3-(N-p-methoxy cinnamidopropyl-N,N-dimethyl ammonium)-2-hydroxypropane-1-sulphonate

Endpoint summary

Description of key information

(Q)SAR predictions of skin sensitisation were calculated for Galaxy SunBeat using OECD QSAR Toolbox 3.3.3, Biovia Discovery Studio (TOPKAT) 4.5 (extensible), VEGA NIC 1.1.1, and DEREK Nexus 3.0.1.

The results are summarised in the following table.

OECD Toolbox	TOPKAT	VEGA	DEREK
Sensitiser	Non-sensitiser	Sensitiser	Sensitiser

Key value for chemical safety assessment

Skin sensitisation

Link to relevant study records

Reference

Reference 1

Endpoint:

skin sensitisation, other

Remarks:

in silico

Type of information:

(Q)SAR

Adequacy of study:

weight of evidence

Study period:

18 January 2017

Reliability:

2 (reliable with restrictions)

Rationale for reliability incl. deficiencies:

results derived from a valid (Q)SAR model, but not (completely) falling into its applicability domain, with adequate and reliable documentation / justification

Justification for type of information:

1. SOFTWARE
Derek for Windows - Skin sensitisation

2. MODEL (incl. version number)
Derek for Windows version 13 was released in December 2010 and included updates to the skin sensitisation endpoint; Knowledge Base: Derek KB 2014 1.0

3. SMILES OR OTHER IDENTIFIERS USED AS INPUT FOR THE MODEL
SMILES C1=CC(=CC=C1C=CC(NCCC[N+](CC(O)CS(=O)(=O)[O-])(C)C)=O)OC
InChI 1/C18H28N2O6S/c1-20(2,13-16(21)14-27(23,24)25)12-4-11-19-18(22)10-7-15-5-8-17(26-3)9-6-15/h5-10,16,21H,4,11-14H2,1-3H3,(H-,19,22,23,24,25)/b10-7+
Stereochemical features: Four stereoisomers

4. SCIENTIFIC VALIDITY OF THE (Q)SAR MODEL
- Defined endpoint: Skin Sensitisation; Dependent variable: Not applicable; expert rules
- Unambiguous algorithm: QMRF identifier (ECB Inventory):Q13-34-36-315
- Defined domain of applicability:
i. With log Kp < -5 the query compound may be out of domain.
ii. Alert based on quaternary ammonium salt.
iii. Quaternary ammonium salts may exhibit skin sensitisation if the cation is suitably lipophilic which does not apply for the query compound .
- Appropriate measures of goodness-of-fit and robustness and predictivity: The alert is equivocal. According to DEREK mammalian skin sensitisation predictions in Derek associated with a reasoning level of equivocal or above have been considered positive
- Mechanistic interpretation: The presence of a skin sensitisation structural alert within a molecule indicates the molecule has the potential to cause skin sensitisation. Whether or not the molecule will be a skin sensitiser will also depend upon its percutaneous absorption. Generally, small lipophilic molecules are more readily absorbed into the skin and are therefore more likely to cause sensitisation.

5. APPLICABILITY DOMAIN
- Descriptor domain: Calculated
- Similarity with analogues in the training set:
No sensitisers firing the same alert provided/found in the DEREK database. Benzyldimethyl-2-methyl-4(1,1,3,3, tetramethylbutylphenoxyethoxyethylammonium chloride) is given as structure with positive experimental result.

6. ADEQUACY OF THE RESULT
Result is directly applicable since no conversion of the result is required.

Reason / purpose for cross-reference:

(Q)SAR model reporting (QMRF)

Guideline:

other:

Version / remarks:

REACH Guidance on QSARs R.6

Principles of method if other than guideline:

For the prediction of skin sensitisation of Galaxy SunBeat the four models OECD QSAR Toolbox 3.3.5, Biovia Discovery Studio (TOPKAT) 4.5 (extensible), within VEGA NIC 1.1.1 and DEREK were applied in order to provide a weight-of-evidence approach.

