MOSAICbioacc




The MOSAICbioacc application is a turn-key web tool providing bioaccumulation metrics (BCF/BMF/BSAF) from a toxicokinetic (TK) model fitted to accumulation-depuration data. It is designed to fulfil the requirements of regulators when examining applications for market authorization of active substances.
Learn more

Toxicokinetic/toxicodynamic (TKTD) models are used to describe and predict the toxicity and the effects of chemical substances on individual traits based on experimental data. The toxicokinetic (TK) part describes the relationship between the exposure medium and the organism, considering various processes such as ADME (accumulation, depuration, metabolization and excretion)1. Regulation No 283/2013 (EU)2 defines the data requirements for active substances of plant protection products in marketing authorization applications. In particular, a bioaccumulation study on fish is required following OECD guideline 3053. Achieved in agreement with EFSA's scientific opinion on good modeling practices4,5, this ready-to-use on-line service allows to easily estimate bioaccumulation metrics (namely, BCF/BMF/BSAF) as required in a regulatory framework, accounting for bioaccumulation of parent compounds and their metabolites through biotransformation. MOSAICbioacc does not expect any input besides the accumulation-depuration datasets according to exposure concentrations. The service automatically fits the TK model, suitably selected for the user by the application, and optimizes its parameters. Then, the service provides the corresponding bioaccumulation metrics, as well as all goodness-of-fit criteria required to carefully check the reliability of the results6. All calculations are based on JAGS software and companion R packages rjags and jagsUI 7,8,9. More details about the underlying modeling framework can be found in the user guide and companion research papers10.





Contact: sandrine.charles@univ-lyon1.fr

Gamma version (updated on 2021-04-13)


This work is supported by the EUR H2O’Lyon (ANR-17-EURE-0018) of Université de Lyon (UdL), within the program “Investissements d’Avenir” operated by the French National Research Agency (ANR).

Data upload




Time unit:


Lipid fraction
Is lipid fraction available?
Mean lipid fraction (based on wet weight):

Duration of accumulation phase:


Data Visualisation

Exposure:

Model and parameters


Accumulation:

(Select one or several exposure media)

Depuration:

(Select involved processes)


Click here for more information about parameters meaning.

Please click here
after each change

This is the deterministic part of the model with:



Click here for more information about the model (determistic and stochastic parts) and here for solving equations.

Click here for more information about the type of fitting algorithm used, the number of iterations required...


Bioaccumulation metric calculation


Click here for more information about the BCF calculation.


Probability density of the bioaccumulation metric. The middle dotted line represents the median value, and the dotted lines at the right and left are the 2.5 and 97.5% quantiles.



Median and 95% uncertainty limits of bioaccumulation metrics:





From the fit below, if your data are very close to the asymptotic steady-state at the end of the accumulation phase, you can ask for the BCFss :


Click here for more information about the BCFpw calculation.


Probability density of the bioaccumulation metric. The middle dotted line represents the median value, and the dotted lines at the right and left are the 2.5 and 97.5% quantiles.



Median and 95% uncertainty limits of bioaccumulation metrics:





From the fit below, if your data are very close to the asymptotic steady-state at the end of the accumulation phase, you can ask for the BCFss :


Click here for more information about the BSAF calculation.


Probability density of the bioaccumulation metric. The middle dotted line represents the median value, and the dotted lines at the right and left are the 2.5 and 97.5% quantiles.



Median and 95% uncertainty limits of bioaccumulation metrics:





From the fit below, if your data are very close to the asymptotic steady-state at the end of the accumulation phase, you can ask for the BSAFss :


Click here for more information about the BMF calculation.


Probability density of the bioaccumulation metric. The middle dotted line represents the median value, and the dotted lines at the right and left are the 2.5 and 97.5% quantiles.



Median and 95% uncertainty limits of bioaccumulation metrics:







From the fit below, if your data are very close to the asymptotic steady-state at the end of the accumulation phase, you can ask for the BMFss :


Fitting results





Measured (black dots) and predicted contaminant concentrations in the organism (μg.g-1). Median predictions are symbolized by the orange plain line and the uncertainty bands by the gray zone which is delimited by the 2.5% and 97.5% quantiles in orange dotted lines. The black dotted vertical line indicates the separation between the accumulation phase and the depuration phase.

