Phylogenetic Tree Modelling and Inference using Bayesian Methods applied to Reccurent Follicular Lymphoma Tumor Data

Specialeforsvar ved Adrian Fabich Balk-Møller

Titel: Phylogenetic Tree Modelling and Inference using Bayesian Methods applied to Reccurent Follicular Lymphoma Tumor Data

Abstract: In this thesis we consider the phenomena of a single ancestral cell evolving into a cancer tumor, which is a whole population of cells. Throughout this process we imagine single cells divide, die and mutate. To model this process we combined a special case of birth-death processes and the Jukes Cantor mutation model. We analyzed extracted DNA sequences from 42 recurrent follicular lymphoma tumors from 20 different patients and estimated model parameters using algorithms based on Approximate Bayesian Compu- tation (ABC). We conducted simulation studies to investigate various aspect of the ABC algorithms, in particular estimation accuracy and computation time, and the final algo- rithms were chosen with respect to a satisfactory trade-o_ between the two. Estimated model parameters were used to estimate growth and age ratios between first occurring tumors and relapse tumors under some assumptions on unobserved parameters. These estimated ratios suggested that relapse tumors are in general faster growing and older. However, the parameter governing the birth-death process that was involved in this com- putation was estimated with large uncertainty, making these conclusions questionable. We believe this uncertainty can be attributed to components of the ABC algorithms, in particular the used summary statistics. 

Vejleder: Carsten Wiuf
Censor:    Lars Nørvang Andersen, Aarhus Universitet