Identifying uncertainty in physical-chemical property estimation with IFSQSAR.

This study describes the development and evaluation of six new models for predicting physical-chemical (PC) properties that are highly relevant for chemical hazard, exposure, and risk estimation: solubility (in water SW and octanol SO), vapor pressure (VP), and the octanol-water (KOW), octanol-air (KOA), and air-water (KAW) partition ratios. The models are implemented in the Iterative Fragment Selection Quantitative Structure-Activity Relationship (IFSQSAR) python package, Version 1.1.0. These models are implemented as Poly-Parameter Linear Free Energy Relationship (PPLFER) equations which combine experimentally calibrated system parameters and solute descriptors predicted with QSPRs. Two other ancillary models have been developed and implemented, a QSPR for Molar Volume (MV) and a classifier for the physical state of chemicals at room temperature. The IFSQSAR methods for characterizing applicability domain (AD) and calculating uncertainty estimates expressed as 95% prediction intervals (PI) for predicted properties are described and tested on 9,000 measured partition ratios and 4,000 VP and SW values. The measured data are external to IFSQSAR training and validation datasets and are used to assess the predictivity of the models for "novel chemicals" in an unbiased manner. The 95% PI intervals calculated from validation datasets for partition ratios needed to be scaled by a factor of 1.25 to capture 95% of the external data. Predictions for VP and SW are more uncertain, primarily due to the challenges in differentiating their physical state (i.e., liquids or solids) at room temperature. The prediction accuracy of the models for log KOW, log KAW and log KOA of novel, data-poor chemicals is estimated to be in the range of 0.7 to 1.4 root mean squared error of prediction (RMSEP), with RMSEP in the range 1.7-1.8 for log VP and log SW. Scientific contributionNew partitioning models integrate empirical PPLFER equations and QSARs, allowing for seamless integration of experimental data and model predictions. This work tests the real predictivity of the models for novel chemicals which are not in the model training or external validation datasets.

Modeling sorption of environmental organic chemicals from water to soils.

Reliable estimation of chemical sorption from water to solid phases is an essential prerequisite for reasonable assessments of chemical hazards and risks. However, current fate and exposure models mostly rely on algorithms that lack the capability to quantify chemical sorption resulting from interactions with multiple soil constituents, including amorphous organic matter, carbonaceous organic matter, and mineral matter. Here, we introduce a novel, generic approach that explicitly combines the gravimetric composition of various solid constituents and poly-parameter linear free energy relationships to calculate the solid-water sorption coefficient (Kd) for non-ionizable or predominantly neutral organic chemicals with diverse properties in a neutral environment. Our approach demonstrates an overall statistical uncertainty of approximately 0.9 log units associated with predictions for different types of soil. By applying this approach to estimate the sorption of 70 diverse chemicals from water to two types of soils, we uncover that different chemicals predominantly exhibit sorption onto different soil constituents. Moreover, we provide mechanistic insights into the limitation of relying solely on organic carbon normalized sorption coefficient (KOC) in chemical hazard assessment, as the measured KOC can vary significantly across different soil types, and therefore, a universal cut-off threshold may not be appropriate. This research highlights the importance of considering chemical properties and multiple solid constituents in sorption modeling and offers a valuable theoretical approach for improved chemical hazard and exposure assessments.

Developing an internal threshold of toxicological concern (iTTC).

