Cytotoxic activity of 47 semi-synthetic Taxoteres against four different cancer cell lines, such as ovarian cancer (A121), human colon adenocarcinoma (HT29), human breast cancer (MCF7) and drug-resistant human breast cancer (MCF7R), was used to develop a quantitative structure-cytotoxicity relationship (QSCR). The data were divided into training (38) and test sets (9), in which the former was used to develop the QSCR and the later was used for the prediction of three of the cell lines namely A121, HT29 and MCF7. In the case of MCF7R, the training set contained 36 compounds and the test set contained nine compounds. The statistical measures such as squared correlation coefficient (r 2=0.73-0.81), adjusted squared correlation coefficient (r 2 adj =0.72-0.80), cross-correlation coefficient (q2 =0.62-0.75) and F-ratio (17.49-27.49) were found to be satisfactory. Other statistical diagnoses such as n/p ratio, fraction of the variance in terms of r2 and quality factor (Q) were also satisfactory. Thermodynamic, lipophilic/hydrophilic, spatial (Jurs descriptors) and topological (Kier and Hall connectivity indices) descriptors contributed to the activity. Functional groups required at R1-R3 positions for enhanced activity are discussed. The developed QSCR models can be used in the design and development of newer semi-synthetic Taxotere derivatives for the cancer therapy. © 2011 Taylor & Francis.