采用贝叶斯网络法评价化妆品原料眼刺激性

采用贝叶斯网络(BN)替代评价方法将非试验数据和替代试验数据组合形成评估框架,用于化妆品原料的眼刺激性评估和分级。基于综合测试策略(ITS),在替代试验的基础上增加了非试验数据,包括构效关系和理化性质等。通过搜集104种化妆品原料的数据,建立了化妆品原料眼刺激性评价的贝叶斯网络,并进行验证试验。将5种化妆品原料分别进行替代试验和家兔眼刺激试验,通过将非试验数据和替代试验数据输入贝叶斯网络,得到分级预测结果,并与家兔眼刺激试验结果进行比较。结果表明,所建立的基于贝叶斯网络的替代评价方法能够有效整合非试验数据和体外试验数据,有助于提高评价的可靠性。

A safety evaluation method of eye irritation evaluation of cosmetic ingredients based on Bayesian Network

This study applied a Bayesian Network (BN) to predict eye irritation of cosmetic ingredients. It bases on the Integrated Testing Strategy (ITS) and combines evidence from different information sources such as QSAR model predictions, in vitro alternative method results and in vivo test results in order to increase the reliability of a conclusion on the eye irritation of cosmetic ingredients. The parameters of BN were calculated from an extended data set of 104 cosmetic ingredients. A detailed validation analysis and a case study of five cosmetic ingredients were carried out to demonstrate the utility of the BN for practical application. These results suggest that BN displays good predictive capacity in eye irritation of cosmetic ingredients and is of well application value as a prediction model in cosmetic safety evaluation.