中文摘要：

[目的]应用定量构效关系研究方法分别构建两个可用于预测金属氧化物纳米材料对人正常肝细胞（L02细胞）和人肝癌细胞（Hep G2细胞）毒性的预测模型。[方法]在16种金属氧化物纳米材料中,随机选取12种纳入测试集用于模型构建,另4种纳入验证集用于模型验证,尝试在所研究金属氧化物纳米材料的结构参数和其对L02细胞及Hep G2细胞半数抑制浓度（IC50）间分别构建出两个具有统计学意义的纳米定量构效关系模型。[结果]成功地运用一个结构参数核核排斥能构建了一个可用于预测金属氧化物纳米材料对L02细胞毒性的预测模型lg IC50=-0.000 056 2ECORE＋3.34（拟合统计量：n=12,F=35.38,R2=0.72,P〈0.005）;用传导能以及最高轨道能和最低轨道能总和的二分之一构建了一个能够用来预测金属氧化物纳米材料对Hep G2细胞毒性的预测模型lg IC50=-0.1EEc＋0.307EShift＋3.67（拟合统计量：n=12,F=10.53,R2=0.70,P〈0.005）。[结论]本次构建的两个模型R2均大于0.6,符合模型构建要求,对金属氧化物纳米材料的设计和安全性评价具有一定的参考价值。

英文摘要：

[Objective] To build two quantitative nanostructure activity relationship（QNAR） models to predict the cytotoxicity of metal oxide nanomaterials towards human normal liver cells（L02 cells） and human liver cancer cells（Hep G2 cells）.[Methods] Sixteen metal oxide nanomaterials were selected,twelve of them were put into training set to build model,and the other four were put into validation set to validate model.The study attempted to correlate the eighteen structural descriptors of these metal oxide nanomaterials with their median inhibition concentration（IC50） towards L02 cells and Hep G2 cells by conducting multiple regression analysis.[Results] Finally,core-core repulsion energy（CORE） was used to build a statistically significant QNAR model for L02 cells,lg IC50=-0.000 056 2ECORE＋3.34（fit statistics：n=12,F=35.38,R2=0.72,P 〈 0.005）.Conduction band（Ec） and Schuurmann MO shift alpha（shift） were used to build a statistically significant QNAR model for Hep G2 cells,lg IC50=-0.1EEc＋0.307EShift＋3.67（fit statistics：n=12,F=10.53,R2=0.70,P 〈 0.005）.[Conclusion] The correlation coefficients of the two models are larger than 0.6,which meet the criterion of a good model,and these two models can provide supportive information on the design and safety assessment of metal oxide nanomaterials.