Registration Dossier

Toxicological information

Skin sensitisation

Currently viewing:

Administrative data

Endpoint:
skin sensitisation, other
Remarks:
in silico
Type of information:
(Q)SAR
Adequacy of study:
weight of evidence
Reliability:
2 (reliable with restrictions)
Rationale for reliability incl. deficiencies:
results derived from a (Q)SAR model, with limited documentation / justification, but validity of model and reliability of prediction considered adequate based on a generally acknowledged source
Justification for type of information:
This endpoint study record is part of a Weight of Evidence approach comprising of two QSAR model predictions and a human Repeated Insult Patch (RIP) studies. The results of the QSAR and the RIP studies agree as to the skin sensitisation potential and are sufficient to fulfil the information requirements as further explained in the provided skin sensitisation endpoint summary.

QSAR Prediction:
1. SOFTWARE: Toxtree (estimation of Toxic Hazard – A Decision Tree Approach) v.2.6.6

2. MODEL (incl. version number): Toxtree Module (v.2.6.6): Skin sensitisation reactivity domains

3. SMILES OR OTHER IDENTIFIERS USED AS INPUT FOR THE MODEL: OC(C)CCO

4. SCIENTIFIC VALIDITY OF THE (Q)SAR MODEL
- Defined endpoint: Yes, the alerts provide grouping into reactivity mode of action and do not predict skin sensitisation potential. Identification of mechanisms of toxic action for skin sensitisation using a SMARTS pattern based approach.
Enoch, S. J., J. C. Madden, and M. T. D. Cronin. "Identification of mechanisms of toxic action for skin sensitisation using a SMARTS pattern based approach." SAR and QSAR in Environmental Research 19.5-6 (2008): 555-578.

5. APPLICABILITY DOMAIN
the applicability domain is not defined

6. ADEQUACY OF THE RESULT
The alerts provide grouping into reactivity mode of action and do not predict skin sensitisation potential
Cross-referenceopen allclose all
Reason / purpose:
reference to other study
Reference
Endpoint:
sensitisation data (humans)
Type of information:
experimental study
Adequacy of study:
weight of evidence
Reliability:
4 (not assignable)
Rationale for reliability incl. deficiencies:
documentation insufficient for assessment
Justification for type of information:
This endpoint study record is part of a Weight of Evidence approach comprising of two QSAR model predictions and a human Repeated Insult Patch (RIP) studies. The results of the QSAR and the RIP studies agree as to the skin sensitisation potential and are sufficient to fulfil the information requirements as further explained in the provided skin sensitisation endpoint summary.
Reason / purpose:
reference to other study
Related information:
Composition 1
Reason / purpose:
reference to other study
Related information:
Composition 1
Reference:
Composition 0
Type of sensitisation studied:
skin
Study type:
study with volunteers
Guideline:
other: Repeated Insult Patch test
GLP compliance:
not specified
Test material information:
Composition 1
Type of population:
general
Ethical approval:
not specified
Route of administration:
dermal
Results of examinations:
No signs of sensitisation were observed in several repeated patch tests with products containing the test substance in concentrations up to 16% (Schwartz-Peck prophetic patch test: 104 subjects, formulation with 5%; Draize-Shelanski RIPTs: 108 subjects, formulation with 16%; 49 subjects, 5% formulation; 108 subjects, formulation with 3%). There was also no photosensitisation.
Conclusions:
No sensitisation or photosensitisation was observed in several human studies with cosmetic formulations containing the test substance.
Executive summary:

No sensitisation or photosensitisation was observed in controlled studies with formulations containing the test substance in concentration up to 16%. Several hundred volunteers have been tested in total.

