Bayesian network models of interdependent chromosomal aberrations in unstable cancer genomes.
Doctoral thesis, UCL (University College London).
Cancer results from the sequential accumulation of heterogeneous mutations and epigenetic changes, leading to “onco-selection” of cells with survival/growth advantages in specific microenvironments. Chemotherapy and radiotherapy act by damaging DNA; uncoupling DNA damage from cell death can cause treatment resistance. Deregulation of the DNA damage response (DDR) can result from aberrant expression of DDR-related genes, due to chromosomal aberrations. Solid tumours often have multiple, recurrent chromosomal aberrations, suggesting coordinated somatic evolution of cancer genomes via co-selection of interdependent chromosomal aberrations. This thesis investigates the hypothesis that “onco-selection” acts on unstable cancer genomes by selecting for multiple changes together. A comprehensive DDR signalling network was constructed. The cytogenetic location of its genes was analysed. The non-random genomic organisation of DDR genes partly coincided with regulation by shared transcription factors, notably oncogenic SNAIL. SNAIL expression was investigated in relation to hypoxia and apoptosis in xenografts of four human colorectal cancer cell lines. SW1222 and LS174T had weak and strong SNAIL expression, respectively, so could be used to investigate the role of SNAIL in treatment resistance through regulation of DDR genes. Bayesian Networks (BNs) were constructed to model probabilistic dependencies between the proposed co-selected, DDR-related chromosomal aberrations in breast, colorectal, lung, ovarian and prostate cancer. The significance of the models was assessed using newly-developed software. Breast cancer involved both co-selection of aberrations on different chromosomes and loss/gain of adjacent regions, whereas loss/gain of adjacent chromosomal regions predominated in the other cancers. A therapeutic combination of DDR targets (RELA, DVL1, E2Fs, CDC45L and MAP2K1/2) was proposed by mapping the predicted dependencies in the breast cancer BN model onto the DDR network, and analysing gene function and network topology. DDR network motif analysis further supported the importance of targeting RELA in combination therapies. This thesis provides insight into cancer resistance to DNA damaging therapies, and may help to predict new personalized combinations of targets to overcome treatment resistance.
|Title:||Bayesian network models of interdependent chromosomal aberrations in unstable cancer genomes|
|Additional information:||Authorisation for digitisation not received|
|UCL classification:||UCL > School of Life and Medical Sciences > Faculty of Medical Sciences > Wolfson Institute and Cancer Institute Administration > Cancer Institute|
Archive Staff Only