GLP compliance:

not specified

Vehicle controls validity:

not applicable

Negative controls validity:

not applicable

Positive controls validity:

not applicable

Remarks on result:

other: Sensitiser. There is only moderate confidence in the prediction as the alert is considered equivocal.

Remarks:

There are indications that the query compound may have a skin sensitising potential. However, for further conclusions predictions from other QSAR may be taken into account.

Interpretation of results:

other: Classification based on weight of evidence approach.

Conclusions:

The substance is considered to be a sensitiser however there is only moderate confidence in the prediction as the alert is considered equivocal.

Reference 2

Endpoint:

skin sensitisation, other

Remarks:

in silico

Type of information:

other: Expert Statement

Adequacy of study:

weight of evidence

Study period:

18 January 2017

Reliability:

other: A weight of evidence approach has been used to fulfil the sensitisation requirements. QSAR testing and review of applicable data has been used.

Rationale for reliability incl. deficiencies:

test procedure in accordance with national standard methods with acceptable restrictions

Reason / purpose for cross-reference:

assessment report

Reason / purpose for cross-reference:

assessment report

Reason / purpose for cross-reference:

assessment report

Reason / purpose for cross-reference:

assessment report

Qualifier:

no guideline available

Principles of method if other than guideline:

For the prediction of skin sensitisation of Galaxy SunBeat the four models OECD QSAR Toolbox 3.3.5, Biovia Discovery Studio (TOPKAT) 4.5 (extensible), within VEGA NIC 1.1.1 and DEREK were applied in order to provide a weight-of-evidence approach.

GLP compliance:

not specified

Type of study:

other: QSAR

Vehicle controls validity:

not applicable

Negative controls validity:

not applicable

Positive controls validity:

not applicable

Remarks on result:

other: Sensitiser

Remarks:

Out of AD Low confidence since both mol weight and log Kow are out of domain. Moreover, similarity of structures used for prediction is low.

Vehicle controls validity:

not applicable

Negative controls validity:

not applicable

Positive controls validity:

not applicable

Remarks on result:

other: Non-Sensitiser

Remarks:

Moderate confidence: Due to an intermediate probability, low similarity between structures in the training set, low concordance of actual results of similar structures in the training set there is only moderate confidence in the prediction. However, the prediction may be supported by significant Bayesian statistics and enrichment value.

Vehicle controls validity:

not applicable

Negative controls validity:

not applicable

Positive controls validity:

not applicable

Remarks on result:

other: Sensitiser

Remarks:

Out of AD The prediction is considered low reliable. CAESAR's internal AD index indicates that the query compound is out of the models applicability domain.

Vehicle controls validity:

not applicable

Negative controls validity:

not applicable

Positive controls validity:

not applicable

Remarks on result:

other: Sensitiser

Remarks:

Equivocal result, low confidence

Other effects / acceptance of results:

The query molecule has two structural features to make assumption for the skin sensitization potential: the quaternary ammonium fragment and the double bond with neighbouring electron-withdrawing group favourable for Michael addition. Therefore, the majority of models predict the structure as sensitizer. Concerning the applicability of the models for the given structure, the reliability is low.

With OECD Toolbox the query compound was predicted to be a sensitiser. However, the query compound is out of applicability domain since all 6 category members possess obviously lower molecular weight, which is one of determining parameter of the applicability domain in the OECD Toolbox assessment. In addition, this molecular property is especially important for the description of the dermal absorption rate. A reduced dermal absorption may be assumed for chemicals with a molecular weight > 500 and log Kow smaller than 0 or higher than 4i. The query compound has a molecular weight of 400 and the log Kow is below 0 which are not very favourable for the dermal absorption.

The DEREK Nexus prediction for skin sensitization submits an equivocal reasoning, which is not conclusive for the inherent property of the query compound. The alert is related with the quaternary ammonium salt feature of the query compound. Nevertheless, the prediction based on expert rules highlights that lipophilicity and low molecular size are the conditions for the likelihood to cause sensitization. Regarding the water solubility, the compound is highly water soluble (above 10’000 mg/L); therefore with a Log Kow below 0 the substance may be too hydrophilic to cross the lipid rich environment of the stratum corneumi. In addition, it is known that ionized species do not penetrate the skin very well. As example, quaternary ammonium ion may bind to skin components slowing the uptake of the query compound.

