Samaneh Hassanzadeh viyaee - Department of Chemistry,Central Tehran Branch, Islamic Azad University,P.O.BOX 14676-86831 ,Tehran ,IRAN
Maryam daghighi Asli - Department of Chemistry,Central Tehran Branch, Islamic Azad University,P.O.BOX 14676-86831 ,Tehran ,IRAN
Robabeh Sayyadi kord Abadi - Department of Chemistry , Rasht Branch ,Islamic Azad University,P.O.BOX 41335-3516,Rasht,IRAN

Twenty five different Paclitaxel anticancer derivatives were selected as a sample set and the geometry of the complexes were optimized using Gaussian 03W and Polarized continuummodel (PCM) was applied to consider the non-specific solvent effect, and all molecules were optimized in H2O solvent. The activity of the 25 different Paclitaxel derivatives was estimated by means of multiple linear regression (MLR), artificial neural network (ANN), and genetic algorithm (GA) techniques inwater solvent. The results obtained using the GA-ANN were compared with those obtained using MLR- PLS1, MLR-ANN ,GA-MLR and GA-ANN methodes. A high predictive ability was observed for the MLR-PLS1, MLR-ANN ,GA-MLR GA-ANN models, with root mean sumsquare errors (RMSE) of 0.997, 0.497, 0.494 , 0.469. respectively (N=25). The results obtained using the GA-ANN method indicated that the activityof the derivatives of Paclitaxel depends on different parameters such as Pipc03, BIC4 descriptors in solvent phase. In summary, a comparison of the quality of ANN with different MLR methods showed that has ANN a betterPredictive ability

Paclitaxel, Antitumor drugs, QSAR, Gentic Algorithm, Solvent Effect