The Combined QSAR-ICE Models: Practical Application in Ecological

Citation data is made available by participants in Crossref's Cited-by Linking service. For a more comprehensive list of citations to this article, us...
0 downloads 0 Views 826KB Size
Viewpoint pubs.acs.org/est

The Combined QSAR-ICE Models: Practical Application in Ecological Risk Assessment and Water Quality Criteria Jia He,† Zhi Tang,† Yuanhui Zhao,‡ Ming Fan,§ Scott D. Dyer,§ Scott E. Belanger,§ and Fengchang Wu*,† †

State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences, Beijing 100012, China ‡ School of Environment, Northeast Normal University, Changchun, 130027, China § Global Product Stewardship, Procter and Gamble Company, 8700 Mason Montgomery Road, Mason, Ohio 45040, United States Independent QSAR and ICE models both have advantages and disadvantages. QSAR models can be used to predict physicochemical and biological (e.g., toxicological) properties of molecules from knowledge of chemical structure, which also represent theoretical significance.3 Unfortunately, QSAR models are only able to predict different chemical toxicities for certain species or taxa types (e.g., Daphnia and fish). ICE models describe the relationship with the acute toxicity value (e.g., LC50) for a range of chemicals tested between two species. If toxicity data are available for surrogate species, toxicity to the predicted taxon can be estimated using the ICE models for a particular interspecies pair.4 However, the majority of the current ICE models available are based on North American native species. We combined QSAR and ICE models together with an aim of leveraging the advantages of their corresponding prediction abilities. The combined QSAR-ICE models were developed from models based on surrogate species using the following steps: First, available acute toxicity values for aquatic species (medium effective/lethal concentration (EC/LC50)) and wildlife species (medium lethal dose (LD50)) were used to establish least-squares regression of the relationships between surrogate and predicted taxa (see ICE eq 1). Second, all available acute toxicity values for ne of the greatest challenges both in ecological risk surrogate species were used to establish a number of linear assessment (ERA) and water quality criteria (WQC) is to regressions of relationships between toxicity values of surrogate extrapolate across a broad range of species exposed to species and structural descriptors of chemicals (see QSAR eq 2). contaminants using toxicity data for only a limited number of Third, QSAR models of surrogates were substituted into the commonly tested species.1,2 Over the past 40 years, thousands of appropriate ICE models, resulting in the combined QSAR-ICE tests with aquatic biota have been conducted to assess potential models (eq 3). risks of chemicals. However, hundreds of new synthetic chemicals are being introduced into the market every year, and ICE: log10(predicted taxa toxicity) it is impossible to test required testing species for such an = a + b × log10(surrogate taxa toxicity) enormous number of existing and new chemicals. Quantitative (1) structure activity relationship (QSAR) models provide opportunities to estimate ecotoxicity values (i.e., lethal concentration QSAR: log10(surrogate toxicity) of 50% (LC50)) for the three main aquatic trophic levels (i.e., = a0 + a1f1 + a 2f2 + ··· + anfn algae, invertebrates and fish) based on knowledge of chemical (2) structures or properties. Additionally, the development of statistical extrapolation methods, such as interspecies correlation QSAR − ICE: log10(predicted toxicity) estimation (ICE) models, that use available toxicity data of = a + b × (a0 + a1f1 + a 2f2 + ··· + anfn ) surrogate species (i.e., Danio rerio) to predict untested species2 has drawn global attention. QSAR results coupled with ICE = A 0 + A1f1 + A 2 f2 + ··· + A nfn (3) models could greatly expand the ability to predict untested chemicals and potential effects on untested species allowing for development of screening level of ERA and/or WQC across Received: May 26, 2017 multiple species.