BACKGROUND: Threshold of Toxicological Concern (TTC) approaches are used for chemical safety assessment and risk-based priority setting for data poor chemicals. TTCs are derived from in vivo No Observed Effect Level (NOEL) datasets involving an external administered dose from a single exposure route, e.g., oral intake rate. Thus, a route-specific TTC can only be compared to a route-specific exposure estimate and such TTCs cannot be used for other exposure scenarios such as aggregate exposures. OBJECTIVE: Develop and apply a method for deriving internal TTCs (iTTCs) that can be used in chemical assessments for multiple route-specific exposures (e.g., oral, inhalation or dermal) or aggregate exposures. METHODS: Chemical-specific toxicokinetics (TK) data and models are applied to calculate internal concentrations (whole-body and blood) from the reported administered oral dose NOELs used to derive the Munro TTCs. The new iTTCs are calculated from the 5th percentile of cumulative distributions of internal NOELs and the commonly applied uncertainty factor of 100 to extrapolate animal testing data for applications in human health assessment. RESULTS: The new iTTCs for whole-body and blood are 0.5 nmol/kg and 0.1 nmol/L, respectively. Because the iTTCs are expressed on a molar basis they are readily converted to chemical mass iTTCs using the molar mass of the chemical of interest. For example, the median molar mass in the dataset is 220 g/mol corresponding to an iTTC of 22 ng/L-blood (22 pg/mL-blood). The iTTCs are considered broadly applicable for many organic chemicals except those that are genotoxic or acetylcholinesterase inhibitors. The new iTTCs can be compared with measured or estimated whole-body or blood exposure concentrations for chemical safety screening and priority-setting. SIGNIFICANCE: Existing Threshold of Toxicological Concern (TTC) approaches are limited in their applications for route-specific exposure scenarios only and are not suitable for chemical risk and safety assessments under conditions of aggregate exposure. New internal Threshold of Toxicological Concern (iTTC) values are developed to address data gaps in chemical safety estimation for multi-route and aggregate exposures.

Application of an Iterative Fragment Selection (IFS) Method to Estimate Entropies of Fusion and Melting Points of Organic Chemicals.

The main objective of this study is to develop and evaluate novel Quantitative Structure-Property Relationships (QSPRs) for predicting entropy of fusion (ΔSM ) and melting point (TM ) of organic chemicals from chemical structure. The QSPRs are developed using the Iterative Fragment Selection (IFS) method that requires only 2D structural information from the user (SMILES codes) for property prediction. The QSPRs also provide information on the applicability domain for each calculation and uncertainty estimates for the predictions. The root mean square error (RMSE) for the external validation sets are 11.8 J mol-1 K-1 and 46.9 K for the ΔSM and TM QSPRs, respectively. The performance of the new QSPRs is comparable to other predictive methods but has advantages with respect to availability and ease of use as well as the guidance on applicability domain for each prediction. Limitations of the new QSPRs are discussed. The QSPRs are coded as a user-friendly, freely available tool.

A toxicokinetic model for fish including multiphase sorption features.

In the scientific field of physiologically based toxicokinetic modeling the complexity of the model used depends on the complexity of the problem to be handled, leading to a broad range of existing models from simple 1-box models to complex multicompartment models. Most of these models work with lumped parameters, for example, an uptake efficiency parameter that can only be obtained with a fit of experimental data. The authors' goal was a model that is completely based on well-defined physiological and physicochemical parameters. Lumped parameters fitted on training data sets would limit the model's applicability. This would enable a new view on process understanding for uptake, distribution, and elimination procedures. Eventual goals are a better localization of chemicals within the organism itself, and to set the stage for future extensions toward ionic compounds and active transport across membranes. The model evaluation reported in the present study has shown that uptake, clearance, and bioaccumulation data for nonpolar chemicals are well predicted. Environ Toxicol Chem 2017;36:1538-1546. © 2016 SETAC.

Dermal permeation data and models for the prioritization and screening-level exposure assessment of organic chemicals.

High-throughput screening (HTS) models are being developed and applied to prioritize chemicals for more comprehensive exposure and risk assessment. Dermal pathways are possible exposure routes to humans for thousands of chemicals found in personal care products and the indoor environment. HTS exposure models rely on skin permeability coefficient (KP; cm/h) models for exposure predictions. An initial database of approximately 1000 entries for empirically-based KP data was compiled from the literature and a subset of 480 data points for 245 organic chemicals derived from testing with human skin only and using only water as a vehicle was selected. The selected dataset includes chemicals with log octanol-water partition coefficients (KOW) ranging from -6.8 to 7.6 (median=1.8; 95% of the data range from -2.5 to 4.6) and molecular weight (MW) ranging from 18 to 765g/mol (median=180); only 3% >500g/mol. Approximately 53% of the chemicals in the database have functional groups which are ionizable in the pH range of 6 to 7.4, with 31% being appreciably ionized. The compiled log KP values ranged from -5.8 to 0.1cm/h (median=-2.6). The selected subset of the KP data was then used to evaluate eight representative KP models that can be readily applied for HTS assessments, i.e., parameterized with KOW and MW. The analysis indicates that a version of the SKINPERM model performs the best against the selected dataset. Comparisons of representative KP models against model input parameter property ranges (sensitivity analysis) and against chemical datasets requiring human health assessment were conducted to identify regions of chemical properties that should be tested to address uncertainty in KP models and HTS exposure assessments.