Quantiles of estimated parameters:



Goodness-of-fit criteria




Goodness-of-fit criteria are given below in our prioritized order; the PPC and the prior-posterior comparison are the most important to check; if they do not correspond to the expectation, you must consider your results with an even more particular attention. More details.

If goodness-of-fit criteria do not fully match the expectations, go to the appendix of the user guide to learn more about the interpretation.



Posterior Predictive Check




The PPC shows the observed values against their corresponding estimated predictions (black dots), along with their 95% credible interval (vertical segments). If the fit is correct, we expect to see 95% of the data within the intervals. Ideally observations and predictions should coincide, so we would expect to see black dots along the first bisector y = x (plain black line). The 95% credible intervals are colored in green if they overlap this line, in red otherwise. User guide.


Priors and Posteriors




The prior distribution is represented by the gray area and the posterior distribution by the orange area. The accuracy of the model parameter estimation can be visualized by comparing prior and posterior distributions: the overall expectation is to get a narrower posterior distribution compared to the prior one, what reflects that data contributed enough to precisely estimate parameters. The deterministic part represents the mean trend, while the stochastic part describes the variability around this mean trend. User guide.








   


Correlations between parameters




By default, a colored matrix allows to see at a glance the most correlated (green, Pearson coefficient closed to 1), anti-correlated (orange, Pearson coefficient closed to -1) or not correlated (Pearson coefficient closed to 0) parameters.
Correlations between parameters are also visualized by projecting the joint posterior distribution in a plot matrix with planes of parameter pairs (lower triangular elements), marginal posterior distribution of each model parameter (diagonal), and Pearson correlation coefficients (upper triangular elements). Correlations are expected to be low (reflected by “potatoid” shapes of density lines in orange); a leaning elliptical shape translates high correlation (positive if leaning to the right, negative if leaning to the left). The deterministic part represents the mean trend, while the stochastic part describes the variability around this mean trend. User guide.










Potential Scale Reduction Factors




Convergence of the MCMC chains can be check with the Gelman-Rubin diagnostic expressed with the potential scale reduction factor (PSRF). Approximate convergence is better diagnosed when the PSRF is below to 1.0111, 12. User guide.


Watanabe–Akaike information criterion




Information criteria offer a computationally appealing way of estimating the generalization performance of the model. A fully Bayesian criterion is the widely applicable information criterion (WAIC) by Watanabe a penalized deviance statistics accounting for the uncertainty in the parameters and can be used also for singular models. WAIC is widely used in model comparison for a same dataset (e.g., with or without \(k_{ee}\)). Sub-models with lower WAIC values will be preferred13. User guide.




Deviance Information Criterion




This criteria, denoted DIC, is a penalized deviance statistics accounting for the number of parameters for use in model comparison for a same dataset. As for the WAIC criterion, sub-models with lower DIC values will be preferred13. User guide.




Traces of MCMC iterations




A traceplot is an essential plot for assessing convergence and diagnosing of MCMC chains. It shows the time series of the sampling process leading to the posterior distribution. Different colors are used for each of the chains (here 3) to assess within-chain convergence. The three chains are expected to converge towards the same values. The deterministic part represents the mean trend, while the stochastic part describes the variability around this mean trend.
User guide.




Downloads


Download all outputs (.zip)

