Octanol/Air Partition Coefficient─A General-Purpose Fragment Model to Predict Log Koa from Molecular Structure.

The octanol/air partition coefficient Koa is important for assessing the bioconcentration of airborne xenobiotics in foliage and in air-breathing organisms. Moreover, Koa informs about compound partitioning to aerosols and indoor dust, and complements the octanol/water partition coefficient Kow and the air/water partition coefficient Kaw for multimedia fate modeling. Experimental log Koa at 25 °C has been collected from literature for 2161 compounds with molecular weights from 16 to 959 Da. The curated data set covers 18.2 log units (from -1.0 to 17.2). A newly developed fragment model for predicting log Koa from molecular structure outperforms COSMOtherm, EPI-Suite KOAWIN, OPERA, and linear solvation energy relationships (LSERs) regarding the root-mean-squared error (rms) and the maximum negative and positive errors (mne and mpe) (rms: 0.57 vs 0.86 vs 1.09 vs 1.19 vs 1.05-1.53, mne: -2.55 vs -3.95 vs -7.51 vs -7.54 vs (-5.63) - (-7.34), mpe: 2.91 vs 5.97 vs 7.54 vs 4.24 vs 6.89-10.2 log units). The prediction capability, statistical robustness, and sound mechanistic basis are demonstrated through initial separation into a training and prediction set (80:20%), mutual leave-50%-out validation, and target value scrambling in terms of temporarily wrong compound-Koa allocations. The new general-purpose model is implemented in a fully automatized form in the ChemProp software available to the public. Regarding Koa indirectly determined through Kow and Kaw, a new approach is developed to convert from wet to dry octanol, enabling higher consistency in experimental (and thus also predicted) Koa.

Henry's Law Constant─A General-Purpose Fragment Model to Predict Log Kaw from Molecular Structure.

Henry's law constant is important for assessing the environmental fate of organic compounds, including polar accumulation, indoor contamination, and the impact of airborne predominance on persistence. Moreover, it can be used in the context of alternative 3R bioassays to inform about the compound loss through volatilization as a confounding factor. For 2636 compounds, curated experimental log Kaw (air/water partition coefficient) data at 25° covering 23.6 orders of magnitude (from -18.6 to 5.0) have been collected from the literature. Subsequently, a new fragment model for predicting log Kaw from molecular structures has been developed. According to the root-mean-squared error (rms) and the maximum negative and positive errors (mne and mpe), this general-purpose model outperforms COSMOtherm, EPISuite HENRYWIN, OPERA, and LSER with calculated input parameters significantly (rms 0.50 vs 0.92 vs 1.25 vs 1.28 vs 1.38, mne -2.74 vs -6.78 vs -9.11 vs -6.24 vs -6.27, mpe 2.25 vs 6.22 vs 8.27 vs 11.5 vs 7.69 log units). Initial separation into a training and prediction set (80%:20%), mutual leave-50%-out validation, and target value scrambling (temporarily wrong compound-Kaw allocations) demonstrate the prediction capability, statistical robustness, and mechanistically sound basis of the fragment scheme. The new model is available to the public in fully computerized form through the ChemProp software, and can be combined with a separate existing model to extend the log Kaw prediction to temperatures different from 25 °C.

An Integrated Data-Driven Strategy for Safe-by-Design Nanoparticles: The FP7 MODERN Project.

The development and implementation of safe-by-design strategies is key for the safe development of future generations of nanotechnology enabled products. The safety testing of the huge variety of nanomaterials that can be synthetized is unfeasible due to time and cost constraints. Computational modeling facilitates the implementation of alternative testing strategies in a time and cost effective way. The development of predictive nanotoxicology models requires the use of high quality experimental data on the structure, physicochemical properties and bioactivity of nanomaterials. The FP7 Project MODERN has developed and evaluated the main components of a computational framework for the evaluation of the environmental and health impacts of nanoparticles. This chapter describes each of the elements of the framework including aspects related to data generation, management and integration; development of nanodescriptors; establishment of nanostructure-activity relationships; identification of nanoparticle categories; hazard ranking and risk assessment.

