We apply an improved QSAR method to study the toxicity of alcohols by resor
ting to the employment of the following properties: surface area, volume, l
og P, molar refractivity, and polarizability. The idea behind the choice of
these physicochemical descriptors is to use realistic quantum chemical mol
ecular quantities, which can, in principle, express all of the electronic a
nd geometric properties of molecules and their interactions. The log LC50 v
alues for a set of 12 alkanols were used, and we examined the whole set of
possibilities for one-, two-, three-, four- and five-variable relationships
between the toxicity index (log LC50) and the molecular properties up to t
he third order. The analysis of computed data shows that molecular descript
ors are extremely well fitted tools for assessing the toxicity of saturated
alcohols. In particular, we have verified that using higher level relation
ships (i.e. quadratic and/or cubic) in several-variable equations give exce
llent predictions and their corresponding accuracy is greater than that pro
duced by the usual topological indices. (C) 2000 Elsevier Science B.V. All
rights reserved.