TOPKAT has predicted the query compound as non-sensitizer. The prediction is characterised by some uncertainties in the prediction statistics and low concordance between actual results of similar structures and the query compound. However, the query compound is within applicability domain and all structural features were found in the training set. Therefore, the TOPKAT result may be regarded as moderate reliable.

VEGA (CAESAR) has predicted the query compound as sensitizer. The model reports moderate similarity of the compounds in the training set as well as 1 unknown fragment and 3 infrequent fragments in the compounds of the training set. In addition, the predicted compound is outside the applicability domain of the model, which suggests a rather inconclusive prediction. The result is therefore regarded as not reliable.

Interpretation of results:

other: Classification based on weight of evidence approach.

Conclusions:

In conclusion, the discussed properties of the query compound suggest that the dermal absorption is likely to be low. In addition, there are no local effects (no skin and eye irritation) which may cause an increased absorption throughout the skin. According to the acute dermal assessment in rabbit, no clinical signs of intoxication were observed after 5 mL/kg dossing. This may suggest a low toxicity of the query compound and/or low systemic uptake.
The absence of sensitizing properties of the query compound may be supported by the repeat insult patch-test conducted with different formulations containing up to 4% SunBeat, where no dermal irritation or sensitization was observed.

Executive summary:

(Q)SAR predictions of skin sensitisation were calculated for Galaxy SunBeat using OECD QSAR Toolbox 3.3.3, Biovia Discovery Studio (TOPKAT) 4.5 (extensible), VEGA NIC 1.1.1, and DEREK Nexus 3.0.1. The results are summarised in the following table.

 

OECD Toolbox	TOPKAT	VEGA	DEREK
Sensitiser	Non-sensitiser	Sensitiser	Sensitiser

Reference 3

Endpoint:

skin sensitisation, other

Remarks:

in silico

Type of information:

(Q)SAR

Adequacy of study:

weight of evidence

Study period:

29 February 2016

Reliability:

2 (reliable with restrictions)

Rationale for reliability incl. deficiencies:

results derived from a valid (Q)SAR model, but not (completely) falling into its applicability domain, with adequate and reliable documentation / justification

Justification for type of information:

1. SOFTWARE
OECD QSAR Toolbox

2. MODEL (incl. version number)
3.3.3

3. SMILES OR OTHER IDENTIFIERS USED AS INPUT FOR THE MODEL
CN{+}(C)(CCCNC(=O)C=Cc1ccc(OC)cc1)CC(O)CS(=O)(=O)O{-}

4. SCIENTIFIC VALIDITY OF THE (Q)SAR MODEL
[Explain how the model fulfils the OECD principles for (Q)SAR model validation. Consider attaching the QMRF or providing a link]
- Defined endpoint: Skin Sensitisation; EC3; S M W N Materials & methods / type of method: in vivo
- Unambiguous algorithm: Read-across from category members Experimental values for the target chemical (if any) were not used in prediction calculations.
- Defined domain of applicability: As described in the attached document.
- Appropriate measures of goodness-of-fit and robustness and predictivity: The prediction is based on 5 neighbours' values, 4 of them equal to prediction. Prediction confidence is measured by the p-value = 0.188 (no confidence)
- Mechanistic interpretation: Chemical and biological mechanisms (OECD Principle 5)

5. APPLICABILITY DOMAIN
- Descriptor domain: Applicability domain (OECD Principle 3): The target chemical DOES NOT FALL within applicability domain
- Structural and mechanistic domains: The query compound is out of the applicability domain since its molecular weight is greater than of the neighbours used for prediction. In addition, the log Kow of the query compound is lower than of the neighbours. Overall, there is low confidence in the prediction.

6. ADEQUACY OF THE RESULT
Protein binding alerts for skin sensitisation may indicate a potential for skin sensitisation. However, the prediction is not reliable and predictions with other QSAR are recommended.