Organic chemicals jeopardize the health of freshwater ecosystems on the continental scale.

Organic chemicals can contribute to local and regional losses of freshwater biodiversity and ecosystem services. However, their overall relevance regarding larger spatial scales remains unknown. Here, we present, to our knowledge, the first risk assessment of organic chemicals on the continental scale comprising 4,000 European monitoring sites. Organic chemicals were likely to exert acute lethal and chronic long-term effects on sensitive fish, invertebrate, or algae species in 14% and 42% of the sites, respectively. Of the 223 chemicals monitored, pesticides, tributyltin, polycyclic aromatic hydrocarbons, and brominated flame retardants were the major contributors to the chemical risk. Their presence was related to agricultural and urban areas in the upstream catchment. The risk of potential acute lethal and chronic long-term effects increased with the number of ecotoxicologically relevant chemicals analyzed at each site. As most monitoring programs considered in this study only included a subset of these chemicals, our assessment likely underestimates the actual risk. Increasing chemical risk was associated with deterioration in the quality status of fish and invertebrate communities. Our results clearly indicate that chemical pollution is a large-scale environmental problem and requires far-reaching, holistic mitigation measures to preserve and restore ecosystem health.

Contribution of waste water treatment plants to pesticide toxicity in agriculture catchments.

Pesticide residues are frequently found in water bodies and may threaten freshwater ecosystems and biodiversity. In addition to runoff or leaching from treated agricultural fields, pesticides may enter streams via effluents from wastewater treatment plants (WWTPs). We compared the pesticide toxicity in terms of log maximum Toxic Unit (log mTU) of sampling sites in small agricultural streams of Germany with and without WWTPs in the upstream catchments. We found an approximately half log unit higher pesticide toxicity for sampling sites with WWTPs (p < 0.001). Compared to fungicides and insecticides, herbicides contributed most to the total pesticide toxicity in streams with WWTPs. A few compounds (diuron, terbuthylazin, isoproturon, terbutryn and Metazachlor) dominated the herbicide toxicity. Pesticide toxicity was not correlated with upstream distance to WWTP (Spearman's rank correlation, rho = - 0.11, p > 0.05) suggesting that other context variables are more important to explain WWTP-driven pesticide toxicity. Our results suggest that WWTPs contribute to pesticide toxicity in German streams.

Fish embryo toxicity test: identification of compounds with weak toxicity and analysis of behavioral effects to improve prediction of acute toxicity for neurotoxic compounds.

The fish embryo toxicity test has been proposed as an alternative for the acute fish toxicity test, but concerns have been raised for its predictivity given that a few compounds have been shown to exhibit a weak acute toxicity in the fish embryo. In order to better define the applicability domain and improve the predictive capacity of the fish embryo test, we performed a systematic analysis of existing fish embryo and acute fish toxicity data. A correlation analysis of a total of 153 compounds identified 28 compounds with a weaker or no toxicity in the fish embryo test. Eleven of these compounds exhibited a neurotoxic mode of action. We selected a subset of eight compounds with weaker or no embryo toxicity (cyanazine, picloram, aldicarb, azinphos-methyl, dieldrin, diquat dibromide, endosulfan, and esfenvalerate) to study toxicokinetics and a neurotoxic mode of action as potential reasons for the deviating fish embryo toxicity. Published fish embryo LC50 values were confirmed by experimental analysis of zebrafish embryo LC50 according to OECD guideline 236. Except for diquat dibromide, internal concentration analysis did not indicate a potential relation of the low sensitivity of fish embryos to a limited uptake of the compounds. Analysis of locomotor activity of diquat dibromide and the neurotoxic compounds in 98 hpf embryos (exposed for 96 h) indicated a specific effect on behavior (embryonic movement) for the neurotoxic compounds. The EC50s of behavior for neurotoxic compounds were close to the acute fish toxicity LC50. Our data provided the first evidence that the applicability domain of the fish embryo test (LC50s determination) may exclude neurotoxic compounds. However, neurotoxic compounds could be identified by changes in embryonic locomotion. Although a quantitative prediction of acute fish toxicity LC50 using behavioral assays in fish embryos may not yet be possible, the identification of neurotoxicity could trigger the conduction of a conventional fish acute toxicity test or application of assessment factors while considering the very good fish embryo-acute fish toxicity correlation for other compounds.

