The University of Manchester; 2013
This thesis with the title:”Logical models of DNA damage induced pathways to cancer” was completed by Kun Tian for his PhD degree in the University of Manchester and submitted in October 2013. Chemotherapy is commonly used in cancer treatments, however only 25 % of cancers are responsive and a significant proportion develops resistance. The p53 tumour suppressor is crucial for cancer development and therapy, but has been less amenable to therapeutic applications due to the complexity of its action reflected in 67,000 papers describing its function. Here we provide a systematic approach to integrate this information by constructing large-scale logical models of the p53 interactome using extensive database and literature integration. Initially we generated models using manual curation to demonstrate the feasibility of the approach. This was followed by creation of the next generation models by automatic text mining results retrieval. Final model PKT205/G3 was generated by choosing the size of the interactome that could be analysed with current available computing power and by linking upstream nodes to input environmental signals such as DNA damage and downstream nodes to output signal such as apoptosis. This final version of the PKT205/G3 model contains 205 nodes representing genes or proteins, DNA damage input and apoptosis output, and 677 logical interactions. Predictions from in silico knock-outs and steady state model analysis were validated using literature searches and in vitro experiments. We identify an up regulation of Chk1, ATM and ATR pathways in p53 negative cells and 58 other predictions obtained by knockout tests mimicking mutations. The comparison of model simulations with microarray data demonstrated a significant rate of successful predictions ranging between 52 % and 71 % depending on the cancer type. Growth factors and receptors FGF2, IGF1R, PDGFRB and TGFA were identified as factors contributing selectively to the control of U2OS osteosarcoma and HCT116 colon cancer cell growth. In summary, we provide the proof of principle that this versatile and predictive model has vast potential for use in cancer treatment by identifying pathways in individual patients that contribute to tumour growth, defining a sub population of “high” responders and identification of shifts in pathways leading to chemotherapy resistance.