1. Quantitative modelling of human potency
D W Roberts* A M Api and T W Schultz
With acknowledgement to Nora Aptula

2. How well does LLNA model human potency?
Api, Lalko and Basketter, Correlation between experimental human and murine skin sensitization induction thresholds Cutaneous and Ocular Toxicology 34 (4)
Overall “good agreement” between EC3 and NOEL, with some outliers
Group 1 – “good agreement” – 42 cases
Group 2 – LLNA underpredicts NOEL – 7 cases
Group 3 – LLNA overpredicts NOEL – 7 cases

3. For regression analysis
Group 1 (good agreement)
Remove cases where EC3 or NOEL is given as <x or >x
Remove cases footnoted “No sensitization was observed in human predictive studies. Doses reported reflect the highest concentration tested, not necessarily the highest achievable NOEL”
This leaves 31 cases

4. logNOEL = 1.04(0.10)logEC3 – 0.09((0.34)
Regression analysis
logNOEL = 1.04(0.10)logEC3 – 0.09((0.34)
n = 31, R2 = 0.786, AdjR2 = 0.779, s = 0.24, F = 107

5. Interpretation of regression equation
logNOEL = 1.04(0.10)logEC3 – 0.09((0.34)
n = 31, R2 = 0.786, AdjR2 = 0.779, s = 0.24, F = 107
s (standard deviation of residuals) = 0.24 corresponds to 95% confidence limits of a factor of 3 on the NOEL as predicted from the EC3
Slope and intercept not significantly different from 1 and 0 respectively, i.e.….
EC3 can be used directly to define NOEL…
…If outliers can be recognised

6. Can we model human potency directly from chemistry?
20 Aliphatic aldehydes* with human NOEL data:
a-Methyl-phenylacetaldehyde; 1,2,3,4,5,6,7,8-Octahydro-8,8-dimethyl-2-naphthaldehyde; Phenylacetaldehyde; Citral; Cuminyl acetaldehyde; Bourgeonal; p-Methylhydrocinnamic aldehyde; p-Isobutyl-a-methyl hydrocinnamaldehyde; Hydroxycitronellal; Lilial; Landolal (Lyral); Methoxy dicyclopentadiene carboxaldehyde; Triplal; 2-Methyl-3-(p-methoxyphenyl)propanal; Cyclamen aldehyde; Heptanal, 6-methoxy-2,6-dimethyl-; 3-Phenylbutanal; Citronellal; Isocyclocitral; a-Methyl-1,3-benzodioxole-5-propionaldehyde
From Basketter et al., . (2014) Categorization of chemicals according to their relative human skin sensitizing potency. Dermatitis 25 (1), and IFRA IFRA standard 48th amendment
* “Aliphatic aldehydes” are defined as those not having an aromatic carbon bonded directly to the C=O group

7. Modelling parameters
Reactivity Taft substituent constants for groups bonded to carbonyl, Ss* Hydrophobicity Calculated (Leo and Hansch method) logP (octanol/water)

8. QMM plot
pNOEL = 2.34 Ss* logP
Outlier

9. QMM equation
pNOEL = 2.34(±0.33) Ss* (±0.07) logP (±0.22) n = 19, R2 = 0.770, R2(adj) = 0.741, s = 0.20, F = 27 Error limits (from s value) correspond to a factor of <2.2 between observed and calculated NOEL levels – similar to variability of LLNA Except for one outlier, ca 8 times as potent as calculated…

10. Outlier has 6 allylic Hydrogens able to give tert-allylic hydroperoxides

11. Conclusions LLNA EC3 predicts NOEL directly for most chemicals
Underpredictions of potency can be attributed to and anticipated for:
Aromatic Schiff base electrophiles
Chemicals likely to contain impurities/by-products from synthesis
Pro-/pre-haptens with complex activation pathways
Overpredictions of potency can be attributed to and anticipated for:
Chemicals readily susceptible to autoxidation under LLNA conditions
Physical-organic chemistry principles underlying LLNA potency also apply to human potency
Other reaction mechanistic domains need to investigated similarly