The OSIRIS Weight of Evidence approach: ITS mutagenicity and ITS carcinogenicity.

Risk assessment of chemicals usually implies data evaluation of in vivo tests in rodents to conclude on their hazards. The FP7 European project OSIRIS has developed integrated testing strategies (ITS) for relevant toxicological endpoints to avoid unnecessary animal testing and thus to reduce time and costs. This paper describes the implementation of ITS mutagenicity and carcinogenicity in the public OSIRIS webtool. The data requirements of REACH formed the basis for these ITS. The main goal was to implement procedures to reach a conclusion on the adequacy and validity of available data. For the mutagenicity ITS a quantitative Weight of Evidence approach based on Bayesian statistics was developed and implemented. The approach allows an overall quality assessment of all available data for the five types of mutagenicity data requirements: in vitro bacterial mutagenicity, in vitro and in vivo chromosome aberration, in vitro and in vivo mammalian mutagenicity. For the carcinogenicity ITS a tool was developed to evaluate the quality of studies not conforming (entirely) to guidelines. In a tiered approach three quality aspects are assessed: documentation (reliability), study design (adequacy) and scope of examination (validity). The quality assessment is based on expert and data driven quantitative Weight of Evidence.

The OSIRIS Weight of Evidence approach: ITS for the endpoints repeated-dose toxicity (RepDose ITS).

In the FP6 European project OSIRIS, Integrated Testing Strategies (ITSs) for relevant toxicological endpoints were developed to avoid new animal testing and thus to reduce time and costs. The present paper describes the development of an ITS for repeated-dose toxicity called RepDose ITS which evaluates the conditions under which in vivo non-guideline studies are reliable. In a tiered approach three aspects of these "non-guideline" studies are assessed: the documentation of the study (reliability), the quality of the study design (adequacy) and the scope of examination (validity). The reliability is addressed by the method "Knock-out criteria", which consists of four essential criteria for repeated-dose toxicity studies. A second tool, termed QUANTOS (Quality Assessment of Non-guideline Toxicity Studies), evaluates and weights the adequacy of the study by using intra-criterion and inter-criteria weighting. Finally, the Coverage approach calculates a probability that the detected Lowest-Observed-Effect-Level (LOEL) is similar to the LOEL of a guideline study dependent on the examined targets and organs of the non-guideline study. If the validity and adequacy of the non-guideline study are insufficient for risk assessment, the ITS proposes to apply category approach or the Threshold of Toxicological Concern (TTC) concept, and only as a last resort new animal-testing.

A comparative survey of chemistry-driven in silico methods to identify hazardous substances under REACH.

This paper presents an inventory of in silico screening tools to identify substance properties of concern under the European chemicals' legislation REACH. The objective is to support the selection and implementation of appropriate tools as building blocks within integrated testing strategies (ITS). The relevant concerns addressed are persistence, bioaccumulation potential, acute and long-term aquatic toxicity, PBT/vPvB properties ((very) persistent, (very) bioaccumulative, toxic), CMR (carcinogenicity, mutagenicity, reproductive toxicity), endocrine disruption and skin sensitisation. The inventory offers a comparative evaluation of methods with respect to the underlying algorithms (how does the method work?) and the applicability domains (when does the method work?) as well as their limitations (when does the method not work?). The inventory explicitly addresses the reliability of predictions of different in silico models for diverse chemicals by applicability domain considerations. The confidence in predictions can be greatly improved by consensus modelling that allows for taking conflicting results into account. The inventory is complemented by a brief discussion of socio-economic tools for assessing the potential efficiency gains of using in silico methods compared to traditional in vivo testing of chemical hazards.