Using model-based screening to help discover unknown environmental contaminants.

Of the tens of thousands of chemicals in use, only a small fraction have been analyzed in environmental samples. To effectively identify environmental contaminants, methods to prioritize chemicals for analytical method development are required. We used a high-throughput model of chemical emissions, fate, and bioaccumulation to identify chemicals likely to have high concentrations in specific environmental media, and we prioritized these for target analysis. This model-based screening was applied to 215 organosilicon chemicals culled from industrial chemical production statistics. The model-based screening prioritized several recognized organosilicon contaminants and generated hypotheses leading to the selection of three chemicals that have not previously been identified as potential environmental contaminants for target analysis. Trace analytical methods were developed, and the chemicals were analyzed in air, sewage sludge, and sediment. All three substances were found to be environmental contaminants. Phenyl-tris(trimethylsiloxy)silane was present in all samples analyzed, with concentrations of ∼50 pg m(-3) in Stockholm air and ∼0.5 ng g(-1) dw in sediment from the Stockholm archipelago. Tris(trifluoropropyl)trimethyl-cyclotrisiloxane and tetrakis(trifluoropropyl)tetramethyl-cyclotetrasiloxane were found in sediments from Lake Mjøsa at ∼1 ng g(-1) dw. The discovery of three novel environmental contaminants shows that models can be useful for prioritizing chemicals for exploratory assessment.

Estimating screening-level organic chemical half-lives in humans.

Relatively few measured data are available for the thousands of chemicals requiring hazard and risk assessment. The whole body, total elimination half-life (HLT) and the whole body, primary biotransformation half-life (HLB) are key parameters determining the extent of bioaccumulation, biological concentration, and risk from chemical exposure. A one-compartment pharmacokinetic (1-CoPK) mass balance model was developed to estimate organic chemical HLB from measured HLT data in mammals. Approximately 1900 HLs for human adults were collected and reviewed and the 1-CoPK model was parametrized for an adult human to calculate HLB from HLT. Measured renal clearance and whole body total clearance data for 306 chemicals were used to calculate empirical HLB,emp. The HLB,emp values and other measured data were used to corroborate the 1-CoPK HLB model calculations. HLs span approximately 7.5 orders of magnitude from 0.05 h for nitroglycerin to 2 × 10(6) h for 2,3,4,5,2',3',5',6'-octachlorobiphenyl with a median of 7.6 h. The automated Iterative Fragment Selection (IFS) method was applied to develop and evaluate various quantitative structure-activity relationships (QSARs) to predict HLT and HLB from chemical structure and two novel QSARs are detailed. The HLT and HLB QSARs show similar statistical performance; that is, r(2) = 0.89, r(2-ext) = 0.72 and 0.73 for training and external validation sets, respectively, and root-mean-square errors for the validation data sets are 0.70 and 0.75, respectively.

Effect of wind on the chemical uptake kinetics of a passive air sampler.

