PURPOSE: In radiation risk analysis the state-of-the-art approach is based on descriptive models which link excess rates of cancer incidence and mortality to radiation exposure by statistical association. To estimate the number of sporadic and radiation-induced cases descriptive models apply parametric dose response function which directly determine the radiation risk. In biologically-based models of cancer risk (BBCR models) dose responses are implemented for key events on the biological level such as early mutations or clonal expansion of initiated cells. Influenced by radiation these events then shape the risk response on the epidemiological level. Although BBCR models facilitate a more comprehensive consideration of biological processes for risk assessment, their range of application in radiation research remains limited. Therefore, we emphasize their ability to improve understanding of radiation-related carcinogenesis by integrating molecular biology with epidemiology. We highlight the potential of BBCR models to harness information from adverse outcome pathways (AOPs) for risk estimation with closer links to radiobiology. The AOP concept originates from toxicology and may be applied profitably in radiation research. CONCLUSION: The conceptual design of BBCR models can be guided by the high-dimensional data environment provided by AOPs. Risk estimates from BBCR models pertain not only to classical radioepidemiological covariables such as radiation dose or attained age but also to well characterized molecular pathways. By additionally deploying biological information BBCR models facilitate finer risk stratification for a more personalized risk assessment. They leave behind the one-size-fits-all approach of descriptive modeling with the downside of more involved model development. Importantly, predictions from BBCR models can be validated against molecular measurements. Validated predictions would confirm the model design and strengthen the link between molecular biology and epidemiology. But the availability of cancer tissue in good quality from patients with known radiation exposure constitutes a major bottleneck. More ambitious initiative is needed to recover stored tissue samples and make them available for molecular investigations. To conclude, risk estimation will not only on rely on statistical association but will be quantitatively informed with radiobiological insight. Combined with the AOP framework BBCR models could improve accuracy and reduce uncertainty of radiation risk estimates in future research.