Registration Dossier

Environmental fate & pathways

Bioaccumulation: aquatic / sediment

Currently viewing:

Administrative data

Endpoint:
bioaccumulation: aquatic / sediment
Type of information:
(Q)SAR
Adequacy of study:
weight of evidence
Reliability:
4 (not assignable)
Rationale for reliability incl. deficiencies:
other: Methods were validated by US EPA using statistical external validation; based on the mean absolute errors of the models the confidence in the predicted results is low.
Justification for type of information:
QSAR prediction: migrated from IUCLID 5.6

Data source

Referenceopen allclose all

Reference Type:
other: calculation
Title:
Unnamed
Year:
2014
Report Date:
2014
Reference Type:
other: Estimation software
Title:
T.E.S.T. (Toxicity Estimation Software Tool), v4.1
Author:
US EPA
Year:
2012
Bibliographic source:
United States Environmental Protection Agency, Washington, DC, USA

Materials and methods

Principles of method if other than guideline:
T.E.S.T. is a toxicity estimation software tool. The program requires only the molecular structure of the test item, all other molecular descriptors which are required to estimate the toxicity are calculated within the tool itself. The molecular descriptors describe physical characteristics of the molecule (e.g. E-state values and E-state counts, constitutional descriptors, topological descriptors, walk and path counts, connectivity, information content, 2d autocorrelation, Burden
eigenvalue, molecular property (such as the octanol-water partition coefficient), Kappa, hydrogen bond acceptor/donor counts, molecular distance edge, and molecular fragment counts). Each of the available methods uses a different set of these descriptors to estimate the toxicity.
The bioaccumulation factor (BCF) was estimated using several available methods: hierarchical method; FDA method, single model method; group contribution method; nearest neighbor method; consensus method. The methods were validated using statistical external validation using separate training and test data sets.
The experimental data set was obtained from several different databases (Dimitrov et al., 2005; Arnot and Gobas, 2006; EURAS; Zhao, 2008). From the available data set salts, mixtures and ambiguous compounds were removed. The final data set contained 676 chemicals.

References:
- Dimitrov, S., N. Dimitrova, T. Parkerton, M. Combers, M. Bonnell, and O. Mekenyan. 2005. Base-line model for identifying the bioaccumulation potential of chemicals. SAR and QSAR in Environmental Research 16:531-554.
- Arnot, J.A., and F.A.P.C. Gobas. 2006. A review of bioconcentration factor (BCF) and bioaccumulation factor (BAF) assessments for organic chemicals in aquatic organisms. Environ. Rev. 14:257-297.
- EURAS. Establishing a bioconcentration factor (BCF) Gold Standard Database. EURAS [cited 5/20/09]. Available from http://www.euras.be/eng/project.asp?ProjectId=92.
- Zhao, C.; Boriani, E.; Chana, A.; Roncaglioni, A.; Benfenati, E. 2008. A new hybrid system of QSAR models for predicting bioconcentration factors (BCF). Chemosphere 73:1701-1707.
GLP compliance:
no

Test material

Reference
Name:
Unnamed
Type:
Constituent
Type:
Constituent
Details on test material:
- Name of test material (as cited in study report): DL-alpha-Tocopherol

Test organisms

Test organisms (species):
other: fish

Test conditions

Details on estimation of bioconcentration:
BASIS FOR CALCULATION OF BCF
- Estimation software: US EPA T.E.S.T. v4.1

Applied estimation methods:
- Hierarchical method : The toxicity for a given query compound is estimated using the weighted average of the predictions from several different cluster models.
- FDA method : The prediction for each test chemical is made using a new model that is fit to the chemicals that are most similar to the test compound. Each model is generated at runtime.
- Single model method : Predictions are made using a multilinear regression model that is fit to the training set (using molecular descriptors as independent variables).
- Group contribution method : Predictions are made using a multilinear regression model that is fit to the training set (using molecular fragment counts as independent variables).
- Nearest neighbor method : The predicted toxicity is estimated by taking an average of the 3 chemicals in the training set that are most similar to the test chemical.
- Consensus method : The predicted toxicity is estimated by taking an average of the predicted toxicities from the above QSAR methods (provided the predictions are within the respective applicability domains; recommended method by T.E.S.T. for providing the most accurate predictions).

Results and discussion

Bioaccumulation factor
Type:
BCF
Value:
272
Remarks on result:
other: method: consensus (average of reasonable results from all models); log BCF = 2.43; Based on the mean average error, the confidence in the predicted BCF values is low.

Any other information on results incl. tables

Model details:

Method

Predicted value

Model statistics

MAE (in log10)

External test set

Training set

log BCF

BCF

No. of chemicals

Entire set

SC >= 0.5

Entire set

SC >= 0.5

Consensus method

2.43

271.96

-

-

-

0.51

0.69

0.42

0.52

Hierarchical clustering

2.33

215.22

(46.65-993.04)

0.662 - 0.776

0.569 - 0.733

112 - 540

(cluster models: 5)

0.54

0.83

0.23

0.31

Single model

2.52

332.30

(22.72-4860.96)

-

0.733

540

0.54

0.61

0.53

0.65

Group contribution

3.50

3.173.63

(121.30-83034.19)

-

0.527

499

0.62

0.82

0.60

0.52

FDA

2.66

455.76

(55.38-3750.46)

0.854

0.785

59

0.57

0.78

0.53

0.68

Nearest neighbor

1.16

14.38

-

-

3

0.60

0.87

0.55

0.67

*: Insufficient nearest neighbors in the training set were available to make a prediction.

SC = Similarity Coefficient

r² = correlation coefficient

q² = leave one out correlation coefficient

Confidence in predicted results

The MAEs for the predicted results based on similar chemicals (SC >= 0.5) of the external test set and the training set are generally higher than the MAEs for the predictions based on the entire set of chemicals. There is only one exception (MAE for training sets from Group contribution). Overall, the confidence in the predicted BCF values is low.

Applicant's summary and conclusion