QSAR Models for P‑Glycoprotein Transport Based on a Highly Consistent Data Set

P-Glycoprotein (Pgp) is involved in the elimination and
in the
disposition of a significant portion of marketed drugs. So far, publicly
available data sets used for modeling Pgp transport included compounds
tested in different assays, different cell lines, and different protocols.
In this work, we present a collection of 478 Efflux Ratios (ERs) in
MDCK-MDR1 cell lines, and from this collection we define a data set
of 187 compounds that were tested in the Borst-derived MDCK-MDR1 cell
lines. Of the 23 models resulting from the use of different descriptors,
classification algorithms, and variable selection techniques, the
4 most accurate in external validation (∼0.86) are based on
VolSurf+ (VS+) descriptors. Two of these models are Naïve Bayes
(NB) classifiers using 4 descriptors that were selected through a
new technique hereby first time extensively described.