Background: Metal nanomaterials are widely used in various fields, including targeted therapy and diagnosis. They are extensively used in targeted drug delivery and local treatments. However, the toxicity associated with these materials could lead to severe adverse health effects.

Methods: In this study, we investigated the relationships between the toxicity and structures of metal nanoparticles by using theoretical calculations and quantitative structure-activity relationships. Twenty four physicochemical descriptors and toxicity data of 23 types of metal nanoparticles were selected as samples, and a multiple linear regression model was established to obtain a toxicity prediction equation with 5 descriptors with an R2 of 0.910. Structures of copper nanoparticles were designed based on the model, and the structure with low toxicity was searched. The multiple nonlinear regression model was used to further improve the prediction accuracy.

Results: The R2 values were 0.995 in the training set and 0.988 in the test set, which indicated that the prediction accuracy improved. Based on the result of multiple linear regression, we designed copper nanoparticles with low toxicity.

Conclusion: The study confirmed that the quantitative structure-activity relationship was a reasonable method for predicting the toxicity and designing the structures with low toxicity of metal nanoparticles.

背景：金属纳米材料被广泛用于各种领域，包括靶向治疗和诊断。它们被广泛用于靶向药物输送和局部治疗。但是，与这些材料相关的毒性可能导致严重的不利健康影响。

方法：在这项研究中，我们通过理论计算和定量构效关系研究了金属纳米颗粒的毒性与结构之间的关系。选择23种金属纳米粒子的二十四个理化指标和毒性数据作为样本，建立了多元线性回归模型，得到了具有五个描述符的R2为0.910的毒性预测方程。基于该模型设计了铜纳米颗粒的结构，并寻找低毒的结构。使用多元非线性回归模型进一步提高了预测准确性。

结果：训练集的R2值为0.995，测试集的R2值为0.988，表明预测准确性有所提高。基于多元线性回归的结果，我们设计了低毒性的铜纳米粒子。

结论：研究证实定量构效关系是预测金属纳米颗粒毒性和设计低毒性结构的合理方法。