Download

Download an output
Download
Download a table

Download
Build a report
Including data table of analysis?
Download


Download R script

Download

References


[1] Mackay, D. and Fraser, A. 2000. Bioaccumulation of persistent organic chemicals: Mechanisms and models. Environmental Pollution. 110:375-391. https://doi.org/10.1016/S0269-7491(00)00162-7
[2] Commission Regulation (EU) No 283/2013 of 1 March 2013 setting out the data requirements for active substances, in accordance with Regulation (EC) No 1107/2009 of the European Parliament and of the Council concerning the placing of plant protection products on the market Text with EEA relevance. Pages 1–84. Official Journal of the European Union, 93, 3.4.2013. https://eur-lex.europa.eu/legal-content/EN/ALL/?uri=CELEX%3A32013R0283
[3] OECD (2012) Test No. 305: Bioaccumulation in Fish: Aqueous and Dietary Exposure, OECD Guidelines for the Testing of Chemicals, Section 3, OECD Edition, Paris, https://eur-lex.europa.eu/legal-content/EN/ALL/?uri=CELEX%3A32013R0283
[4] Charles, S., P. Veber, and M. L. Delignette-Muller. 2018. MOSAIC: a web-interface for statistical analyses in ecotoxicology. Environmental Science and Pollution Research, 25:11295-11302. https://doi.org/10.1007/s11356-017-9809-4
[5] EFSA. 2014. Scientific Opinion on good modelling practice in the context of mechanistic effect models for risk assessment of plant protection products. EFSA Journal 12. https://www.efsa.europa.eu/en/efsajournal/pub/3589
[6] Fernández-i-Marín, X. (2016). ggmcmc: Analysis of MCMC samples and Bayesian inference.Journal of Statistical Software, 70(9), 1-20. https://doi.org/10.18637/jss.v070.i09
[7] Plummer, M. (2019). rjags: Bayesian Graphical Models using MCMC. R package version 4-10. https://CRAN.R-project.org/package=rjags
[8] R Core Team (2019). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/.
[9] Kellner, K. (2019). jagsUI: A Wrapper Around 'rjags' to Streamline 'JAGS' Analyses. Version 1.5.1. https://cran.r-project.org/web/packages/jagsUI/index.html
[10] Ratier, A. et al. (2021). A new on-line tool for the calculation of bioaccumulation metrics of active substances within living organisms: MOSAICBioacc. bioRxiv. https://www.biorxiv.org/content/10.1101/2020.07.07.185835v1.
[11] Gelman, A. and Rubin, D. (1992). Inference from iterative simulation using multiple sequences. Statistical Science, 7:457-511. https://www.jstor.org/stable/2246093
[12] Vehtari, A. et al. (2019). Rank-normalization, folding, and localization: An improved R for assessing convergence of MCMC. arXiv:1903.08008v3. https://arxiv.org/ct?url=https%3A%2F%2Fdx.doi.org%2F10.1214%2F20-BA1221&v=8607c76e
[13] Spiegelhalter, D., Best, N., Carlin, B. and van der Linde, A. (2002). Bayesian measures of model complexity and fit (with discussion). Journal of the Royal Statistical Society Series B, 64: 583-639. https://doi.org/10.1111/1467-9868.00353.
[14] Ciccia, T. (2019). Accumulation et devenir du mercure chez l'espèce sentinelle Gammarus fossarum : de l'expérimentation au développement d'un modèle toxicocinétique multi-compartiments. Rapport de stage de Master 2, INRAE.
[15] Crookes, M. and Brooke D. (2011). Estimation of fish bioconcentration factor (BCF) from depuration data. Report. Environmental Agency, Bristol, UK. https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/291527/scho0811buce-e-e.pdf.
<<<<<<< HEAD [16] Ashauer, R. et al. (2012). Significance of xenobiotic metabolism for bioaccumulation kinetics of organic chemicals in Gammarus pulex. Environmental Science Technologie, 46: 3498-3508. https://doi.org/10.1021/es204611h.
[17] Schuler, L.J. et al. (2003). Toxicokinetics of sediment-sorbed benzo[a]pyrene and hexachlorobiphenyl using the freshwater invertebrates Hyalella azteca, Chironomus tentans, and Lumbriculus variegatus. Environmental Toxicology and Chemistry, 22: 439-449. https://doi.org/10.1002/etc.5620220227. ======= [16] Larisch, W., Brown, T.N. and Goss, K.‐U. (2017). A toxicokinetic model for fish including multiphase sorption features. Environmental Toxicology and Chemistry, 3: 1538-1546. https://doi.org/10.1002/etc.36.
[17] Ashauer, R. et al. (2012). Significance of xenobiotic metabolism for bioaccumulation kinetics of organic chemicals in Gammarus pulex. Environmental Science Technologie, 46: 3498-3508. https://doi.org/10.1021/es204611h.
[18] Schuler, L.J. et al. (2003). Toxicokinetics of sediment-sorbed benzo[a]pyrene and hexachlorobiphenyl using the freshwater invertebrates Hyalella azteca, Chironomus tentans, and Lumbriculus variegatus. Environmental Toxicology and Chemistry, 22: 439-449. https://doi.org/10.1002/etc.5620220227. >>>>>>> Gamma

Staff and Contributors

Aude Ratier, post-doc at LBBE
Aurélie Silberchicot, Software Engineer at LBBE
Christelle Lopes, Associate Professor at University Lyon 1
Gauthier Multari, bachelor student at University Lyon 1
Sandrine Charles, Professor at University Lyon 1
Stéphane Delmotte, Software Engineer at LBBE