摘要
Since it is unrealistic to do an experimental mixture assessment on every possible combination, mathematical model plays an important role in predicting the mixture toxicity. The present study is devoted to the further application of linear concentration addition（CA）-based model（LCA） and independent action（IA）-based model（LIA） to predict the non-interactive mixture toxicity. The 26 mixtures including 312 data points were used to evaluate the predictive powers of LCA and LIA models. The models were internally validated using the leave-one-out cross-validation and y-randomization test, and the external validations were evaluated by the test tests. Both LCA and LIA models agree well with the experimental values for all mixture toxicity, and present high internally（R2 and Q2 〉 0.98） and externally（Q2F1, Q2F2, and Q2F3 〉 0.99） predictive power. The use of LCA and LIA led to improved predictions compared to the estimates based on the CA and IA models. Both LCA and LIA were found to be appropriate methods for modeling toxicity of non-interactive chemical mixtures.
Since it is unrealistic to do an experimental mixture assessment on every possible combination, mathematical model plays an important role in predicting the mixture toxicity. The present study is devoted to the further application of linear concentration addition（CA）-based model（LCA） and independent action（IA）-based model（LIA） to predict the non-interactive mixture toxicity. The 26 mixtures including 312 data points were used to evaluate the predictive powers of LCA and LIA models. The models were internally validated using the leave-one-out cross-validation and y-randomization test, and the external validations were evaluated by the test tests. Both LCA and LIA models agree well with the experimental values for all mixture toxicity, and present high internally（R2 and Q2 〉 0.98） and externally（Q2F1, Q2F2, and Q2F3 〉 0.99） predictive power. The use of LCA and LIA led to improved predictions compared to the estimates based on the CA and IA models. Both LCA and LIA were found to be appropriate methods for modeling toxicity of non-interactive chemical mixtures.
基金
supported by the National Natural Science Foundation of China(21407032,21667013,51578171)
Natural Science Foundation of Guangxi Province(2014GXNSFBA118233)
Guilin Scientific Research and Technology Development Program(2016012505)
