans of lineal aliphatic aldehydes and helional as gauged by dose-response functions.

We have measured concentration detection (i.e., psychometric) functions to determine the odor detectability of homologous aliphatic aldehydes (propanal, butanal, hexanal, octanal, and nonanal) and helional. Subjects (16 < or = n < or = 18) used a 3-alternative forced-choice procedure against carbon-filtered air (blanks), under an ascending concentration approach. Generation, delivery, and control of each vapor were achieved via an 8-station vapor delivery device. Gas chromatography served to quantify the concentrations presented. Group and individual functions were modeled by a sigmoid (logistic) equation. Odor detection thresholds (ODTs) were defined as the concentration producing a detectability (P) halfway (P = 0.5) between chance (P = 0.0) and perfect detection (P = 1.0). ODTs decreased with carbon chain length: 2.0, 0.46, 0.33, and 0.17 ppb, respectively, from propanal to octanal, but the threshold increased for nonanal (0.53 ppb), revealing maximum sensitivity for the 8-carbon member. The strong olfactory receptor (OR) ligands octanal and helional (0.14 ppb) showed the lowest thresholds. ODTs fell at the lower end of previously reported values. Interindividual variability (ODT ratios) amounted to a factor ranging from 10 to 50, lower than typically reported, and was highest for octanal and hexanal. The behavioral dose-response functions emerge at concentrations 2-5 orders of magnitude lower than those required for functions tracing the activation of specific human ORs by the same aldehydes in cell/molecular studies, after all functions were expressed as vapor concentrations.

Solute descriptors for phenoxide anions and their use to establish correlations of rates of reaction of anions with iodomethane.

Partition coefficients from water to organic solvents for phenoxide anions have been obtained using a method based on the variation of pK(a) of phenols with solvent. From these partition coefficients, the Abraham descriptors for phenoxide anions, on the same scale as descriptors for neutral molecules, have been obtained. Equations have been constructed for the prediction of descriptors, and values of the Abraham descriptors E, S, A (equal to zero for the phenoxide anions we have studied), B, V, and J(-) have been determined or estimated for some 60 substituted phenoxide anions. These anions are characterized by very large values of the dipolarity/polarizability descriptor S and the hydrogen bond basicity descriptor B. In order to show the utility of the obtained phenoxide descriptors, we demonstrate that they can be combined with descriptors for carboxylate anions and permanent anions to yield equations for the Finkelstein reaction of anions with iodomethane in methanol, N,N-dimethylformamide, and water.

Equations for the transfer of neutral molecules and ionic species from water to organic phases.

Equations that can be used for partition coefficients of both neutral molecules and ions have been revised, and the term that is specific to ions has been re-evaluated. A new method has been devised for the determination of partition coefficients from water to organic solvents for carboxylate anions that is based on the variation of pK(a) for carboxylic acids with solvent. Using these partition coefficients, descriptors for carboxylate anions, on the same scale as descriptors for neutral molecules, have been obtained for 71 such anions. For 13 carboxylate anions in a test set, descriptors could be predicted that in turn led to predictions of 78 log P values over six water-solvent systems with an absolute error of 0.13 and a standard deviation of only 0.55 log units. Descriptors have been obtained for 26 protonated amines as a training set and descriptors predicted for 17 protonated amines as a test set. The predicted descriptors in turn led to the prediction of 18 log P values for protonated amines with an absolute error of 0.09 and a standard deviation of 0.39 log units. The carboxylate anions are the strongest monofunctional hydrogen bond bases, and the protonated amines are the strongest monofunctional hydrogen bond acids that we have studied.

Hydrogen bonding between solutes in solvents octan-1-ol and water.