Reason / purpose for cross-reference:

(Q)SAR model reporting (QMRF)

Guideline:

other:

Version / remarks:

REACH Guidance on QSARs R.6

Guideline:

other:

Version / remarks:

OECD Guidance Document on the Validation of (Quantitive) Structure-Activity Relationships Models

Principles of method if other than guideline:

Toxicity of the target chemical (Positive) is predicted from category members using read-across based on 5 values (Positive x4, Negative x1) from 5 nearest neighbours compared by prediction descriptors. Category members are single chemicals or mixtures and are selected based on the profile of the target chemical. Only chemicals having experimental data are listed in the category.

GLP compliance:

not specified

Vehicle controls validity:

not applicable

Negative controls validity:

not applicable

Positive controls validity:

not applicable

Remarks on result:

other: Protein binding alerts for skin sensitisation may indicate a potential for skin sensitisation. However, the prediction is not reliable and predictions with other QSAR are recommended.

Other effects / acceptance of results:

The query compound is out of the applicability domain since its molecular weight is greater than of the neighbours used for prediction. In addition, the log Kow of the query compound is lower than of the neighbours. Overall, there is low confidence in the prediction.

Interpretation of results:

other: Classification based on weight of evidence approach.

Conclusions:

The query compound is out of the applicability domain since its molecular weight is greater than of the neighbours used for prediction. In addition, the log Kow of the query compound is lower than of the neighbours. Overall, there is low confidence in the prediction.

Protein binding alerts for skin sensitisation may indicate a potential for skin sensitisation. However, the prediction is not reliable and predictions with other QSAR are recommended.

Reference 4

Endpoint:

skin sensitisation, other

Remarks:

in silico

Type of information:

(Q)SAR

Adequacy of study:

weight of evidence

Study period:

18 January 2017

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 limited documentation / justification

Justification for type of information:

1. SOFTWARE
Biovia Discovery Studio (TOPKAT) 4.5 (extensible)

2. MODEL (incl. version number)
4.5

3. SMILES OR OTHER IDENTIFIERS USED AS INPUT FOR THE MODEL
SMILES C1=CC(=CC=C1C=CC(NCCC[N+](CC(O)CS(=O)(=O)[O-])(C)C)=O)OC
InChI 1/C18H28N2O6S/c1-20(2,13-16(21)14-27(23,24)25)12-4-11-19-18(22)10-7-15-5-8-17(26-3)9-6-15/h5-10,16,21H,4,11-14H2,1-3H3,(H-,19,22,23,24,25)/b10-7+
Stereochemical features: Four stereoisomers

4. SCIENTIFIC VALIDITY OF THE (Q)SAR MODEL
- Defined endpoint: Skin Sensitisation (None vs Sensitiser), Classification as sensitiser or non-sensitiser
- Unambiguous algorithm: Toxicity Prediction (Extensible) Toxicity Prediction (Extensible) Skin sensitisation (None vs Sensitiser)
- Defined domain of applicability:
i. All properties and OPS components are within expected ranges
ii. All fingerprint features of the query molecule are found in the training set
iii. Considerations on the mechanism domain are not applicable since the contributing FCFP_12 features of the model are selected purely on their Bayesian score (statistical model)
- Appropriate measures of goodness-of-fit and robustness and predictivity: Uncertainty may be indicated due to an intermediate probability, low similarity of analogue structures, and low concordance between actual results of the similar structures and the predicted value of the query structure.
- Mechanistic interpretation: Not applicable since statistical model

5. APPLICABILITY DOMAIN
- Descriptor domain:
i. All properties and OPS components are within expected ranges
ii. All fingerprint features of the query molecule are found in the training set
- Similarity with analogues in the training set:
i. Aspartame
ii. Trimethylammonium-ethyloxycarboxybenzenesulfonate
iii. Trimethylammonium-hexanoyloxy-4-benzenesulfonate
iv. Triethyl citrate
With 16% the average similarity of the four analogues to the query structure is considered low. Three out of four structures are sensitiser thus indicating low concordance with the query structure. Accuracy between predicted and actual result is high as all are predicted correctly.

