Cancer is a heterogeneous disease. Even tumors arising at the same site can differ drastically in terms of their underlying biology, prognosis, and response to treatment. Moreover, the prevalence of some cancer subtypes differs by ethnicity, which may in part account for disparities in incidence and mortality. Understanding the sources and consequences of cancer heterogeneity is thus important to individuals for cancer prevention and treatment of individual and for overall population health. Recent cancer genomic studies have shed light on tumor heterogeneity, but the distinct origins of different genomic subtypes and the relationship between these subtypes and modifiable risk factors (including environmental exposures, diet, smoking and drinking behaviors, and physical activity, among others) remains largely unknown. This workshop will bring together experts in cancer genomics, epidemiology, oncology, pathology, and related data sciences to discuss (a) the state of the science linking prediagnostic risk factors to cancer subtypes, and (b) the opportunities to address these questions using large cancer epidemiology cohorts with extensive risk factor information and archival tumor specimens.