QSAR Studies on Toxicity of Organic Compounds to Chlorella vulgaris

The quantitative structure-activity relationships (QSAR) studies on toxicity of 91 organic compounds to Chlorella vulgaris have been performed by using ν-support vector machine(ν-SVM) algorithm and taking the 2D-autocorrelation descriptors as the structural parameters based on variable selection with particle swarm optimization(PSO) methed. The correlation coefficient(R²) and Q_cv² of PSO-ν-SVM model in the training set are respectively 0.9469 and 0.7216, and R_a² in the test set is 0.9446 while the R², Q_cv² and R_a² of training set and test set of the reference model are 0.9340, 0.9090 and 0.9290, respectively. The result shows that the QSAR model has better stability and prediction ability, so this model is a good reference for the study on toxicity of organic compounds to Chlorella vulgaris.