一种新型in silico模型在化妆品原料皮肤致敏性初筛中的应用

Application of New in silico Model in Initial Screening of Skin Sensitizing Potential of Cosmetic Ingredients

用Derek EC3模型对OECD TG429和TG442C中32个验证化合物、114个外部数据集中已有致敏结果和120个已有小鼠局部淋巴结试验(LLNA)和直接肽反应(DPRA)试验数据的化妆品原料化合物进行致敏性评价,通过比较Derek结果与致敏性结果(LLNA或DPRA)的一致性对Derek模型的适用性进行评价。对于LLNA试验的标准验证物质,Derek方法的预测特异度为50%(3/6),灵敏度为81.25%(13/16),准确度为72.73%(16/22);对于DPRA试验的标准验证物质,Derek方法的预测特异度为75%(3/4),灵敏度为100%(6/6),准确度为90%(9/10);Derek方法对109个外部数据集中化合物致敏性预测特异度为79.01%(64/81),灵敏度为71.43%(20/28),准确度为77.06%(84/109);与LLNA比较,120个化合物致敏性评价结果中Derek和DPRA方法特征参数分别为特异度60.71%/74.07%、灵敏度84.27%/67.42%、准确度78.63%/68.97%,组合这两种方法特异度为51.72%(15/29),灵敏度为90.11%(82/91),准确度为80.83%(97/120)。结果表明,Derek EC3模型适用于化妆品原料的皮肤致敏性初筛。

The skin sensitizing potential of given chemicals,including 22 validated compounds in OECD TG429,10 validated compounds in OECD TG442C,114 cosmetic ingredients from external data set and 120 compounds with LLNA/DPRA results that already had relevant allergenicity data were predicted using the Derek EC3 model.Then,the accuracy of the Derek model was assessed by the result consistency with the sensitive results(provided by LLNA or DPRA methods).Compared to LLNA,Derek model was able to predict the skin sensitization of 22 validated compounds with a method specificity of 50%(3/6),a sensitivity of 81.25%(13/16),and an accuracy of 72.73%(16/22);Compared with DPRA,Derek model can predict the skin sensitization of 10 validated compounds with a method specificity of 75%(3/4)and a sensitivity of 100%(6/6).The accuracy was 90%(9/10).For 109 chemicals in external data set,Derek model can predict compounds with a specificity of 79.01%(64/81),a sensitivity of 71.43%(20/28)and its accuracy was 77.06%(84/109).Among the 120 compounds,Derek model can predict 117 compounds with a specificity of 60.71%,a sensitivity of 84.27%and its accuracy was 78.63%while DPRA can predict compounds with a specificity of 74.07%,a sensitivity of 67.42%and its accuracy was 68.97%.And combination of these two methods can get a specificity of 51.72%(15/29),a sensitivity of 90.11%(82/91)and improve the accuracy to 80.83%(97/120).Results showed that the Derek EC3 model is suitable for predicting the skin sensitizing potential of cosmetic ingredients.