Speaker
Dr
Elizabeth Ainsbury
(Health Protection Agency)
Description
The classical methods for dose estimation in radiation cytogenetics are now extremely well defined. However, classical statistical methodology requires that chromosome aberration yields are considered as something “fixed.” This 'frequentist' approach takes into account only the chromosomal damage which has been identified in a particular sample, and as such does not fully consider the intrinsically stochastic nature of aberrations.
The alternative to the classical methodology is to use a Bayesian approach to data analysis. In the Bayesian framework, probability of an event is described in terms of previous beliefs and uncertainty. Previously existing, or prior, information is used in combination with experimental results to infer probabilities or the likelihood that a hypothesis is true. A number of authors have suggested that a Bayesian approach may be preferable for analysis of cytogenetic data and it has been shown that the Bayesian approach increases both the accuracy and quality assurance of radiation dose estimates.
A software tool has been developed in order to bring Bayesian analysis to laboratory cytogeneticists. CytoBayesJ takes a number of Bayesian or ‘Bayesian like’ methods that have been proposed in the literature and presents them to the user in the form of simple user-friendly tools. These include:
- Testing for the most appropriate model for distribution of chromosome aberrations amongst cells from the Poisson, Negative Binomial, Neyman type-A and Hermite models;
- Calculations of posterior probability distributions for yields of chromosome aberrations;
- Calculations of posterior probability distributions for radiation dose;
- Bayesian methods of calculating detection limit and decision threshold.
The software will be presented, together with details of testing with respect to cytogenetic data collected in a number of different radiation exposure scenarios by researchers who are active in the field of radiation cytogenetics. Overall, the applicability of the Bayesian approach to radiation cytogenetics will be further highlighted.
Primary author
Dr
Elizabeth Ainsbury
(Health Protection Agency)
Co-author
Dr
Volodymyr Vinnikov
(bGrigoriev Institute for Medical Radiology of the Academy of Medical Science of Ukraine)
