构建基于GHS标准的黑头呆鱼(Pimephales promelas)急性毒性二元分类模型
Development of GHS-based Binary Classification Models for Predicting Acute Toxicity of Fathead Minnow (Pimephales promelas)
-
摘要: 鱼类急性毒性参数是进行化学品生态风险评估、分类标签等工作不可或缺的毒性指标。本文选取634个有机化学品对黑头呆鱼(Pimephales promelas)的急性毒性数据,并依据"全球化学品统一分类和标签制度"(GHS)中推荐的分类标准,将急性毒性值小于和大于100 mg·L-1的物质分别划分为有毒物质和无毒物质。以分类结果为建模指标,构建了基于欧几里德距离的K最近邻(kNN)二元分类模型。评估结果表明,模型训练集和验证集的预测准确度(Q)、敏感性(Sn)和特异性(Sp)参数均大于0.7,说明模型具有较好的预测能力。因而,在化学品分类标签工作中,可使用该模型预测缺失的鱼类急性毒性类别。
-
关键词:
- 黑头呆鱼 /
- 急性毒性 /
- kNN /
- 欧几里德距离 /
- 全球化学品统一分类和标签制度
Abstract: The acute toxicity data of aquatic organisms are indispensable parameters in the ecological risk assessment, chemical classification and labelling. In the present study, the acute toxicity data of fathead minnow (Pimephales promelas) for 634 organic chemicals was collected. Then, the model compounds with their acute toxicity data ≤ 100 mg·L-1 and > 100 mg·L-1 were classified as toxic and non-toxic, respectively, according to the classification criteria recommend in globally harmonized system of classification and labelling of chemicals (GHS). Then, binary classification models were developed by using Euclidean distances-based k-nearest neighbor method (kNN). The predictive accuracy (Q), sensitivity (Sn) and specificity (Sp) values for the training set and validation set was > 0.7, indicating the obtained optimum models had high predictive ability. Thus, the missing data gap for acute toxicity data of fish could be filled by employing the model developed here in classification and labelling chemicals. -
-
点击查看大图
计量
- 文章访问数: 2077
- HTML全文浏览数: 2077
- PDF下载数: 17
- 施引文献: 0
下载:
