The BioHIT Group of SKEL Lab organises, an open to all, online presentation on Thursday 1 October 2020 at 14.00 with visiting speaker Dimitris Papakonstantinou, entitled Using Machine learning techniques for asserting cellular damage induced by particle high-LET radiation .
This is a pilot study concerning the use of Machine Learning (ML) techniques for ascertaining the effect of particle ionizing irradiation to cell survival and DNA damage. Our main goal is to create a ML-based prediction model of the characteristic radiobiological alpha and beta coefficients of the basic Linear Quadratic (LQ) model and several key metrics that quantify DNA damage. The LQ model is the most common way to associate the response of a population of cells that are being irradiated with the radiation dosage. The surviving population follows the S = e^ −( αD + βD ) mathematical equation in a semi logarithmic scaled graph in relation to the dosage measured in Grays (Gy). Our main effort is to develop a computational prediction framework that will be able to assess key radiobiophysical quantities without sacrificing their interactive nature, and on the other hand to interpret biophysically to the best extent the model and its predictions. Complex DNA damage is calculated using original publicly available datasets using the fast Monte Carlo simulation code (MCDS). MCDS provides metrics that relate to the quantity and distribution of DNA double strand breaks (DSBs) ιn a genome, something which is widely accepted to be closely related to cell survival and more broadly to the pernicious biological effects of IR and even more in the case of particle irradiation. We critically discuss the varying importance of physical parameters like ion atomic number (Z), charge, linear energy transfer (LET) and the uncertainties of our predictions as well as future directions and the dynamic of our approach. Our endeavor is to produce a ML prediction model that can take into consideration the intrinsic complexities and stochastic effects produced by the interaction of the particle ionizing radiation with biological tissues. Current empirical models do not always take into account such interactions, thus our work will provide a useful prediction algorithm and an interpretation framework in the case of exposure to particle radiations.
Dimitris Papakonstantinou holds an BEng degree in Applied Mathematics and Physical Sciences from the National Technical University of Athens and a MSc in Bioinformatics and Computational Biology from the Faculty of Biology, National Kapodistrian University of Athens. His MSc thesis was on Machine Learning Techniques in predicting Cell survival and DNA damage complexity after irradiation. His academic and professional interests involve Bioinformatics, Statistical Learning and Information Theory. His main scientific interest involves quantifying and modeling biological response utilizing artificial intelligence, graph theory and complex systems theory, with the goal to make an interpretation framework that bridges the gap between the scale of the molecular/genetic to the the systemic scale. He is also very interested in being involved in building open-access systems to scientific community. He also works as a web developer, and has a thorough knowledge of a broad spectrum of web technologies.