Reason / purpose:
reference to other study
Reference
Endpoint:
skin sensitisation, other
Remarks:
in silico
Type of information:
(Q)SAR
Adequacy of study:
weight of evidence
Reliability:
2 (reliable with restrictions)
Rationale for reliability incl. deficiencies:
results derived from a valid (Q)SAR model and falling into its applicability domain, with adequate and reliable documentation / justification
Justification for type of information:
This endpoint study record is part of a Weight of Evidence approach comprising of two QSAR model predictions and a human Repeated Insult Patch (RIP) studies. The results of the QSAR and the RIP studies agree as to the skin sensitisation potential and are sufficient to fulfil the information requirements as further explained in the provided skin sensitisation endpoint summary.

QSAR Prediction:
1. SOFTWARE
VEGA (IFRMN)

2. MODEL (incl. version number)
Skin Sensitization model (CAESAR) 2.1.6

3. SMILES OR OTHER IDENTIFIERS USED AS INPUT FOR THE MODEL
OCCC(O)C

4. SCIENTIFIC VALIDITY OF THE (Q)SAR MODEL
- Defined endpoint:
Skin Sensitization – positive or negative

- Unambiguous algorithm:
SAR/QSAR based on calculator

- Defined domain of applicability:
The applicability domain of predictions is assessed using an Applicability Domain Index (ADI) that has values from 0 (worst case) to 1 (best case). The ADI is calculated by grouping several other indices, each one taking into account a particular issue of the applicability domain. Most of the indices are based on the calculation of the most similar compounds found in the training and test set of the model, calculated by a similarity index that consider molecule's fingerprint and structural aspects (count of atoms, rings and relevant fragments).

- Appropriate measures of goodness-of-fit and robustness and predictivity:
Training set: n = 167; Accuracy = 0.91; Specificity = 0.74; Sensitivity = 0.95
Test set: n = 42; Accuracy = 0.93; Specificity = 0.75; Sensitivity = 0.97

- Mechanistic interpretation:
The model consists in an Adaptive Fuzzy Partition (AFP) based on 8 descriptors. The AFP produces as output two values O(positive) and O(negative) that represent the belonging degree respectively to the sensitizer and non-sensitizer classes. The input compound is assigned to the class having this degree value higher than 0.5, unless the difference between the values of the two degrees is lower than the threshold 0.001; in this case, the belonging to one class or the other is not sure, thus no prediction is made. Full reference and details of the used formulas can be found in: Chaudhry, Q., Piclin, N., Cotterill, J., Pintore, M., Price, N. R., Chrétien, J. R. and Roncaglioni, A. (2010). Global QSAR models of skin sensitisers for regulatory purposes., Chem Cent J 4 Suppl 1, S5.

5. APPLICABILITY DOMAIN
The descriptors used by the AFP model are the following:
- nN: Number of nitrogen atoms
- GNar: Narumi geometric topological index
- MDDD: Mean Distance Degree Deviation
- X2v: Valence connectivity index chi-2
- EEig10r: Eigenvalue 10 from the edge adjacency matrix weighted by resonance integrals
- GGI8: Topological charge index of order 8
- nCconj: Number of non aromatic conjugated C(sp2)
- O-058: Atom-centred fragment =O
- Similar molecules with known experimental value. This index takes into account how similar are the first two most similar compounds found. Values near 1 mean that the predicted compound is well represented in the dataset used to build the model, otherwise the prediction could be an extrapolation.
- Accuracy of prediction for similar molecules. This index takes into account the classification accuracy in prediction for the two most similar compounds found. Values near 1 mean that the predicted compounds falls in an area of the model's space where the model gives reliable predictions (no misclassifications), otherwise the lower is the value, the worse the model behaves.
- Concordance for similar molecules . This index takes into account the difference between the predicted value and the experimental values of the two most similar compounds. Values near 0 mean that the prediction made disagrees with the values found in the model's space, thus the prediction could be unreliable.
- Atom Centered Fragments similarity check. This index takes into account the presence of one or more fragments that aren't found in the training set, or that are rare fragments. First order atom centered fragments from all molecules in the training set are calculated, then compared with the first order atom centered fragments from the predicted compound; then the index is calculated as following: a first index RARE takes into account rare fragments (those who occur less than three times in the training set), having value of 1 if no such fragments are found, 0.85 if up to 2 fragments are found, 0.7 if more than 2 fragments are found; a second index NOTFOUND takes into account not found fragments, having value of 1 if no such fragments are found, 0.6 if a fragments is found, 0.4 if more than 1 fragment is found. Then, the final index is given as the product RARE * NOTFOUND.
- Model descriptors range check. This index checks if the descriptors calculated for the predicted compound are inside the range of descriptors of the training and test set. The index has value 1 if all descriptors are inside the range, 0 if at least one descriptor is out of the range.
- Global AD Index. The final global index takes into account all the previous indices, in order to give a general global assessment on the applicability domain for the predicted compound.

