Each cancer is unique: it reflects the underlying genetic make-up of the patient and the stochastic mutational processes that occur within the tumour. This uniqueness suggests that each patient should receive a personalized type of therapy. Current predictions of a cancer patient’s outcome or prognosis are highly inaccurate. To aid in the prediction of patient prognosis based on highthroughput molecular datasets I have worked to optimize each step of the experimental pipeline: platform annotation, experimental design, consideration of tumour heterogeneity, data pre-processing and statistical analysis, and feature selection. First, a 12k CpG Island clone library was sequenced and annotated using a BLAT analysis. Second, microarrays built using this library were used in a fully-saturated study to evaluate the importance of ChIP-chip experimental design parameters. Third, intra-tumour heterogeneity was shown to influence specific pathways in a large fraction of genes. Fourth, a systematic empirical evaluation of 19,446 combinations of microarray analysis methods identified key steps of the analysis process and provided insight into their optimization. Finally, the combination of a two-stage experimental design and a novel semi-supervised algorithm yielded a six-gene, mRNA abundance-based classifier that could divide non-small cell lung cancer patients into two groups with significantly different outcomes in four independent validation cohorts. Further, a permutation study showed that millions of six-gene markers exist, but that ours ranked amongst the top 99.98% of all six-gene markers. The knowledge gained from these studies provides a key foundation for the development of personalized therapies for cancer patients.
| Identifer | oai:union.ndltd.org:TORONTO/oai:tspace.library.utoronto.ca:1807/17734 |
| Date | 23 September 2009 |
| Creators | Boutros, Paul Christopher |
| Contributors | Penn, Linda Z., Jurisica, Igor |
| Source Sets | University of Toronto |
| Language | en_ca |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.0021 seconds