Prediction of the dissociation constant pKa of organic acids from local molecular parameters of their electronic ground state.

A quantum chemical method has been developed to estimate the dissociation constant pK(a) of organic acids from their neutral molecular structures by employing electronic structure properties. The data set covers 219 phenols (including 29 phenols with intramolecular H-bonding), 150 aromatic carboxylic acids, 190 aliphatic carboxylic acids, and 138 alcohols, with pK(a) varying by 16 units (0.38-16.80). Optimized ground-state geometries employing the semiempirical AM1 Hamiltonian have been used to quantify the site-specific molecular readiness to donate or accept electron charge in terms of both charge-associated energies and energy-associated charges, augmented by an ortho substitution indicator for aromatic compounds. The resultant regression models yield squared correlation coefficients (r(2)) from 0.82 to 0.90 and root-mean-square errors (rms) from 0.39 to 0.70 pK(a) units, corresponding to an overall (subset-weighted) r(2) of 0.86. Simulated external validation, leave-10%-out cross-validation and target value scrambling demonstrate the statistical robustness and prediction power of the derived model suite. The low intercorrelation with prediction errors from the commercial ACD package provides opportunity for a consensus model approach, offering a pragmatic way for further increasing the confidence in prediction significantly. Interestingly, inclusion of calculated free energies of aqueous solvation does not improve the prediction performance, probably because of the limited precision provided by available continuum-solvation models.

Quantitative read-across for predicting the acute fish toxicity of organic compounds.

Read-across enables the interpolation of a property for a target chemical from respective experimental data of sufficiently similar compounds. Employing a set of 692 organic compounds with experimental values for the 96 h fish toxicity toward the fathead minnow in terms of LC(50) (lethal concentration 50%) values, a read-across method has been developed that is based on atom-centered fragments (ACFs) for evaluating chemical similarity. Prediction of log LC(50) proceeds through reading across the toxicity enhancement over predicted narcosis-level toxicity in terms of the respective logarithmic ratio, log T(e), and adding the respective baseline narcosis LC(50) estimated from log K(ow) (octanol/water partition coefficient). Depending on the minimum similarity imposed on a compound to serve as read-across basis for the target chemical, three different standard settings have been introduced, allowing one to perform screening-level estimations as well as predictions with intermediate and good confidence. The respective squared correlation coefficients (r(2)) are 0.73, 0.78, and 0.87, with root-mean square errors (rms) of 0.73, 0.60, and 0.39 log units, respectively. As a general trend, increasing the ACF minimum similarity increases the prediction quality at the cost of decreasing the application range. The method has the potential to assist in the predictive evaluation of fish toxicity for regulatory purposes such as under the REACH legislation.

Occurrence and toxicity of 331 organic pollutants in large rivers of north Germany over a decade (1994 to 2004).

We analyzed the detection frequencies and concentrations for 331 organic compounds measured between 1994 and 2004 in the four largest rivers of north Germany, the Elbe, Weser, Aller, and Ems Rivers, and we assessed the potential risk for aquatic fauna using experimental and predicted acute toxicity data for the green alga Pseudokirchneriella subcapitata, the crustacean Daphnia magna, and the fish Pimephales promelas. The detection frequency for most compounds decreased significantly from 1994 to 2004. Polycyclic aromatic hydrocarbons (PAHs) were most frequently detected, while pesticides were the most important chemical group concerning toxicity for the standard test organisms. The predicted toxicity for D. magna was significantly higher than for the other organisms and reached levels envisaging acute toxic effects on the invertebrate fauna, still in 2004. Most of the compounds responsible for potential acute effects on aquatic organisms are currently not considered as priority substances in the European Union, while only 2 of 25 priority substances that have been measured occurred at levels that may be relevant in terms of toxicity for the selected test organisms. We conclude that attenuation of pesticides and other organic toxicants should play an increased role in river basin management.

