| Despite the advancements in therapies, next-generation sequencing, and our knowledge, breast cancer is claiming hundreds of thousands of lives around the world every year. We have therapy options that work for only a fraction of the population due to the heterogeneity of the disease. It is still overwhelmingly challenging to match a patient with the appropriate available therapy for the optimal outcome. This dissertation work focuses on using biomedical informatics approaches to development of pathway-based biomarkers to predict personalized drug response in breast cancer and assessment of feasibility integrating such biomarkers in current electronic health records to better implement genomics-based personalized medicine.;The uncontrolled proliferation in breast cancer is frequently driven by HER2/PI3K/AKT/mTOR pathway. In this pathway, the AKT node plays an important role in controlling the signal transduction. In normal breast cells, the proliferation of cells is tightly maintained at a stable rate via AKT. However, in cancer, the balance is disrupted by amplification of the upstream growth factor receptors (GFR) such as HER2, IGF1R and/or deleterious mutations in PTEN, PI3KCA. Overexpression of AKT leads to increased proliferation and decreased apoptosis and autophagy, leading to cancer. Often these known amplifications and the mutation status associated with the disease progression are used as biomarkers for determining targeting therapies. However, downstream known or unknown mutations and activations in the pathways, crosstalk between the pathways, can make the targeted therapies ineffective. For example, one third of HER2 amplified breast cancer patients do not respond to HER2-targeting therapies such as trastuzumab, possibly due to downstream PTEN loss of mutation or PIK3CA mutations. To identify pathway aberration with better sensitivity and specificity, I first developed gene-expression-based pathway biomarkers that can identify the deregulation status of the pathway activation status in the sample of interest. Second, I developed drug response prediction models primarily based on the pathway activity, breast cancer subtype, proteomics and mutation data. Third, I assessed the feasibility of including gene expression data or transcriptomics data in current electronic health record so that we can implement such biomarkers in routine clinical care. |
