Ancestral Inference and Cancer Stem Cell Dynamics in Colorectal Tumor

Tumorigenesis is the process by which a tumor is formed. It starts with the transformation of normal cells into cells with uncontrolled growth, coupled with genetic, epigenetic, and cellular changes. One of the challenges that researchers have to face when studying tumor growth is that only the end point of the growth process is observed. Namely, we are not able to examine the clonal expansion of the first transformed cell. To design an effective therapeutic strategy to treat tumors and to prevent tumorigenesis, basic understanding of the entire developmental details of tumors is needed. Intratumor heterogeneity (ITH) of sequencing data, i.e. the fact that not all parts of a given tumor are the same, has been observed in tumors in various genomic data sets. Thus all tumors are not the same, and it is likely that they should not all be treated in the same way. In this thesis, I aim to better understand tumor growth and the variations in this process between individuals by viewing tumor growth as an evolutionary process. I will decipher the past of a given tumor using techniques borrowed from molecular phylogeny using the information about that past that is contained within the ITH.;In summary, here I describe methods and models to study ITH contained in methylation data, exome sequencing data and SNP array data in order to decipher the ancestry of colorectal tumors and estimate several important parameters, such as methylation error rate, demethylation error rate, mutation rate, number of cancer stem cells, and the probability that a stem cell undergoes symmetric division or asymmetric division.;In Chapter 1, I give a general introduction to the field. In Chapter 2, I demonstrate that DNA methylation data that are collected from colorectal tumors can be used to infer the ancestral methylation state, the number of cancer stem cells, and methylation/demethylation error rate. In addition, I demonstrate that methylation data have limitations for parameter inference. In Chapter 3, I present work I performed as part of the development of an alternative tumor growth model, a "big bang" model, for colorectal tumors, which is supported by genomic data from spatial sampling. In Chapter 4, I explore the possibility that the DNA mutations from single gland exome sequencing data can be utilized to infer the division patterns of cancer stem cells and to infer whether or not there is a mutation burst in the early stage of tumorigenesis. Finally, in Chapter 5, I summarize the work presented in my thesis and discuss future questions of interest.