Effects of pesticides monitored with three sampling methods in 24 sites on macroinvertebrates and microorganisms.

Grab water samples, sediment samples, and 2,2,4-trimethylpentane passive samplers (TRIMPS) were used to determine the exposure to 97 pesticides in 24 southeast Australian stream sites over 5 months. Macroinvertebrate communities and selected microorganisms (bacteria, flagellates, ciliates, amoebas, nematodes, and gastrotrichs) were sampled to detect relationships with pesticide toxicity. Sediment samples had the highest estimated toxicities in terms of toxic units (TU) for Daphnia magna (TUDM) and for Selenastrum capricornutum (TUSC). The pesticide-selective SPEARpesticides and the general SIGNAL index for macroinvertebrates exhibited negative linear relationships (r(2) = 0.67 and 0.36, respectively) with pesticide contamination in terms of log maximum TUDM (log mTUDM), suggesting macroinvertebrate community change due to pesticide exposure. Pesticide contamination was the only measured variable explaining variation in ecological quality. Variation in the densities of several microbial groups was best explained by environmental variables other than log TUs. The log mTUDM values derived from sediment concentrations were most important to establish a link with effects on macroinvertebrates, whereas log mTUDM of grab water samples had only minor contribution. Current-use insecticides and fungicides can affect macroinvertebrate communities and monitoring of sediment and continuous water sampling is needed to detect these effects.

Comparative analysis of QSAR models for predicting pK(a) of organic oxygen acids and nitrogen bases from molecular structure.

For 1143 organic compounds comprising 580 oxygen acids and 563 nitrogen bases that cover more than 17 orders of experimental pK(a) (from -5.00 to 12.23), the pK(a) prediction performances of ACD, SPARC, and two calibrations of a semiempirical quantum chemical (QC) AM1 approach have been analyzed. The overall root-mean-square errors (rms) for the acids are 0.41, 0.58 (0.42 without ortho-substituted phenols with intramolecular H-bonding), and 0.55 and for the bases are 0.65, 0.70, 1.17, and 1.27 for ACD, SPARC, and both QC methods, respectively. Method-specific performances are discussed in detail for six acid subsets (phenols and aromatic and aliphatic carboxylic acids with different substitution patterns) and nine base subsets (anilines, primary, secondary and tertiary amines, meta/para-substituted and ortho-substituted pyridines, pyrimidines, imidazoles, and quinolines). The results demonstrate an overall better performance for acids than for bases but also a substantial variation across subsets. For the overall best-performing ACD, rms ranges from 0.12 to 1.11 and 0.40 to 1.21 pK(a) units for the acid and base subsets, respectively. With regard to the squared correlation coefficient r², the results are 0.86 to 0.96 (acids) and 0.79 to 0.95 (bases) for ACD, 0.77 to 0.95 (acids) and 0.85 to 0.97 (bases) for SPARC, and 0.64 to 0.87 (acids) and 0.43 to 0.83 (bases) for the QC methods, respectively. Attention is paid to structural and method-specific causes for observed pitfalls. The significant subset dependence of the prediction performances suggests a consensus modeling approach.

Tautomer identification and tautomer structure generation based on the InChI code.