Passive air samplers (PASs) operate in different types of environment under various wind conditions, which may affect sampling rates and thus introduce uncertainty to PAS-derived air concentrations. To quantify the effect of wind speed and angle on the uptake in cylindrical PASs using XAD-resin as the sampling medium, we measured the uptake kinetics of polychlorinated biphenyls (PCBs) in XAD and of water in silica-gel, both under quasi wind-still condition and with lab-generated wind blowing toward the PASs at various speeds and angles. Passive sampling rates (PSRs) of PCBs under laboratory generated windy conditions were approximately 3-4 times higher than under wind-still indoor conditions. The rate of water uptake by silica-gel increased with wind speed, following a logarithmic function so that PSRs are more strongly influenced at lower wind speed. PSRs of both PCBs and water varied little with wind angle, which is consistent with computational fluid dynamic simulations showing that different angles of wind incidence cause only minor variations of air velocities within the cylindrical sampler housing. Because modifications of the design of the cylindrical PAS were not successful in eliminating the wind speed dependence of PSRs at low wind levels, indoor and outdoor deployments require different sets of PSRs. The effect of wind speed and angle on the PSRs of the cylindrical PAS are much smaller than what has been reported for the double-bowl polyurethane foam PAS. PSRs of the cylindrical XAD-PAS therefore tend to vary much less between sampling sites exposed to different wind conditions.

General model for estimating partition coefficients to organisms and their tissues using the biological compositions and polyparameter linear free energy relationships.

Equilibrium partition coefficients of organic chemicals from water to an organism or its tissues are typically estimated by using the total lipid content in combination with the octanol-water partition coefficient (K(ow)). This estimation method can cause systematic errors if (1) different lipid types have different sorptive capacities, (2) nonlipid components such as proteins have a significant contribution, and/or (3) K(ow) is not a suitable descriptor. As an alternative, this study proposes a more general model that uses detailed organism and tissue compositions (i.e., contents of storage lipid, membrane lipid, albumin, other proteins, and water) and polyparameter linear free energy relationships (PP-LFERs). The values calculated by the established PP-LFER-composition-based model agree well with experimental in vitro partition coefficients and in vivo steady-state concentration ratios from the literature with a root mean squared error of 0.32-0.53 log units, without any additional fitting. This model estimates a high contribution of the protein fraction to the overall tissue sorptive capacity in lean tissues (e.g., muscle), in particular for H-bond donor polar compounds. Direct model comparison revealed that the simple lipid-octanol model still calculates many tissue-water partition coefficients within 1 log unit of those calculated by the PP-LFER-composition-based model. Thus, the lipid-octanol model can be used as an order-of-magnitude approximation, for example, for multimedia fate modeling, but may not be suitable for more accurate predictions. Storage lipid-rich phases (e.g., adipose, milk) are prone to particularly large systematic errors. The new model provides useful implications for validity of lipid-normalization of concentrations in organisms, interpretation of biomonitoring results, and assessment of toxicity.

Elimination half-life as a metric for the bioaccumulation potential of chemicals in aquatic and terrestrial food chains.

The assessment of chemicals as bioaccumulative in the regulatory process makes use of the bioconcentration factor as a metric. However, this metric does not account for the dietary uptake route and therefore cannot be applied to terrestrial food chains. In recent years, the biomagnification factor (BMF) and the trophic magnification factor (TMF) have been suggested as standard metrics for bioaccumulation. For regulatory purposes, though, the BMF and the TMF also suffer from a number of shortcomings. They are not applicable to assess uptake routes other than the diet (e.g., dermal uptake, as is important for personal care products). When measured in the field, they depend largely on biological and ecological factors and less so on the chemical's properties, and they are difficult to normalize and standardize. In the present study, the authors suggest the elimination half-life (EL0.5 ) of a chemical as an alternative metric for bioaccumulation. The EL0.5 is equivalent to the depuration rate constant (k2 ) that is measured in various bioaccumulation and bioconcentration tests. This metric can be applied to air- and water-breathing animals, and it is valuable for all uptake routes. It has a number of practical advantages over the BMF and the TMF. In combination with a standard uptake scenario, the EL0.5 can also be linked directly to a BMF threshold of unity. Thus, the EL0.5 as a bioaccumulation metric overcomes the shortcomings of the BMF and the TMF while still conserving the advantages of the latter metrics.

Exploring the role of shelf sediments in the Arctic Ocean in determining the Arctic contamination potential of neutral organic contaminants.

