›› 2010, Vol. 53 ›› Issue (12): 1436-1441.

• 研究论文 • 上一篇    下一篇


王志明, 谭显胜, 周玮, 袁哲明   

  • 出版日期:2011-01-17 发布日期:2010-12-20
  • 通讯作者: 袁哲明

Bioassay data analysis based on support vector regression

WANG Zhi-Ming, TAN Xian-Sheng, ZHOU Wei, YUAN Zhe-Ming   

  • Online:2011-01-17 Published:2010-12-20
  • Contact: YUAN Zhe-Ming


生物测定是生物学、 医学、 毒理学的重要内容与基础。常用的定量生物测定数据分析方法时间-剂量-死亡率模型(TDM)不能对复杂生测数据建立统一模型, 信息利用不充分。本文基于支持向量回归(SVR), 提出了一种能对不同供试因子、 不同供试对象和不同环境条件下复杂生测数据统一建模的新方法。14个简单生测数据和2套复杂生测数据的对比分析结果表明, SVR模型拟合与留一法预测精度均优于TDM模型, 估计的LD50LT50等指标更为可信。SVR模型有望作为TDM模型的有益补充, 在定量生物测定数据分析中得到广泛应用。

关键词: 时间-剂量-死亡率模型, 互补重对数模型, 支持向量回归, 留一法, 生物测定


Bioassay plays an important role in the studies of biology, medicine and toxicology. The time-dose-mortality model (TDM) widely applied to quantitative bioassay data analysis can not construct a unified model for complex bioassay data, and has the disadvantage of utilizing the information incompletely. Based on support vector regression (SVR), a novel quantitative bioassay model has been developed, which can construct a unified model for complex data with different test factors, different test objects and different environment factors. We compared the prediction performance between SVR and TDM using 14 simple data and 2 complex data. The results showed that SVR achieved better precision than TDM not only in self-consistency test but also in jackknife test, implying that the estimated values of LD50 and LT50 by the former are more reliable. As a useful supplement to TDM, SVR has the potential to be widely used for quantitative bioassay data analysis.

Key words: Time-dose-mortality model (TDM), complementary log-log model (CLL), support vector regression, Leave-One-Out method, bioassay