An algorithm is introduced that enables a fast generation of all possible prototropic tautomers resulting from the mobile H atoms and associated heteroatoms as defined in the InChI code. The InChI-derived set of possible tautomers comprises (1,3)-shifts for open-chain molecules and (1,n)-shifts (with n being an odd number >3) for ring systems. In addition, our algorithm includes also, as extension to the InChI scope, those larger (1,n)-shifts that can be constructed from joining separate but conjugated InChI sequences of tautomer-active heteroatoms. The developed algorithm is described in detail, with all major steps illustrated through explicit examples. Application to approximately 72,500 organic compounds taken from EINECS (European Inventory of Existing Commercial Chemical Substances) shows that around 11% of the substances occur in different heteroatom-prototropic tautomeric forms. Additional QSAR (quantitative structure-activity relationship) predictions of their soil sorption coefficient and water solubility reveal variations across tautomers up to more than two and 4 orders of magnitude, respectively. For a small subset of nine compounds, analysis of quantum chemically predicted tautomer energies supports the view that among all tautomers of a given compound, those restricted to H atom exchanges between heteroatoms usually include the thermodynamically most stable structures.

Quantitative and qualitative models for carcinogenicity prediction for non-congeneric chemicals using CP ANN method for regulatory uses.

The new European chemicals regulation Registration, Evaluation, Authorization and Restriction of Chemicals entered into force in June 2007 and accelerated the development of quantitative structure-activity relationship (QSAR) models for a variety of endpoints, including carcinogenicity. Here, we would like to present quantitative (continuous) and qualitative (categorical) models for non-congeneric chemicals for prediction of carcinogenic potency. A dataset of 805 substances was obtained after a preliminary screening of findings of rodent carcinogenicity for 1,481 chemicals accessible via Distributed Structure-Searchable Toxicity (DSSTox) Public Database Network originated from the Lois Gold Carcinogenic Potency Database (CPDB). Twenty seven two-dimensional MDL descriptors were selected using Kohonen mapping and principal component analysis. The counter propagation artificial neural network (CP ANN) technique was applied. Quantitative models were developed exploring the relationship between the experimental and predicted carcinogenic potency expressed as a tumorgenic dose TD(50) for rats. The obtained models showed low prediction power with correlation coefficient less than 0.5 for the test set. In the next step, qualitative models were developed. We found that the qualitative models exhibit good accuracy for the training set (92%). The model demonstrated good predicted performance for the test set. It was obtained accuracy (68%), sensitivity (73%), and specificity (63%). We believe that CP ANN method is a good in silico approach for modeling and predicting rodent carcinogenicity for non-congeneric chemicals and may find application for other toxicological endpoints.

Computational material flow analysis for thousands of chemicals of emerging concern in European waters.

Knowledge of exposure to a wide range of chemicals, and the spatio-temporal variability thereof, is urgently needed in the context of protecting and restoring aquatic ecosystems. This paper discusses a computational material flow analysis to predict the occurrence of thousands of man-made organic chemicals on a European scale, based on a novel temporally and spatially resolved modelling framework. The goal was to increase understanding of pressures by emerging chemicals and to complement surface water monitoring data. The ambition was to provide a first step towards a "real-life" mixture exposure situation accounting for as many chemicals as possible. Comparison of simulated concentrations and chemical monitoring data for 226 substance/basin combinations showed that the simulated concentrations were accurate on average. For 65% and 90% of substance/basin combinations the error was within one and two orders of magnitude respectively. An analysis of the relative importance of uncertainties revealed that inaccuracies in use volume or use type information contributed most to the error for individual substances. To resolve this, we suggest better registration of use types of industrial chemicals, investigation of presence/absence of industrial chemicals in wastewater and runoff samples and more scientific information exchange.

Modeling the H bond donor strength of -OH, -NH, and -CH sites by local molecular parameters.