13. Outliers: potency underpredicted by LLNA
Chemical
EC3 (mg/cm2)
NOEL (mg/cm2)
Benzaldehyde
> 6250
590
Vanillin
> 1250
1100
Trans-2-Hexenal
1012 24
6-Methyl-3,5-heptadiene-2-one
1250
110
2-Methoxy-4-methylphenol
1450
118
Methyl 2-nonynoate
625
Treemoss absolute
> 5000
700
Treemoss absolute not considered further – potency variable depending on composition, particularly atranol and chloratranol content

14. Underpredicted – benzaldehyde and vanillin
Schiff base electrophiles, aromatic Most aromatic aldehydes are weak or NS in LLNA, weaker than predicted by the QMM for SB: pEC3 = 1.12 Ss* logP – 0.62 (Roberts et al 2007), developed from data on aliphatics

15. QMM prediction for benzaldehyde and vanillin
Although aromatic aldehydes are outside the applicability domain…
The QMM predicts (assuming NOEL = predicted EC3):
Benzaldehyde NOEL (actual 590) – factor of 1.8
Vanillin NOEL (actual 1181) – factor of 3.3
Within/close to 95% confidence limits of logNOEL vs logEC3 regression

16. Underpredicted potency: trans-2-hexenal and 6-methyl-3,5-heptadienal
Michael acceptors, volatile but NOEL potency < predicted from Michael acceptor QMM, so…
volatility alone cannot explain the large underpredictions
Impurities (eg from aldol dimerisation) in samples tested in HRIPT may be responsible

17. Underpredicted potency – 2-methoxy-4-methylphenol
Pro- or pre-electrophile, activated by oxidation, either after metabolic demethylation or directly to quinone methide. Free radical mechanisms also possible.
Variety of possible mechanisms makes inter-species variation more likely

18. Underpredicted potency – 2-methoxy-4-methylphenol
Pro- or pre-electrophile, activated by oxidation, either after metabolic demethylation or directly to quinone methide. Free radical mechanisms also possible.
Variety of possible mechanisms makes inter-species variation more likely

19. Underpredicted potency – methyl 2-nonynoate
EC3
NOEL
MA QMM prediction
Me 2-nonynoate
625 24
450
Me 2-octynoate
<125
118
412
Simplest interpretation:
The EC3 value of 625 for methyl 2-nonynoate is correct
The NOEL value of 24 for methyl 2-nonynoate is anomalous
The EC3 value of <125 for methyl 2-octynoate is anomalous
The NOEL of 118 for methyl 2-octynoate is anomalous
This pattern suggests that the recorded potency values are influenced by potent impurities present in the samples tested, except for methyl 2-nonynoate (LLNA) which must have contained only insignificant levels.
20th century literature says that potency of these –ynoates is low when freshly synthesised but increases with age (English and Rycroft 1988) – consistent with the impurity interpretation
QMM: pEC3 (mol%) = 0.24logk Roberts, D.W., Natsch, A. (2009). High throughput kinetic profiling approach for covalent binding to peptides: application to skin sensitization potency of Michael acceptor electrophiles. Chemical Research in Toxicology 22,
English JS and Rycroft RJ Allergic contact Dermatitis from methyl octine and methyl heptine carbonates. Contact Dermtitis 18:

20. Outliers: potency overpredicted by LLNA
Chemical
EC3 (mg/cm2)
NOEL (mg/cm2)
a-Amyl cinnamal
2420
23600
a-Hexyl cinnamal
2372
Benzyl salicylate
725
17700
Hexyl salicylate 45
35400
Isocyclocitral
1825
7000
a-iso-Methylionone
5450
70000
OTNE
6825
47200
Benzyl and hexyl salicylate EC3 values are anomalous compared to other salicylates (weak or NS). By-products from synthesis suspected.
The other 5 may be explained by autoxidation being favoured under LLNA open application conditions

21. Overpredicted by LLNA – Amyl- and hexyl-cinnamal
Extent of this reaction, and hence degree of sensitization, depends on accessibility to oxygen

22. Overpredicted by LLNA – hydroperoxide precursors

23. Conclusions LLNA EC3 predicts NOEL directly for most chemicals
Underpredictions of potency can be attributed to and anticipated for:
Aromatic Schiff base electrophiles
Chemicals likely to contain impurities/by-products from synthesis
Pro-/pre-haptens with complex activation pathways
Overpredictions of potency can be attributed to and anticipated for:
Chemicals readily susceptible to autoxidation under LLNA conditions
(Consistent with earlier findings for “LLNA false positives”)