The paper presents a novel approach to prediction of the combined therapy outcome for non-small lung cancer patients. A hybrid model is proposed, consisting of two parts. The first one is a mathematical model of tumor response to therapy, whose parameters are estimated by a neural network based regressor, constituting the second component of the hybrid model. Estimation is based on data from mass spectrometry of patient blood plasma samples. Comparison of clinical and simulation-based survival curves is used to evaluate the quality of the model.