A quantum chemical model is introduced to predict the H-bond donor strength of monofunctional organic compounds from their ground-state electronic properties. The model covers -OH, -NH, and -CH as H-bond donor sites and was calibrated with experimental values for the Abraham H-bond donor strength parameter A using the ab initio and density functional theory levels HF/6-31G** and B3LYP/6-31G**. Starting with the Morokuma analysis of hydrogen bonding, the electrostatic (ES), polarizability (PL), and charge transfer (CT) components were quantified employing local molecular parameters. With hydrogen net atomic charges calculated from both natural population analysis and the ES potential scheme, the ES term turned out to provide only marginal contributions to the Abraham parameter A, except for weak hydrogen bonds associated with acidic -CH sites. Accordingly, A is governed by PL and CT contributions. The PL component was characterized through a new measure of the local molecular hardness at hydrogen, eta(H), which in turn was quantified through empirically defined site-specific effective donor and acceptor energies, EE(occ) and EE(vac). The latter parameter was also used to address the CT contribution to A. With an initial training set of 77 compounds, HF/6-31G** yielded a squared correlation coefficient, r(2), of 0.91. Essentially identical statistics were achieved for a separate test set of 429 compounds and for the recalibrated model when using all 506 compounds. B3LYP/6-31G** yielded slightly inferior statistics. The discussion includes subset statistics for compounds containing -OH, -NH, and active -CH sites and a nonlinear model extension with slightly improved statistics (r(2) = 0.92).

Chemical domain of QSAR models from atom-centered fragments.

A methodology to characterize the chemical domain of qualitative and quantitative structure-activity relationship (QSAR) models based on the atom-centered fragment (ACF) approach is introduced. ACFs decompose the molecule into structural pieces, with each non-hydrogen atom of the molecule acting as an ACF center. ACFs vary with respect to their size in terms of the path length covered in each bonding direction starting from a given central atom and how comprehensively the neighbor atoms (including hydrogen) are described in terms of element type and bonding environment. In addition to these different levels of ACF definitions, the ACF match mode as degree of strictness of the ACF comparison between a test compound and a given ACF pool (such as from a training set) has to be specified. Analyses of the prediction statistics of three QSAR models with their training sets as well as with external test sets and associated subsets demonstrate a clear relationship between the prediction performance and the levels of ACF definition and match mode. The findings suggest that second-order ACFs combined with a borderline match mode may serve as a generic and at the same time a mechanistically sound tool to define and evaluate the chemical domain of QSAR models. Moreover, four standard categories of the ACF-based membership to a given chemical domain (outside, borderline outside, borderline inside, inside) are introduced that provide more specific information about the expected QSAR prediction performance. As such, the ACF-based characterization of the chemical domain appears to be particularly useful for QSAR applications in the context of REACH and other regulatory schemes addressing the safety evaluation of chemical compounds.

Prediction of the intrinsic hydrogen bond acceptor strength of chemical substances from molecular structure.

Hydrogen bonding affects the partitioning of organic compounds between environmental and biological compartments as well as the three-dimensional shape of macromolecules. Using the semiempirical quantum chemical AM1 level of calculation, we have developed a model to predict the site-specific hydrogen bond (HB) acceptor strength from ground-state properties of the individual compounds. At present, the model parametrization is confined to compounds with one HB acceptor site of the following atom types: N, O, S, F, Cl, and Br that act as lone-pair HB acceptors, and pi-electron (aromatic or conjugated) systems with the associated C atoms as particularly weak HB acceptors. The HB acceptor strength is expressed in terms of the Abraham parameter B and calculated from local molecular parameters, taking into account electrostatic, polarizability, and charge transfer contributions according to the Morokuma concept. For a data set of 383 compounds, the squared correlation coefficient r2 is 0.97 when electrostatic potential (ESP) derived net atomic charges are employed, and the root-mean-square (rms) error is 0.04 that is in the range of experimental uncertainty. The model is validated using an extended leave-50%-out approach, and its performance is comparatively analyzed with the ones of earlier introduced ab initio (HF/6-31G**) and density functional theory (B3LYP/6-31G**) models as well as of two increment methods with respect to the total compound set as well as HB acceptor type subsets. The discussion includes an explorative model application to amides and organophosphates that demonstrates the robustness of the approach, and further opportunities for model extensions.