O

© XXXX American Chemical Society

A

DOI: 10.1021/acs.est.7b02736 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Viewpoint

Environmental Science & Technology Where a, a0, A0, and b, a1···an, A1···An are intercepts and slopes of the corresponding models, respectively; f1···f n are specific structural descriptors (e.g., octanol−water partition coefficient, molecular weight etc.). The combined QSAR-ICE models are methods for estimating toxicity of predicted taxa from chemical structural descriptors of surrogate taxa. QSAR-ICE models could simultaneously enlarge the predictive field of species and chemicals as a database supplement methodology for deriving WQC and/or conducting ERA (see Figure 1). Such QSAR-ICE models could be used to

data gaps when deriving such criteria, and increasing overall weight of evidence. The combined models provide practical applications in both the problem formulation stage, when diverse contaminants can be screened for their potential to cause harm to ecological communities, and in characterizing risks to various taxa (i.e., which species may be most vulnerable). Currently, the development of QSAR-ICE models still faces some challenges and needs further improvement. For example, there is the species regional limitation in ICE models as the majority of ICE models were established using toxicity data of North American species.4 Our research has enrolled toxicity data for species native to China (about 60 native species of amphibians and fish), which can broaden application of QSARICE models to be used to derive WQC for China. The importance of biogeographic differences in interspecies toxicity relationships is not fully explored. Additionally, uncertainty of QSAR-ICE models predictions should be fully recognized: Such as if values for toxicities of chemicals to surrogate species or predicted from structural descriptors are outside the range of data used to establish QSAR-ICE models, the predicted values should not be used. Each QSAR-ICE models should be developed using strict statistical validation methods, including base training sets, leave-one-out cross-validation, evaluation of confidence intervals as well as external validation based test sets, to ensure model accuracy and precision. Other considerations include selecting the most significant molecular descriptors, evaluating sample size requirements and assessing impacts of taxonomic distance. To date, the potential application of QSARICE models could satisfy the immediate needs for WQC derivations, as well as help basic data support for ecological risk assessment.



AUTHOR INFORMATION

Corresponding Author

*Phone: +86-10-84915312; e-mail: [email protected]. ORCID

Fengchang Wu: 0000-0003-2615-2849 Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was jointly supported by National Natural Science Foundation of China (41521003, 41630645, 41603113) and Postdoctoral Science Foundation of China (2016M591227).



Figure 1. Development and application of QSAR-ICE models. (Middle graph) Species sensitivity distributions (SSDs) generated from QSARICE models to derive a predicted no-effect concentration in the environment, typically blue dotted line showed the fifth percentile of the SSD, termed the HC5 (hazardous concentration affecting 5% of species) indicated protective of 95% of tested species.

REFERENCES

(1) Arnot, J. A.; Mackay, D.; Webster, E.; Southwood, J. M. Screening level risk assessment model for chemical fate and effect in the environment. Environ. Sci. Technol. 2006, 40, 2316−2323. (2) Dyer, S. D.; Versteeg, D. J.; Belanger, S. E.; Chaney, J. G.; Raimondo, S.; Barron, M. G. Comparison of species sensitivity distributions derived from interspecies correlation models to distributions used to derive water quality criteria. Environ. Sci. Technol. 2008, 42, 3076−3083. (3) Lipnick, R. L. Correlative and mechanistic QSAR models in toxicology. SARQSAR Environ. Res. 1999, 10, 239−248. (4) Raimondo, S.; Jackson, C. R.; Barron, M. G. Influence of taxonomic relatedness and chemical mode of action in acute interspecies estimation models for aquatic species. Environ. Sci. Technol. 2010, 44, 7711−7716. (5) Feng, C. L.; Wu, F. C.; Dyer, S. D.; Chang, H.; Zhao, X. L. Derivation of freshwater quality criteria for zinc using interspecies correlation estimation models to protect aquatic life in China. Chemosphere 2013, 90, 1177−1183.

generate species sensitivity distributions (SSDs), which estimate a hazard level that is protective of the most species within the distributions and SSDs is one of the currently more preferred methods by ecotoxicologists for derivation of WQC because they consolidate all available data for a chemical of interest. For example, zinc experimental toxicity data are limited.5The application of QSAR-ICE models could enable the use of structural descriptors (i.e., softness index (σp) and maximum complex stability constants (log−βn)) to estimate supplemental taxa (including endangered species) sensitivity data, filling zinc B

DOI: 10.1021/acs.est.7b02736 Environ. Sci. Technol. XXXX, XXX, XXX−XXX