Identifying ecotoxicological descriptors to enable predictive hazard assessments of nano-TiO2 from a meta-analysis of ecotoxicological data
| Autor: | Bernd Nowack, Henning Wigger, Yaping Cai |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Hazard (logic)
Future studies Materials Science (miscellaneous) Public Health Environmental and Occupational Health 02 engineering and technology 010501 environmental sciences Nano tio2 021001 nanoscience & nanotechnology computer.software_genre 01 natural sciences 13. Climate action Meta-analysis Linear regression Environmental science Data mining 0210 nano-technology Safety Risk Reliability and Quality Safety Research Exposure duration computer 0105 earth and related environmental sciences Environmental risk assessment |
| Zdroj: | NanoImpact |
| ISSN: | 2452-0748 |
| DOI: | 10.1016/j.impact.2019.100180 |
| Popis: | Hazard assessments for ENMs are made more difficult by the multitude of different nano-forms that have to be tested if a case-by-case approach is applied. Predictive hazard assessments are currently being developed to streamline the environmental risk assessment of ENMs. The present study compiled an ecotoxicological dataset for nano-TiO2 and aimed to identify potential descriptors for the prediction of toxicological effects induced by different nano-forms based on their material properties and experimental conditions. We collected 219 nano-TiO2 data points (in vivo), of which 205 were from freshwater studies. Only 23 of the 65 data points for Daphnia magna—the most investigated species—were considered as high-quality according to the DaNa2.0 criteria. Nano-TiO2's EC50 was predicted using a multiple linear regression (MLR) model for six selected features including intrinsic (primary particle size, crystal composition) and extrinsic parameters (exposure duration, UV and non-UV illumination, concentrations of divalent cations). The EC50 was found to form two main clusters according to the type of illumination, with experiments conducted under UV light resulting in a lower EC50. Nano-TiO2's toxicity to D. magna could be predicted with an R2 of 0.95 (p = 0.15) for the UV dataset and an R2 of 0.55 (p = 0.19) for the non-UV dataset. This was a better performance than the full dataset MLR model which had an R2 of 0.29 (p = 0.41). A one-factor-at-a-time sensitivity analysis identified the share of anatase and the exposure duration as the most sensitive parameters triggering adverse effects. The main impediment to the development of better predictive models is the lack of high-quality datasets with coherent sets of parameters. Future studies will have to overcome several challenges in order to enable a comprehensive approach for measuring and reporting critical parameters and making comparisons between studies. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect