Abstract
Colorectal cancer is one of the most common cancer worldwide with a high mortality rate. Surgery is the primary treatment of choice, however after resection there is still a considerable risk of relapse for patients with stage II and III. While advances in systemic chemotherapy have improved overall survival of patients with stage III colon cancer, it is not well established which stage II patients will benefit the most from adjuvant treatment. The clinicopathological factors used today are still insufficient to identify stage II patients with high risk of recurrence, and stage III patients with low risk, leading to under or over treatment of these patients. Gene expression profiling has elucidated a broad range of mechanisms governing cancer, and more recently, it has had an impact on clinical care. Specially for breast cancer, genomic profiling has improved our capacity of prognostication, and diverse gene-expression-based assays are commercially available. In view of the successful approach for breast cancer and the rapid development of high throughput technologies for genomic profiling, the search for gene expression signatures has increased considerably and similar approaches have been applied to other types of tumors. However, many gene expression signatures established have failed when applied to independent group of samples. Thus, new methods and greater statistical stringency must be considered. Management, integration, and interpretation of genomic data are major issues of modern biomedical research. Currently, the biggest challenge does not rely on the acquisition of genomic data, but mainly on its interpretation and translation to clinical utility. Considering the large number of genomic information available today, meta-analyzes represent a huge potential to improve the collection and analysis of such data. However, analysis of combined datasets are still poorly explored. The aim of this project is to identify a gene expression signature capable to predict response to adjuvant treatment for colon cancer stage II through combined public gene expression data analysis. The first step is to combine clinical and gene expression data from colorectal cancer independent studies enabling reanalysis of the collective primary data. Collected raw data from microarray and high throughput sequencing studies will be processed and normalized. Afterward, multiple combined datasets will be used to build and validate gene expression signatures. Through machine learning, it will be possible to identify expression patterns associated with specific conditions (eg. response to chemotherapy). Finally, gene expression signature will be validated using independent datasets, such as The Cancer Genome Atlas. In summary, the proposed approach will rely on the vast gene expression data available from independent studies, enabling the discovery of recurrent and reliable gene expression signatures. Accordingly, this strategy should avoid weaknesses of single study-derived signatures, and should improve signature accuracy and the confidence of the results. Finally, this project will elucidate a computational approach with great potential to explore the vast genomic data public available and to be extended to other types of tumor and genomic data. (AU)
|