Application and comparison of different chemometric approaches in QSPR modelling of supercooled liquid vapour pressures for chloronaphthalenes.

Molecular descriptors from calculations at the level of Density Functional Theory (B3LYP/6-311++G**) were effectively applied in QSPR estimation of supercooled liquid vapour pressures (P(L)) for individual chloronaphthalene congeners. The estimated values of log P(L) varied from 1.05 Pa to 5.6 x 10(-5) Pa, depending on the number of chlorine substituents present in the molecule and the substitution pattern. Comparison of the five chemometrical methods of modelling (approaches) led to the final conclusion, that the use of relatively simple PLS combined with one of the variable pre-selection algorithms (UVE or GA) seems to be the optimal choice in such computational studies for persistent organic pollutants. The best GA-PLS model was characterized by the value of root mean square error of prediction RMSEP = 0.108 logarithmic Pascal units.