Algorithms and Statistics
Decades of research have yielded very few clinically-useful biomarkers predictive of a patient's response to therapy. However, the emergence of high throughput technology, such as gene profiling, proteomics and metabolomics, is currently revolutionizing the way clinical research is performed. In the future these technologies will also change how patients are diagnosed and treated. Decisions will no longer be based on statistical probabilities, but on protein profiles in the individual's blood and on gene expression of that personís tumor. However the integration of these technologies into clinical research is difficult and presents new methodological challenges in the areas of design of experiments, Bayesian modeling, determination of false discovery rates, and development of algorithms for carrying out the analysis. The goal of our research is to develop reliable and valid methodologies for discovering biomarkers of a disease by using proteome spectra but also by exploiting prior information that can be obtained from other studies or the literature.