This lecture will describe a systems biological approach that has been employed to identify genes, pathways and networks that regulate metastatic disease formation. These studies have uncovered critical roles for vascular-immune interactions in guiding cancer metastatic progression. They have also uncovered unexpected therapeutic opportunities that have been advanced to human clinical testing.

Immunotherapy has become a mainstay of therapy for non-small cell lung cancer (NSCLC) in recent years, with observed response rates around 35%. Despite these initial responses, many of these patients eventually acquire resistance and progress, while others exhibit durable response maintained for >1 year. Our underlying hypothesis is that the observed results from immunotherapy can be improved with sufficient understanding of the evolutionary (cellular and molecular) and ecological (tissue) dynamics that govern response to immunotherapy. Here we present recent results from our integrated approach, developing mathematical models based on data acquired from a Moffitt NSCLC immunotherapy cohort, as well as detailed in vitro experiments. From the patient data, using multispectral imaging, we have characterized the ecology of individual patient tumors pre- and post-treatment, and derived potential biomarkers for predicting response. These analyses have also been used to parameterize a spatial mathematical model of tumor-immune interactions, which allows us to investigate the mechanisms behind treatment response and develop alternative treatment strategies to extend the time to resistance. From the in vitro experiments we characterized the dynamics of tumor-immune interactions by subjecting NSCLC cell lines to immune pressures, and quantified the changes that occur in tumor cells over time (e.g. upregulation of immunosuppressive molecules), and examined the activation and deactivation cycle of T-cells in the presence of immunosuppressive molecules. These studies have then been embedded within two mathematical models (spatial and non-spatial) that bring together the complex dynamics that occur when these populations are mixed. The observed dynamics have important implications for therapy administration, particularly the timing of follow-up doses of immunotherapies. We close by discussing how the integration of mathematical models with pre-therapy data can predict outcomes from anti-PD-L1 immunotherapy, and how calibrating these models from ongoing clinical trials in NSCLC at Moffitt will allow for patient specific predictions regarding immunotherapy responsiveness.

Cancer initiation and progression are controlled by the interplay between tumor, stromal and immune cells. I will describe the use of highly multiplexed immunofluorescence imaging and micro-region transcriptomics to understand the molecular and spatial diversity of immune-tumor interactions. In the case of early melanomas, a wide diversity of states can be identified in a single resection. These range from cancer precursor cells to melanoma in situ to fields in which tumor cells have successfully been immunoedited. Comparison of these states reveals principles of immune evasion and its potential reactivation for therapeutic purposes.

The acceleration of data acquisition and the availability of increasing amounts of both genetic and functional data are changing the face of biomedical research. Computational approaches can take advantage of these data and bring the promise of improved understanding and treatment of cancer.  I will describe some approaches that leverage multiple data types and emphasize the use of more biologically interpretable models to determine the underlying biological mechanisms of specific tumor subtypes, to stratify patients according to their relative risks of relapse, and to identify candidate compounds for new treatments.  I’ll focus on our application of these methods to Medulloblastoma, one of the most common pediatric brain tumors, and our efforts to translate them into the clinic.

Cancer is a complex adaptive system orchestrated by the interactions between tumor cells and their microenvironment. Colorectal cancer (CRC), one of the deadliest cancers in the United States, has a dismal five-year survival rate of less than 15% for metastatic disease. To advance our understanding of tumor progression and improve treatment efficacy, it is vital to employ biomimetic preclinical models that recapitulate the tumor-microenvironment interactions of human CRC, a feature missing in standard “in vitro” experiments. In this seminar, we will describe our use of biologically-inspired 3D model systems (i.e., organoids and organs-on-chips) coupled with quantitative high-content imaging to illuminate the dynamic interactions between colorectal cancer cells and their microenvironment. Specifically, we will highlight how peristalsis-like mechanical forces and patient-derived cancer-associated fibroblasts, a prominent stromal cell type within the tumor microenvironment, influence CRC invasion and chemoresistance.

Network models have proven in recent years to assist in improving our understanding of the coupled dynamics of biomolecules, from individual proteins to supramolecular systems. Among network models, elastic network models (ENMs) found wide usage in molecular biology.  More recently, ENMs proved useful to exploring chromosomal dynamics, using data from Hi-C experiments to reconstruct in silico the connectivity of the chromatin and provide a physical basis for gene regulation transcription and cell type differentiation. Finally, machine learning algorithms that incorporate ENM predictions provide an improved assessment of the effect of mutations on function, compared to those based on sequence and structure exclusively^3,4. These recent developments and future biomedical and pharmacological applications will be discussed.