Francesco Falciani University of Birmingham A bioinformatics approach to investigating cellular interactions in cancer Statistical modelling techniques have been used successfully to link the molecular state of tumour cells to clinical and pathological features of the tumour. These have demonstrated to have the potential to be useful clinical markers and have contributed to reveal novel aspects of the disease physiology. The molecular state of tumour cells is however only representing one of the components that influence the development of the disease. The tumour is a heterogeneous collection of interacting cell types. The physiology of the tumour is influenced by factors produced by stromal cells, endothelial cells and normal epithelial cells. Although it is possible to isolating and profile individual cell populations from clinical samples there still not a systematic approach for the identification of molecular signatures linking the molecular state of normal cells to tumour physiology. This talk describes the development of cell-type specific statistical models linking the molecular state of normal epithelial cells to tumour physiology and clinical outcome in Prostate cancer. We also describe the use of Bayesian networks for pathway inference and its future application to deciphering cell interaction. Time: 22 February 2005 Place: 8.01 EC Stoner, Boardroom, School of Computing. Back to seminar list