Two QSAR models for predicting the toxicity of chemicals towards <i>Tetrahymena pyriformis</i> based on topological-norm descriptors and spatial-norm descriptors

Quantitative structure-activity relationship (QSAR) is important for safe, rapid and effective risk assessment of chemicals. In this study, two QSAR models were established with 1230 chemicals to predict toxicity towards <i>Tetrahymena pyriformis</i> using multiple linear regression (MLR) method. The topological(T)-QSAR model was developed by using topological-norm descriptors generated from the topological structure, and the spatial(S)-QSAR model were built with spatial-norm descriptors obtained from the three-dimensional structure of molecules and topological-norm descriptors. The <i>r</i>²_training and <i>r</i>²_test are 0.8304 and 0.8338 for the T-QSAR model, and 0.8485 and 0.8585 for the S-QSAR model, which means that T-QSAR model and S-QSAR model can be used to predict toxicity quickly and accurately. In addition, we also conducted validation on the developed models. Satisfying validation results and statistical parameters demonstrated that QSAR models based on the topological-norm descriptors and spatial-norm descriptors proposed in this paper could be further utilized to estimate the toxicity of chemicals towards <i>Tetrahymena pyriformis</i>.定量构效关系（Quantitative structure-activity relationship, QSAR）对于化学品的安全、快速且有效的风险评估具有重要意义。本研究基于1230种化学品，采用多元线性回归（multiple linear regression, MLR）方法构建了两种定量构效关系模型，用于预测梨形四膜虫（Tetrahymena pyriformis）的毒性。其中，拓扑（T）-QSAR模型采用由分子拓扑结构生成的拓扑范数描述符（topological-norm descriptors）构建；空间（S）-QSAR模型则结合了由分子三维结构获取的空间范数描述符（spatial-norm descriptors）与拓扑范数描述符进行构建。拓扑-QSAR模型的训练集决定系数（r²_training）与测试集决定系数（r²_test）分别为0.8304和0.8338，空间-QSAR模型的对应值分别为0.8485和0.8585，表明两类模型均可快速、准确地完成毒性预测。此外，本研究还对所构建的模型进行了验证。令人满意的验证结果与统计参数证明，本文提出的基于拓扑范数描述符与空间范数描述符的定量构效关系模型，可进一步用于评估化学品对梨形四膜虫的毒性。