A tumor can be thought of as an evolving unicellular parasite within the host's body. This process is fundamentally an evolutionary process: the progression of cancer from normal tissue to early neoplasms to carcinomas and metastases is driven by the accumulation of somatic mutations; a few of these are drivers and confer fitness advantages to the tumor cells, and the remaining are passenger somatic mutations. Just like in species evolution a phylogenetic tree connects all species, in cancer somatic evolution a cell lineage tree connects all cells within a tissue. Studying cancer from an evolutionary perspective is key. For instance:

• Driver mutations responsible for onset of tumor originate in ancestral branches of the tree
• Proliferation can be quantified across each branch by estimating the number of somatic mutations within the branch
• Extremely "fit" clonal variants, often invasive or metastatic, are characterized by higher mutation load and perhaps lower heterogeneity within the clone
• Treatment should focus on eliminating all the clones, not just the most abundant ones, so that there is no relapse

We study cancer from an evolutionary perspective. We develop the new algorithms and methods required for this approach, and apply them to data that we generate in collaboration with colleagues at the School of Medicine. We have been working on breast cancer, sarcoma, follicular lymphoma, and colorectal cancer in collaboration with the laboratories of Arend Sidow, Rob West, Jim Ford, Ash Alizadeh, and Hanlee Ji. In our ongoing efforts, we apply the 10x Gemcode technology that allows us to get a much more detailed picture of the cancer genome. Some of our ongoing research directions include:

• Development of cancer phylogenetic approaches. See LICHeE, our multi-sample cancer cell lineage tree reconstruction tool, and POMEGRANATE, our cancer lineage tree simulator.
• Development of methods for leveraging 10X Gemcode read cloud data to reconstruct cancer genomes.
• Large-scale studies of breast cancer genomes in collaboration with the our colleagues at the Medical School. 
• Deep learning for integration of existing cancer genomic and functional genomic data.