Electronic Theses and Dissertations

Identifier

86

Author

Xiaowei Yan

Date

2010

Date of Award

7-8-2010

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Mathematical Sciences

Concentration

Applied Statistics

Committee Chair

Tan Wai-Yuan

Committee Member

Deng Lih-Yuan

Committee Member

Wong Seok P.

Committee Member

Xiong Xiaoping

Abstract

With more and more biological mechanisms of cancer development being discovered, in order to improve cancer control and prevention, it becomes necessary to develop effective and efficient mathematical and statistical models and methods to incorporate the biological information, and to identify critical events in the process of carcinogenesis. In this dissertation, the complex nature of carcinogenesis has been represented by stochastic system model; combining this model with information from observations and prior knowledge, we have developed state space models to evaluate cancer gene mutations and cell proliferation at different cancer development stages. Also, we have proposed a generalized Bayesian method via multi-level Gibbs sampling procedure to predict state (stage) variables of the models. In this dissertation, stochastic models have been proposed for initiation, promotion and complete carcinomas experiments; these experiments are most commonly performed in cancer risk assessment of environmental agents. These stochastic models are simple multi-pathway models which are constructed based on biological mechanisms. The estimates we obtained from the models have provided quantitative evaluation of dose related mutation rates of major genes and cells proliferation rates; these results could be used to assess the risk of developing malignant tumor in the environment we live. More complicated stochastic and state space models have been developed for sporadic human colon cancer and for hereditary and non-hereditary human liver cancer. We have utilized the proposed models to fit to Surveillance Epidemiology and End Results (SEER) data. The results imply that our models have effectively incorporated biological information and observations; these models fitted the data very well and the inferences based on estimate were very consistent with biological findings. Furthermore, the models reflected the complex nature of carcinogenesis. We notice that many cancers are developed through multiple-stage multiple-pathway. Our analyses of colon cancer and liver cancer have showed that some pathways are more devastated than others. This suggests thus it would be more efficient to intervene or treat the critical events in the more devastated pathways.

Comments

Data is provided by the student.

Library Comment

dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Share

COinS