Expert Review of In Silico Predictions for Skin Sensitization Potency
Poster for the Tenth World Congress Alternatives and Animal Use in the Life Science (WC10)
Skin sensitization thresholds and potency classification is typically determined based on data from the local lymph node assay (LLNA). Such data are not available for many existing chemicals that may require evaluation within restrictive budgetary and time constraints. For these situations, alternative methods such as predictive in silico programs may be considered.
However, in silico estimates often require an additional level of expert scrutiny. We evaluated the impact of applying expert review to the predicted skin sensitization thresholds (EC3 values) from the in silico program Derek Nexus for a small set of 20 well-known dermal sensitizers. We then compared the predicted EC3 values against potency classifications determined based on readily available guinea pig or human data.
We found the Derek Nexus potency estimates were concordant with study data in 35% of cases (7 out of 20 chemicals). In 15% of cases (3 out of 20 chemicals), in silico estimates were health-protective (i.e., potentially overestimating hazard). However, in the remaining 50% of cases, in silico estimates were judged unreliable due to poor surrogate availability; these were split evenly between weak and moderate sensitizers as determined by actual studies. Our results highlight the importance of expert review in the interpretation and application of in silico estimates for skin sensitization potency.