- Similarity with analogues in the training set:
strongly similar compounds with known experimental value in the training set have been found

6. ADEQUACY OF THE RESULT
Prediction has high reliability (compound into the AD)

Global AD Index
AD index = 0.955
Explanation: the predicted compound is into the Applicability Domain of the model.

Similar molecules with known experimental value
Similarity index = 0.912
Explanation: strongly similar compounds with known experimental value in the training set have been found.

Accuracy of prediction for similar molecules
Accuracy index = 1
Explanation: accuracy of prediction for similar molecules found in the training set is good.

Concordance for similar molecules
Concordance index = 1
Explanation: similar molecules found in the training set have experimental values that agree with the predicted value.

Model's descriptors range check
Descriptors range check = True
Explanation: descriptors for this compound have values inside the descriptor range of the compounds of the training set.

Atom Centered Fragments similarity check
ACF index = 1
Explanation: all atom centered fragment of the compound have been found in the compounds of the training set.
Reason / purpose:
reference to other study
Related information:
Composition 1
Reason / purpose:
reference to other study
Related information:
Composition 1
Reference:
Composition 0
Principles of method if other than guideline:
- Software tool(s) used including version: VEGA (IFRMN)
- Model(s) used: Skin Sensitization model (CAESAR) 2.1.6
- Model description: see field 'Justification for type of information'
- Justification of QSAR prediction: see field 'Justification for type of information'
GLP compliance:
no
Test material information:
Composition 1
Specific details on test material used for the study:
SMILES: OCCC(O)C
Parameter:
other: Skin Sensitization model (CAESAR) 2.1.6
Run / experiment:
QSAR prediction
Remarks on result:
not measured/tested
Remarks:
Prediction is NON-Sensitizer, the result appears reliable.
Interpretation of results:
study cannot be used for classification
Conclusions:
The VEGA Skin Sensitization model (CAESAR) 2.1.6 prediction is NON-Sensitizer, the result appears reliable.

Data source

Reference
Reference Type:
study report
Title:
Unnamed
Year:
2017

Materials and methods

Principles of method if other than guideline:
- Software tool(s) used including version: Toxtree (estimation of Toxic Hazard – A Decision Tree Approach) v.2.6.6
- Model(s) used: Toxtree Module (v.2.6.6): Skin sensitisation reactivity domains
- Model description: see field 'Justification for type of information'
- Justification of QSAR prediction: see field 'Justification for type of information'
GLP compliance:
no

Test material

Reference
Name:
Unnamed
Type:
Constituent
Test material form:
liquid
Details on test material:
no impurities described
Specific details on test material used for the study:
SMILES: OC(C)CCO

Results and discussion

In vitro / in chemico

Results
Parameter:
other: Skin sensitisation reactivity domain alerts identified
Run / experiment:
QSAR prediction
Remarks on result:
not determinable
Remarks:
No skin sensitisation reactivity domain alerts identified

Applicant's summary and conclusion

Interpretation of results:
study cannot be used for classification
Conclusions:
The Toxtree Module (v.2.6.6): Skin sensitisation reactivity domains determined there are no skin sensitisation reactivity domain alerts.