Predicting Toxicities of Diverse Chemical Pesticides in Multiple Avian Species Using Tree-Based QSAR Approaches for Regulatory Purposes

A comprehensive safety
evaluation of chemicals should require toxicity
assessment in both the aquatic and terrestrial test species. Due to
the application practices and nature of chemical pesticides, the avian
toxicity testing is considered as an essential requirement in the
risk assessment process. In this study, tree-based multispecies QSAR
(quantitative-structure activity relationship) models were constructed
for predicting the avian toxicity of pesticides using a set of nine
descriptors derived directly from the chemical structures and following
the OECD guidelines. Accordingly, the Bobwhite quail toxicity data
was used to construct the QSAR models (SDT, DTF, DTB) and were externally
validated using the toxicity data in four other test species (Mallard
duck, Ring-necked pheasant, Japanese quail, House sparrow). Prior
to the model development, the diversity in the chemical structures
and end-point were verified. The external predictive power of the
QSAR models was tested through rigorous validation deriving a wide
series of statistical checks. Intercorrelation analysis and PCA methods
provided information on the association of the molecular descriptors
related to MW and topology. The S36 and MW were the most influential
descriptors identified by DTF and DTB models. The DTF and DTB performed
better than the SDT model and yielded a correlation (R2) of 0.945 and 0.966 between the measured and predicted
toxicity values in test data array. Both these models also performed
well in four other test species (R2 >
0.918). ChemoTyper was used to identify the substructure alerts responsible
for the avian toxicity. The results suggest for the appropriateness
of the developed QSAR models to reliably predict the toxicity of pesticides
in multiple avian test species and can be useful tools in screening
the new chemical pesticides for regulatory purposes.