6. ADEQUACY OF THE RESULT
Result is directly applicable since no conversion of the result is required.

Reason / purpose for cross-reference:

(Q)SAR model reporting (QMRF)

Guideline:

other:

Version / remarks:

REACH Guidance on QSARs R.6

Principles of method if other than guideline:

For the prediction of skin sensitisation of Galaxy SunBeat the four models OECD QSAR Toolbox 3.3.5, Biovia Discovery Studio (TOPKAT) 4.5 (extensible), within VEGA NIC 1.1.1 and DEREK were applied in order to provide a weight-of-evidence approach.

GLP compliance:

not specified

Vehicle controls validity:

not applicable

Negative controls validity:

not applicable

Positive controls validity:

not applicable

Remarks on result:

no indication of skin sensitisation

Other effects / acceptance of results:

Approach for regulatory interpretation of the model result
Result is directly applicable since no conversion of the result is required.

Outcome
Non-Sensitiser. Due to an intermediate probability and low similarity of and concordance with analogues there is only moderate confidence in the prediction. However, the prediction may be supported by significant Bayesian statistics and enrichment value.

Interpretation of results:

other: Classification based on weight of evidence approach.

Conclusions:

There are some indications that the query structure is not sensitising. However, more evidence from e.g. other QSARs and mechanistic interpretations of the structure may additionally be considered.

Reference 5

Endpoint:

skin sensitisation, other

Remarks:

in silico

Type of information:

(Q)SAR

Adequacy of study:

weight of evidence

Study period:

18 January 2017

Reliability:

2 (reliable with restrictions)

Rationale for reliability incl. deficiencies:

results derived from a valid (Q)SAR model, but not (completely) falling into its applicability domain, with adequate and reliable documentation / justification

Justification for type of information:

1. SOFTWARE
Skin Sensitization model (CAESAR) 2.1.6 within VEGA

2. MODEL (incl. version number)
2.1.6

3. SMILES OR OTHER IDENTIFIERS USED AS INPUT FOR THE MODEL
SMILES C1=CC(=CC=C1C=CC(NCCC[N+](CC(O)CS(=O)(=O)[O-])(C)C)=O)OC
InChI 1/C18H28N2O6S/c1-20(2,13-16(21)14-27(23,24)25)12-4-11-19-18(22)10-7-15-5-8-17(26-3)9-6-15/h5-10,16,21H,4,11-14H2,1-3H3,(H-,19,22,23,24,25)/b10-7+
Stereochemical features: Four stereoisomers

4. SCIENTIFIC VALIDITY OF THE (Q)SAR MODEL
- Defined endpoint: Skin Sensitisation (None vs Sensitiser), Classification as sensitiser or non-sensitiser
- Unambiguous algorithm: There is no QMRF available. Instead it is referred to the models user guide attached in the Expert Statement.
- Defined domain of applicability:
i. The predicted compound is outside the Applicability Domain of the model.
ii. A prominent number of atom centered fragments of the compound have not been found in the compounds of the training set or are rare fragments (1 unknown fragments and 3 infrequent fragments found).
iii. Considerations on the mechanism domain are not applicable since statistical model.
- Appropriate measures of goodness-of-fit and robustness and predictivity: Uncertainty may be indicated due to a prominent number of fragments of the query compound not found in the training set and only moderate similarity of structures in the training set. CAESAR's internal AD index indicates that the query compound is out of the models applicability domain.
- Mechanistic interpretation: Not applicable since statistical model

5. APPLICABILITY DOMAIN
- Descriptor domain: Not provided by the software
- Similarity with analogues in the training set:
i. Hexanoic acid, 3,5,5-trimethyl-, 4-sulfophenyl ester, sodium salt
ii. Benzylpenicillin [INN:BAN]
iii. 2-[(2-methyloctadecanoyl)oxy]benzene-1-sulfonic acid
iv. Lauryl gallate
Uncertainty may be indicated due to a prominent number of fragments of the query compound not found in the training set and only moderate similarity of structures in the training set. CAESAR's internal AD index indicates that the query compound is out of the models applicability domain.

