Guided in-silico modelling quantification of drug-induced viral mimicry response in cervical cancer

Abstract

Cervical cancer is one of the most common female cancers with the worldwide number of new cases and deaths reaching hundreds of thousands every year with a rising trend posing a major health issue. This situation creates the need for better understanding of the mechanisms that drive the tumor growth and subsequently promote the study of effective drugs or agents that stop or reduce the growth of the tumor. The gold standard for performing these kinds of analysis is the real world cell cultivation method. However, performing real world cell cultivation experiments comes with high cost it terms of time and resources and new approaches have to be used such as the use of computational methods and specifically the agent based modeling and simulation (ABM/(S)) method. The aim of this thesis is the exploration of the possibility of using decitabine (DAC) a DNA methyltransferase (DNMT) inhibitor as a potential treatment for cervical cancer using the (ABM/(S)) method. For archiving this, we use the biodynamo (ABM/(S)) platform for creating dif- ferent models that try to reflect the results of real world cell cultivation of HELA cells in three different cases that correspond to untreated cells and cells exposed to lower and higher dosages of DAC respectively. In order to fit those real world data, our models explore 10 different properties that affect the cell growth, division and death and this exploration is being done with by use of either Bayesian optimiza- tion or genetic algorithm with different distance functions (L1 and L2) for score calculation in each different model. We have observed that the parameters found by the models that are based on genetic algorithm (and especially those which are using the L2 distance function) make a better fit of the real world data in every different examined case from those models based on the Bayesian optimization algorithm and this is shown by presenting analytical results for every different model as well as model-to-model comparisons in respect to the real world data.