Quantitative Issues in Cancer Research Working Seminar

Alexander Gusev, Ph.D.
Assistant Professor, Medicine, Harvard Medical School
Assistant Professor, Medicine, Dana-Farber Cancer Institute

“Using Thousands of Sequenced Tumors to Predict Patient Response to Cancer Treatment”

ABSTRACT: Targeted tumor sequencing is now common standard of care for many cancers, including nearly >20,000 tumors sequenced at Dana-Farber as part of the Profile project. I will discuss novel methods that leverage this data to understand and predict patient outcomes. First, I’ll present methods that can infer genome-wide germline variation from tumor-only sequencing using “off-target” content that is typically discarded. Second, I’ll show how genetic ancestry inferred from tumors using these methods can modify the prognostic effect of somatic drivers and substantially change patient prognoses. Third, I’ll discuss a machine learning approach to predict the primary site for Cancers of Unknown Primary (CUPs), understand the heritability of CUPs, and predict targeted therapies for CUP patients.