Modelling the Cytotoxicity of Halogenated Aliphatic Hydrocarbons. Quantitative Structure‐Activity Relationships for the IC50 to Human HeLa Cells
Lennart Eriksson, Björn E.R. Sandström, Michael Sjöstróm, Mats Tysklind, Svante Wold · Quantitative Structure-Activity Relationships · 1993
Abstract The cytotoxicity towards human HeLa cells has been determined for a series of halogenated aliphatic hydrocarbons. The cytotoxicity, expressed as the inhibitory concentration lowering cell viability by 50%, IC50, is modelled by means of calculating a multivariate quantitative structure‐activity relationship based on a training set comprising ten chemicals. It is shown that the cytotoxicity of the compounds critically depends on their hydrophobic and steric properties. The QSAR is validated by making predictions for an external validation set consisting of six substances. In this context, a comparison between internal validation, that is crossvalidation, and external validation of the predictability of a QSAR, is made. The results obtained indicate that the latter method is the more demanding and likely leads to a model exhibiting optimal predictive power. Subsequently, the validated QSAR is also used for making predictions for 40 non‐tested compounds of similar chemical properties. The question of variable selection in the context of QSAR calculations is also addressed. Finally, the cytotoxicity endpoint is multivariately related to eight other biological variables for the training set compounds. This, so called, multivariate biological profiling, indicates that the cytotoxicity endpoint mediates information not conveyed by the other eight variables.