Medulloblastomas are the most common malignant brain tumors of childhood, with overall mortality of 40 to 50 percent. Yet their pathogenesis is largely unknown, their cell of origin is debated and their outcome is difficult to predict. My laboratory is addressing these questions, focusing on how genes that regulate cerebellar development become disrupted to promote medulloblastoma growth.
Our lab has measured genome-wide gene expression, somatic gene mutations and chromosome copy-number profiles of medulloblastoma. We have shown that medulloblastomas arise from cerebellar neuronal progenitors, resolving a decades-long dispute on their origin. Examining the molecular profiles further, we have found that as many as six medulloblastoma subtypes can be identified. Each has a unique genetic and molecular "fingerprint" that reflects the mechanisms that drive tumorigenesis and that render the tumor resistant to conventional treatment with chemotherapy and radiation.
Our lab is now investigating these molecular profiles to identify targets for new therapies that can supplement or even replace conventional therapies.
We have used supervised learning classification algorithms to identify differences in gene expression between patients who survived after treatment and those who succumbed to their disease. Outcome prediction models, based on these gene expression profiles, are by far the most accurate predictors of medulloblastoma outcome currently available. These models have been incorporated into therapeutic protocols conducted by the national cancer cooperative, the Children's Oncology Group.