Cancer therapies, like chemotherapy are generally based on heuristic approaches and expert knowledge. Introducing mathematical and engineering methods into the therapy design process has great potentials in therapy optimization. We investigate the application of a discrete time, impulsive therapy generation algorithm for a model that describes living tumor and dead tumor volume dynamics, drug level dynamics, using mixedorder pharmacokinetics and input saturation. We propose an algorithm that calculates low doses of injections that are required to reach or approximate the best results that can be achieved by the application of the drug. The algorithm is tested based on virtual patients (mice) whose parameters are identified based on measurement from experiments with pegylated liposomal doxorubicin as cytotoxic agent and breast cancer as tumor. The algorithm tested in silico shows much better performance than the protocol used in the experiments.