Projects

We construct computational models that capture and ultimately explain the origins of responsiveness and resistance to RAF/MEK inhibitors in melanoma. This class of drugs (which includes vemurafinib, dabrafenib, cobimetinib, and trametinib) represents a triumph of targeted cancer medicine and often induces dramatic tumor shrinkage in patients with BRAF mutant melanoma. However, response is usually transitory and many individuals develop drug-resistant, resurgent disease. We are attempting to understand the initial steps in the emergence of resistance, which are based on reversible adaptive changes in tumor cells, as well as the contributions of adaptive resistance to the emergence of fully resistant tumor cell clones. The projects involves collecting time-series single-cell and population-level data from cells with diverse genotypes followed by temporally resolved modeling using differential equations, logic-based models, and supervised machine learning. These studies are performed primarily in cell lines and mouse models but are informed by data from patients.

Nascent tumors are under continuous surveillance by the immune system and suppression of this surveillance is one of the key steps in the development of malignancy. Drugs acting as immune checkpoint inhibitors (ICIs) restore anti-tumor immunity and result in very durable responses (potentially even cures) in some diseases. The key to understanding why ICIs work in some tumor types and not others is understanding a complex tumor milieu (the tumor microenvironment; TME) in which tumor, stromal and immune cells interact. We are exploiting highly multiplexed imaging methods to immuno-profile tumors from patients before and after treatment with ICIs or other therapeutic drugs. The precise proportions and spatiotemporal arrangements of tumor, stromal and immune cells are being measured in tissue biopsies, and single-cell features extracted and associated with disease progression and therapeutic response using machine learning, deep learning and high-dimensional data analysis. Adverse responses in the skin and gut are also being investigated and compared to therapeutic responses at a mechanistic and clinical level. In this case, the aim is to reduce the burden of ICI therapy on individual patients.

Animal models, primarily syngeneic and genetically engineered mice, are essential for understanding immune checkpoint mechanisms and the drugs that targets these mechanisms. Only in animals is it possible to differentiate between genetic and non-genetic factors influencing immune cell-tumor recognition and test specific mechanistic hypotheses. We use ICIs administered alone or in combination in tumor-bearing mice to evaluate whether highly efficacious responses arise from co-targeting cells of single lineages (e.g. CD8+ effector T cells) or from concurrent targeting of multiple cell lineages (e.g. CD8+ T cells, regulatory T cells), and to identify the tumor settings in which either strategy might prove more effective. Metabolic, signaling, and transcriptional changes associated with cellular responses to ICIs will be assessed and modeled. Agent-based models are used to study non-cell autonomous mechanisms that mediate the therapeutic effects of checkpoint inhibitors, and novel strategies for combining agents that augment efficacy without increasing toxicity will be identified.