Colchicine is a highly toxic plant-derived alkaloid which inhibits microtubule polymerization by binding to tubulin dimers. Currently, the chemotherapeutic value of colchicine is limited by its toxicity against normal cells. Theoretically, this could be remedied by derivatizing colchicine to preferentially bind tubulin isotypes which are more common in cancer cells than in normal body tissues, and particularly in those cancer types which are resistant to conventional therapies. In recent studies, it has been demonstrated that class III ß-Tubulin over-expression is associated with taxane-resistant subsets of non small cell lung cancer, advanced ovarian cancer, breast cancer and cancer of unknown primary origin. Our study investigates the uses of Quantum Mechanics Molecular Mechanics (QMMM) and Molecular Dynamics (MD) modeling to construct derivatives of colchicine which will bind class III ß-Tubulin with increased affinity. Using QMMM and MD modeling techniques, 21 colchicine derivatives were designed to increase affinity for class III ß-Tubulin by offering a better steric fit into the binding pocket . Derivatives were designed and tested in silico before being synthesized by organic chemists at Oncovista Inc. of San Antonio, TX. The colchicine derivatives were then tested in MTS cytotoxicity assays against up to seven different cancer cell lines with differing characteristics and morphologies. Results were obtained by graphing the MTS absorbance readings, and calculating an EC50 Value (drug concentration at which 50% of the drug's effects are seen) using sigmoidal dose-response analysis. Colchicine has an EC50 Value in the range of 10-7 M, and several of our novel derivatives (ie. CH-32, CH-34 and CH-35) were found to have EC50 Values in the range of 10-9 M, while other derivatives (ie. CH-6, CH-7 and CH-21) were found to have EC50 values in the range of 10-5 M to 10-6 M. These results indicate that our derivatives have up to 100X greater and lesser effectiveness than colchicine. Interestingly, comparative derivative cytotoxicity was found to correlate with theoretical QMMM and MD modeling predictions. Successful derivatives warrant continued investigation, screening and development. We propose that our modeling system may be used to design any variety of drugs for specific targets such as vinca alkaloids, taxanes and peloruside.