On the use of local and global QSARs for the prediction of Physico-Chemical Properties of Polybrominated Diphenyl Ethers

Polybrominated diphenyl ethers (PBDEs) are persistent
chemicals that have been among the most marketed
flame retardants used all over the world in the last decades.
PBDEs have been detected in all environmental compartments,
as well as in humans and wildlife, where they
are able to accumulate and exert their toxic effects. At present
only a limited amount of experimental data is available
to characterize the physico-chemical and toxicological behavior
of PBDEs and similar brominated flame retardants.
QSA(P)R approaches are very useful tools to predict missing
data starting from the chemical structure of compounds. In
this study several local QSPR models, developed specifically
for the prediction of logKoa, logKow and melting point of
PBDEs, were compared with predictions by global QSPR
models, such as KoaWIN, KowWIN and MPBPWIN from the
EPI Suite package, and AlogP and MlogP from DRAGON
software, which were trained on heterogeneous and large
datasets. The analysis addressed in the paper supported
the identification of points of strength and weaknesses of
both local models, and global models. The results are relevant
to support decisions made by general QSAR users and
regulators, when they have to select and apply one of the
analyzed models to predict properties for PBDEs.