发育毒物体外筛选试验的聚类分析

Cluster analysis of data obtained from in vitro tests for screening developmental toxicants

以简单配比系数为相似性指标对１５种发育毒物体外筛选试验进行指标型系统聚类分析。结果提示最长、最短和平均距离法３种聚类方法中以最长距离法的聚类效果最为理想，符合发育毒物毒作用的细胞学机制。最终聚类结果为５大类，其中以Ａ／Ｄ比值和以细胞生长为指标的方法聚为一类；以细胞聚集、细胞生长和细胞分化３个终点为联合指标的鸡胚神经视网膜细胞试验和啮齿类动物的全胚胎培养归为一类。但以反映细胞分化和细胞间相互作用以及胚胎形成抑制为终点的方法并未完全归为一类，说明这些单一的终点不可能全面反映细胞增殖、分化和相互作用等相互依赖的细胞行为，因而提示应用其中的两个终点进行组合并不能提高预测价值。为了全面提高发育毒物体外筛选试验的预测价值，应结合各种方法的特异度和灵敏度，并根据经济有效和合理的原则。

The simple matching coefficients(SMCs) was applied as the index of similarity in hierarchical cluster analysis of fifteen in vitro prescreening tests for developmental toxicants.The classification results of complete link method,single link method, and average link method indicated that the result based on the first method was optimal.The 5 groups were formed,which shown that the methods with A/D ratio or cell growth inhibition as endpoints were put into the same groups,respectively,and that the method with combination of cell aggregation,cell growth and cell differentiation as endpoints was clustered into one class with rodents whole embryo culture,but the methods with cell differentiation and cell cell interaction were not classified into one cluster,which verified the theory that three cell behaviors(cell proliferation,cell differentiation and cell cell/cell substrate interactions)are dependant.Therefore,when battery of methods were made,the combination with only two of three endpoints should not be sufficient to improve prediction values of tests.In order to improve prediction values significantly,three or more endpoints should be selected to combine to a test battery according on the principle of low cost, effectiveness and rationale. \ \