On the prediction of cytotoxicity of diverse chemicals for topminnow (Poeciliopsis lucida) hepatoma cell line, PLHC-1^$

Two data sets on the cytotoxicity of diverse chemicals to topminnow (Poeciliopsis lucida) hepatoma cell line (PLHC-1) were modelled with quantitative structure–toxicity relationship (QSTR). The data sets are based on 3-amino-7-dimethylamino-2-methylphenazine hydrochloride (NR) and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assays representing lysosomal damage and metabolic impairment, respectively. The descriptors were calculated with DRAGON 6 and SPARTAN 10 software packages. Descriptor selection was made by ‘all subset’ and genetic algorithm-based features implemented in QSARINS software. The proposed QSTR models were validated both internally and externally. For both endpoints, statistically satisfactory QSTR models were generated with n_Tr = 39; r²_Tr = 0.782; RMSE_Tr = 0.466; n_Test = 18; r²_Test = 0.799; RMSE_Test = 0.360 for NR-based model and n_Tr = 32; r²_Tr = 0.775; RMSE_Tr = 0.460; n_Test = 10; r²_Test = 0.864; RMSE_Test = 0.290 for MTT-based model. Additionally, the QSTR models generated for NR and MTT endpoints were used to predict the cytotoxicity of an external set of 657 and 652 diverse chemicals with structural coverage of 98.6% and 98.3%, respectively. A moderate correlation was observed between the experimental in vivo and predicted in vitro values for external set chemicals. The QSTR models may provide an initial, rapid screening and prioritization of these diverse chemicals for the acute fish toxicity assessment and reduce the need for extensive in vivo toxicity testing.