Ionizing radiation is a health hazard to humans, but is exploited at the same time in various applications, in particular in diagnostic and therapeutic medicine. A profound understanding of the underlying processes, starting from the physical energy deposit up to the biological radiation response, is the basis for a reliable prediction of radiation effects. The subject of this work is the formulation of predictive dose response models.

Special emphasis is set on two aspects, namely the prediction of radiation effects as well as their uncertainties: First, the dependence of radiation effects on physical properties like particle type and energy of the radiation used is discussed. The physical characterization of radiation to which cells or tissues are exposed to is one of the most important determining factors for the effect. Particularly the spatial and temporal pattern of radiation damage induction have to be considered: High ionization densities give rise to clustering of lesions to the DNA, and such complex DNA damage is most critical for the fate of individual cells. Based on experimental findings and theoretic considerations, strategies are developed on how to implement the complex involved physical and biological processes into mathematical models. Such models must be sufficiently simple, testable against experimental data, and practical for application purposes. A set of requirements for radiation effect models has been identified, which is proposed as a list of general criteria for successful models in radiobiology. Based on these requirements, a comprehensive radiobiological model framework for the prediction of radiation damage is introduced. The model applicability to many different aspects of radiobiology is demonstrated, based on one consistent set of concepts, strongly supporting the model assumptions. Second, in addition to effect predictions, strategies to assess the corresponding uncertainties are discussed theoretically and at hand of experimental data. The variability of biological targets as well as errors inferred by model applications are regarded side by side. This is of importance, e.g., for evaluating the accuracy of treatment planning in radiation therapy of cancer.

The developed model framework is put in perspective of current radiobiologic research. The core of the work are six research publications, focusing on various aspects in effect modeling for different radiation qualities. They cover strategies of model set-up and benchmarking using experimental data, as well as aspects of effect uncertainty estimates.