The main objective of this study was to model the contribution of shelf sediments in the Arctic Ocean to the total mass of neutral organic contaminants accumulated in the Arctic environment using a standardized emission scenario for sets of hypothetical chemicals and realistic emission estimates (1930-2100) for polychlorinated biphenyl congener 153 (PCB-153). Shelf sediments in the Arctic Ocean are shown to be important reservoirs for neutral organic chemicals across a wide range of partitioning properties, increasing the total mass in the surface compartments of the Arctic environment by up to 3.5-fold compared to simulations excluding this compartment. The relative change in total mass for hydrophobic organic chemicals with log air-water partition coefficients ≥0 was greater than for chemicals with properties similar to typical POPs. The long-term simulation of PCB-153 generated modeled concentrations in shelf sediments in reasonable agreement with available monitoring data and illustrate that the relative importance of shelf sediments in the Arctic Ocean for influencing surface ocean concentrations (and therefore exposure via the pelagic food web) is most pronounced once primary emissions are exhausted and secondary sources dominate. Additional monitoring and modeling work to better characterize the role of shelf sediments for contaminant fate is recommended.

Prioritizing chemicals and data requirements for screening-level exposure and risk assessment.

BACKGROUND: Scientists and regulatory agencies strive to identify chemicals that may cause harmful effects to humans and the environment; however, prioritization is challenging because of the large number of chemicals requiring evaluation and limited data and resources. OBJECTIVES: We aimed to prioritize chemicals for exposure and exposure potential and obtain a quantitative perspective on research needs to better address uncertainty in screening assessments. METHODS: We used a multimedia mass balance model to prioritize > 12,000 organic chemicals using four far-field human exposure metrics. The propagation of variance (uncertainty) in key chemical information used as model input for calculating exposure metrics was quantified. RESULTS: Modeled human concentrations and intake rates span approximately 17 and 15 orders of magnitude, respectively. Estimates of exposure potential using human concentrations and a unit emission rate span approximately 13 orders of magnitude, and intake fractions span 7 orders of magnitude. The actual chemical emission rate contributes the greatest variance (uncertainty) in exposure estimates. The human biotransformation half-life is the second greatest source of uncertainty in estimated concentrations. In general, biotransformation and biodegradation half-lives are greater sources of uncertainty in modeled exposure and exposure potential than chemical partition coefficients. CONCLUSIONS: Mechanistic exposure modeling is suitable for screening and prioritizing large numbers of chemicals. By including uncertainty analysis and uncertainty in chemical information in the exposure estimates, these methods can help identify and address the important sources of uncertainty in human exposure and risk assessment in a systematic manner.

Iterative fragment selection: a group contribution approach to predicting fish biotransformation half-lives.

There are regulatory needs to evaluate thousands of chemicals for potential hazard and risk with limited available information. An automated method is presented for developing and evaluating Quantitative Structure-Activity Relationships (QSARs) for a range of chemical properties that can be applied for screening level chemical assessments. The method is an integrated algorithm for descriptor generation, data set splitting, cross validation, and model selection. Resulting QSARs are two-dimensional (2D) fragment-based group contribution models. The QSAR development and evaluation method does not require previous expert knowledge for selecting 2D fragments associated with the chemical property of interest. The method includes information on the domain of applicability (structural similarity to the training set) and estimates of the uncertainty in the QSAR predictions. As a demonstration, the method is applied to generate novel QSARs for fish primary biotransformation half-lives (HL(N)). Results from the new HL(N) QSARs are compared to another 2D fragment-based HL(N) QSAR developed with expert judgment, and the predictive powers of the models are found to be similar. The relative merits and limitations of each method are investigated and the new QSAR is found to make comparable predictions with significantly fewer fragments. A coefficient of determination (R(2)) of 0.789 and a root mean squared error (RMSE) of 0.526 were obtained for the training data set and an R(2) of 0.748 and an RMSE of 0.584 were obtained for the validation data set, along with a concordance correlation coefficient (CCC) of 0.857 showing good predictive power.

Screening organic chemicals in commerce for emissions in the context of environmental and human exposure.

