基于主成分回归的整合模型预测重金属混合物毒性

为解决CA和IA模型预测结果共线性的问题，基于主成分回归改进已有整合加和模型ICIM，建立新的混合物整合模型（PCR-IAM），并预测加和、协同和拮抗相互作用的重金属混合物联合毒性.以混合物实验浓度为因变量，浓度加和与独立作用预测混合物效应浓度的主成分回归为自变量，建立了PCR-IAM模型.以4个二元混合物体系（Ni-Fe、Ni-Pb、Ni-Cd和Ni-Cr）共20条混合物射线的联合毒性（共240个样本点）验证PCR-IAM模型的预测能力.结果表明，所有二元混合物的PCR-IAM模型的决定系数（R²）和留一法（LOO）交叉验证相关系数（Q²）值均大于0.95，表明PCR-IAM模型能够准确预测20条加和效应、协同和拮抗作用混合物的联合毒性.因此，经验数学模型PCR-IAM模型可以准确预测加和效应、协同和拮抗作用混合物毒性，为构建更合理的整合模型及环境混合污染物的风险评估提供可靠的技术手段.

Prediction of toxicity of heavy metal mixture by integrated model based on principal component regression

In order to solve the problem of the prediction collinearity from CA and IA models, a new model with principal component regression (PCR-IAM) was developed. The PCR-IAM model is able to predict the joint toxicities of heavy metal mixtures with additive, synergetic and antagonistic effects. The PCR-IAM model was developed by using the experimental mixture concentration as dependent variable, and the principal component regression of concentration addition and independent action predictions as independent variable. Four binary mixture systems (Ni-Fe, Ni-Pb, Ni-Cd, and Ni-Cr) representing 20 mixture rays from 240 sampling points was used to verify the predictive power of the PCR-IAM model. The results showed that the coefficient of determination (R²) and leave-one-out cross-validation correlation coefficient (Q²) were greater than 0.95, which proved that the PCR-IAM model can accurately predict the mixture toxicities of 20mixture rays that presented additive, synergistic, and antagonistic effects. Therefore, the PCR-IAM model can precisely predict additive, synergistic, and antagonistic mixture toxicity, which provides a reliable method for risk assessment of environmental mixtures.