The 1:1 equilibrium constants, K, for the association of hydrogen bond bases and hydrogen bond acids have been determined by using octan-1-ol solvent at 298 K for 30 acid-base combinations. The values of K are much smaller than those found for aprotic, rather nonpolar solvents. It is shown that the log K values can satisfactorily be correlated against α(H)2·ß(H)2, where α(H)2 and ß(H)2 are the 1:1 hydrogen bond acidities and basicities of solutes. The slope of the plot, 2.938, is much smaller than those for log K values in the nonpolar organic solvents previously studied. An analysis of literature data on 1:1 hydrogen bonding in water yields a negative slope for a plot of log K against α(H)2·ß(H)2, thus showing how the use of very strong hydrogen bond acids and bases does not lead to larger values of log K for 1:1 hydrogen bonding in water. It is suggested that for simple 1:1 association between monofunctional solutes in water, log K cannot be larger than about -0.1 log units. Descriptors have been obtained for the complex between 2,2,2-trifluoroethanol and propanone, and used to analyze solvent effects on the two reactants, the complex, and the complexation constant.

Structure-activity relationships on the odor detectability of homologous carboxylic acids by humans.

We measured concentration detection functions for the odor detectability of the homologs: formic, acetic, butyric, hexanoic, and octanoic acids. Subjects (14 ≤ n ≤ 18) comprised young (19-37 years), healthy, nonsmoker, and normosmic participants from both genders. Vapors were delivered by air dilution olfactometry, using a three-alternative forced-choice procedure against carbon-filtered air, and an ascending concentration approach. Delivered concentrations were established by gas chromatography (flame ionization detector) in parallel with testing. Group and individual olfactory functions were modeled by a sigmoid (logistic) equation from which two parameters are calculated: C, the odor detection threshold (ODT) and D, the steepness of the function. Thresholds declined with carbon chain length along formic, acetic, and butyric acid where they reached a minimum (ODTs = 514, 5.2, and 0.26 ppb by volume, respectively). Then, they increased for hexanoic (1.0 ppb) and octanoic (0.86 ppb) acid. Odor thresholds and interindividual differences in olfactory acuity among these young, normosmic participants were lower than traditionally thought and reported. No significant effects of gender on odor detectability were observed. The finding of an optimum molecular size for odor potency along homologs confirms a prediction made by a model of ODTs based on a solvation equation. We discuss the mechanistic implications of this model for the process of olfactory detection.

The transfer of neutral molecules, ions and ionic species from water to wet octanol.

Partition coefficients for transfer from water to wet octanol have been set out for permanent ions, and for ionic species derived from neutral molecules by loss or acceptance of a proton. A previous equation for the partition of neutral compounds from water to wet octanol has been extended to include the partition of permanent ions, and ionic species; the only additional descriptors required are J(+) for cations and J(-) for anions. Fourteen cations and twelve anions are included in the equation. It is shown that differences in partition coefficients between neutral species and the corresponding ionic species, as log P(oct), are not at all constant. For carboxylic acids and the corresponding anion, differences are about 5.0 log units, for phenols and the corresponding anion differences range from 2.2 to 4.9 log units, and for amines and the corresponding protonated cations there is an enormous range of differences from 0.6 to 7.1 log units.

Comments on "Classification of biphasic solvent systems according to Abraham descriptors for countercurrent chromatography".

Abraham model correlations reported by Marlot and coworkers for the 1-octanol/water, 1-butanol/water, ethyl acetate/water, and heptane/methanol biphasic partitioning systems are compared to previously published Abraham model correlations. The previously published correlations for the fore-mentioned partitioning systems are based on more experimental data points, and exhibit much better descriptive ability as evidenced by much smaller standard deviations/standard errors and larger squared correlation coefficients.

Descriptors for terpene esters from chromatographic and partition measurements: Estimation of human odor detection thresholds.

We have used gas chromatographic retention data together with other data to obtain Abraham descriptors for 30 terpene esters. These include the air-water partition coefficient, as log Kw, for which no experimental values are available for any terpene ester. The other descriptors are the ester dipolarity, S, the hydrogen bond basicity, B, (the ester hydrogen bond acidity is zero for the esters studied), and L the logarithm of the air-hexadecane partition coefficient. Both S and B are larger than those for simple aliphatic esters, as expected from the terpene ester structures that include ring systems and ethylenic double bonds. These descriptors can then be used to obtain a large number of physicochemical and environmental properties of terpene esters. We have analyzed experimental results on human odor detection thresholds and have constructed another equation for the calculation of these thresholds, to go with a previous equation that we have reported. Then the descriptors for terpene esters can be used to estimate the important odor detection thresholds.