Quantitative knowledge of organic chemical release into the environment is essential to understand and predict human exposure as well as to develop rational control strategies for any substances of concern. While significant efforts have been invested to characterize and screen organic chemicals for hazardous properties, relatively less effort has been directed toward estimating emissions and hence also risks. Here, a rapid throughput method to estimate emissions of discrete organic chemicals in commerce has been developed, applied and evaluated to support screening studies aimed at ranking and identifying chemicals of potential concern. The method builds upon information in the European Union Technical Guidance Document and utilizes information on quantities in commerce (production and/or import rates), chemical function (use patterns) and physical-chemical properties to estimate emissions to air, soil and water within the OECD for five stages of the chemical life-cycle. The method is applied to 16,029 discrete substances (identified by CAS numbers) from five national and international high production volume lists. As access to consistent input data remains fragmented or even impossible, particular attention is given to estimating, evaluating and discussing uncertainties in the resulting emission scenarios. The uncertainty for individual substances typically spans 3 to 4 orders of magnitude for this initial tier screening method. Information on uncertainties in emissions is useful as any screening or categorization methods which solely rely on threshold values are at risk of leading to a significant number of either false positives or false negatives. A limited evaluation of the screening method's estimates for a sub-set of about 100 substances, compared against independent and more detailed emission scenarios presented in various European Risk Assessment Reports, highlights that up-to-date and accurate information on quantities in commerce as well as a detailed breakdown on chemical function are critically needed for developing more realistic emission scenarios.

Susceptibility of human populations to environmental exposure to organic contaminants.

Environmental exposure to organic contaminants is a complex function of environmental conditions, food chain characteristics, and chemical properties. In this study the susceptibility of various human populations to environmental exposure to neutral organic contaminants was compared. An environmental fate model and a linked bioaccumulation model were parametrized to describe ecosystems in different climatic regions (temperate, arctic, tropical, and steppe). The human body burden resulting from constant emissions of hypothetical chemicals was estimated for each region. An exposure susceptibility index was defined as the body burden in the region of interest normalized to the burden of the same chemical in a reference human from the temperate region eating an average diet. For most persistent chemicals emitted to air, the Arctic had the highest susceptibility index (max 520). Susceptibility to exposure was largely determined by the food web properties. The properties of the physical environment only had a marked effect when air or water, not food, was the dominant source of human exposure. Shifting the mode of emission markedly changed the relative susceptibility of the ecosystems in some cases. The exposure arising from chemical use clearly varies between ecosystems, which makes an understanding of ecosystem susceptibility to exposure important for chemicals management.

Assessment of chemical screening outcomes based on different partitioning property estimation methods.

Screening is widely used to prioritize chemicals according to their potential environmental hazard, as expressed in the attributes of persistence, bioaccumulation (B), toxicity and long range transport potential (LRTP). Many screening approaches for B and LRTP rely on the categorization of chemicals based on a comparison of their equilibrium partition coefficients between octanol and water (K(OW)), air and water (K(AW)) and octanol and air (K(OA)) with a threshold value. As experimental values of the properties are mostly unavailable for the large number of chemicals being screened, the use of quantitative structure-property relationships (QSPRs) and other computational chemistry methods becomes indispensable. Predictions by different methods often deviate considerably, and flawed predictions may lead to false positive/negative categorizations. We predicted the partitioning properties of 529 chemicals, culled from previous prioritization efforts, using the four prediction methods EPI Suite, SPARC, COSMOtherm, and ABSOLV. The four sets of predictions were used to screen the chemicals against various LRTP and B criteria. Screening results based on the four methods were consistent for only approximately 70% of the chemicals. To further assess whether the means of estimating environmental phase partitioning has an impact, a subset of 110 chemicals was screened for elevated arctic contamination potential based on single-parameter and poly-parameter linear free energy relationships respectively. Different categorizations were observed for 5 out of 110 chemicals. Screening and categorization methods that rely on a decision whether a chemical's predicted property falls on either side of a threshold are likely to lead to a significant number of false positive/negative outcomes. We therefore suggest that screening should rather be based on numerical hazard or risk estimates that acknowledge and explicitly take into account the uncertainties of predicted properties.