6. ADEQUACY OF THE RESULT
Result is directly applicable since no conversion of the result is required.

Reason / purpose for cross-reference:

(Q)SAR model reporting (QMRF)

Guideline:

other:

Version / remarks:

REACH Guidance on QSARs R.6

Principles of method if other than guideline:

For the prediction of skin sensitisation of Galaxy SunBeat the four models OECD QSAR Toolbox 3.3.5, Biovia Discovery Studio (TOPKAT) 4.5 (extensible), within VEGA NIC 1.1.1 and DEREK were applied in order to provide a weight-of-evidence approach.

GLP compliance:

not specified

Vehicle controls validity:

not applicable

Negative controls validity:

not applicable

Positive controls validity:

not applicable

Remarks on result:

other: Sensitiser. There is only low confidence in the prediction as the query compound is out the models applicability domain. The prediction is considered low reliable.

Interpretation of results:

other: Classification based on weight of evidence approach.

Conclusions:

The substance was considered to be a sensitiser however there is only low confidence in the prediction as the query compound is out the models applicability domain. As a result of this the prediction is considered low reliable.

Endpoint conclusion

Endpoint conclusion:

no adverse effect observed (not sensitising)

Additional information:

The query molecule has two structural features to make assumption for the skin sensitization potential: the quaternary ammonium fragment and the double bond with neighbouring electron-withdrawing group favourable for Michael addition. Therefore, the majority of

models predict the structure as sensitizer. Concerning the applicability of the models for the given structure, the reliability is low.

With OECD Toolbox the query compound was predicted to be a sensitiser. However, the query compound is out of applicability domain since all 6 category members possess obviously lower molecular weight, which is one of determining parameter of the applicability

domain in the OECD Toolbox assessment. In addition, this molecular property is especially important for the description of the dermal absorption rate. A reduced dermal absorption may be assumed for chemicals with a molecular weight > 500 and log Kow smaller than 0 or higher than 4. The query compound has a molecular weight of 400 and the log Kow is below 0 which are not very favourable for the dermal absorption.

The DEREK Nexus prediction for skin sensitization submits an equivocal reasoning, which is not conclusive for the inherent property of the query compound. The alert is related with the quaternary ammonium salt feature of the query compound. Nevertheless, the prediction based on expert rules highlights that lipophilicity and low molecular size are the conditions for the likelihood to cause sensitization. Regarding the water solubility, the compound is highly water soluble (above 10’000 mg/L); therefore with a Log Kow below 0 the substance may be too hydrophilic to cross the lipid rich environment of the stratum corneumi. In addition, it is known that ionized species do not penetrate the skin very well. As example, quaternary ammonium ion may bind to skin components slowing the uptake of the query compound.

TOPKAT has predicted the query compound as non-sensitizer. The prediction is characterised by some uncertainties in the prediction statistics and low concordance between actual results of similar structures and the query compound. However, the query compound is within applicability domain and all structural features were found in the training set. Therefore, the TOPKAT result may be regarded as moderate reliable.

VEGA (CAESAR) has predicted the query compound as sensitizer. The model reports moderate similarity of the compounds in the training set as well as 1 unknown fragment and 3 infrequent fragments in the compounds of the training set. In addition, the predicted

compound is outside the applicability domain of the model, which suggests a rather inconclusive prediction. The result is therefore regarded as not reliable.

Respiratory sensitisation

Endpoint conclusion

Endpoint conclusion:

no study available

Justification for classification or non-classification

In conclusion, the discussed properties of the query compound suggest that the dermal absorption is likely to be low. In addition, there are no local effects (no skin and eye irritation) which may cause an increased absorption throughout the skin. According to the acute dermal assessment in rabbit, no clinical signs of intoxication were observed after 5mL/kg dosing. This may suggest a low toxicity of the query compound and/or low systemic uptake.

The absence of sensitizing properties of the query compound may be supported by the repeat insult patch-test conducted with different formulations containing up to 4% SunBeat, where no dermal irritation or sensitization was observed.