Despite high rates of initial response to frontline treatment in many human cancers, mortality largely results from relapse or metastasis. Diverse clinical responses are considered to be the result of intratumoral diversity: all cells within a given tumor do not possess the same behavior or response to therapy. Understanding tumor heterogeneity is key to improving outcomes for patients with cancer. Although debate remains as to whether cells with treatment resistance exist as part of the initial tumor at presentation versus develop under the pressure of therapy, many studies suggest it to be the former. Further, this intrinsic heterogeneity observed in primary tissues is not accurately represented or studied through genetic animal models or cell lines and does not lend itself to study of bulk tumor cells. Understanding the intrinsic heterogeneity of tumor populations will improve the ability for clinicians to make prognosis and treatment decisions for patients. This requires using single-cell studies to identify risk-associated individual cell populations across patients.
We apply high-dimensional, single-cell approaches to primary patient samples, primarily in the study of childhood leukemia and solid tumors, to identify treatment resistant cell populations. Further, we are developing newer and more accurate models of clinical risk by utilizing machine learning and neural network modeling of single-cell data in combination with patient level data to learn more about risk of treatment failure and relapse. Finally, by identifying cell populations associated or causative of relapse risk, we interrogate mechanisms of resistance and new therapeutic opportunities.