Etude écotoxicologique des composés organiques sur les espèces aquatiques, accumulation et impact sur la santé humaine. Modélisation QSAR

Abstract

The objective of this present work is to develop reliable, stable and predictive QSAR models for the prediction of two activities of some families of persistent, bioaccumulative and toxic pollutants (PBTs) with respect to to fish. 1- The prediction of the bioconcentration factor of a series of 145 PBTs compounds. 2- The prediction of the toxicity of a series of 33 PBTs compounds. These Persistent Organic Pollutants products are ubiquitous contaminants in the environment due to their wide use in industry and agriculture. The best QSAR models obtained are developed with the PLS (Partial Least Square) method, their stability and explanatory power have been verified by Leave One-Out (LOOcv) cross-validation and Y Randomization. Their predictive power was verified by external validation on the test series not included in the training set and by verifying all of the Tropsha criteria. The field of applicability of the selected models is verified by the "leverage" method and the results obtained show that the QSAR models developed are reliable, stable and have good predictive capacity.