Quantitative Approaches to Understanding and Optimizing Oncolytic Viral Therapy

Tyler Cassidy, McGill University

We use a mathematical model to study the importance of tumour-immune interaction in disease progression and model viral therapy- a recently developed therapeutic technique. We characterize the importance of immune involvement in long-term treatment outcomes and, through a virtual clinical trial, optimize treatment scheduling.

Leukemic stem cell signatures identify novel therapeutics targeting acute myeloid leukemia

Kolja Eppert, McGill University Health Centre

Uncovering Hidden Connections in Hematological Regulatory Networks

Ski Krieger, Harvard University

The process of producing all of the body's blood cells is robustly regulated by an extensive network of cytokines. Generally speaking, the principal (and often secondary, tertiary etc.) role of each of these molecules has been identified. However, given the fundamental role of the hematopoietic system and its constituents to organismal health, many cytokines perform ancillary functions within network structures that remain to be elucidated. Oftentimes, hematopoietic dysregulation in the form of blood pathologies can help to identify dynamical relationships that we are unable to discern in healthy individuals. Cyclic thrombocytopenia is one such dynamical disease. We analyzed time series data from an individual with CTP using standard Fourier and statistical approaches (periodogram and correlation analysis) in addition to convergent cross mapping (Sugihara et al., Detecting causality in complex ecosystems. Science, 2012) to infer causal relationships amongst 64 cytokines (and their gene expressions), uncovering a plethora of novel relationships. These results further refine our understanding of the many networks that support the production of blood cells and immunological responses, and may help to identify novel therapies and drug targets.

A Hybrid Mathematical Approach to Improve Chemotherapy Implant Treatment of Pancreatic Cancer

Adrianne Jenner, University of Sydney

En collaboration avec les Séminaires de l'axe de pharmacométrie et pharmacothérapie de la Faculté de Pharmacie

Presented in collaboration with the Pharmacometrics and Pharmacotherapeutics Seminar Series of the Faculty of Pharmacy

Developing an effective treatment for pancreatic cancer presents a unique challenge for two reasons: (1) intravenous drug delivery results in limited concentration at the tumour site and (2) pancreatic tumours form dense impenetrable heterogeneous masses that inhibit drug diffusion. Chemotherapy-loaded implants are a promising experimental treatment that could overcome these challenges. The implants allow localised sustained delivery of chemotherapy and, by carefully selecting the implant injection site, can bypass the dense tumour structure. Using a hybrid Voronoi Cellular Automaton (VCA) and PDE modelling approach, we optimize two attributes of this therapy: the drug release profile and implant configuration. The implants can be made with different levels of polymer concentrations which affect the drug release profile. From the model, we have determined the optimal release profile as a function of polymer concentration. We have also quantified the dependency of treatment outcome on implant location and configuration. Our model presents a unique visualisation tool for our collaborators and allows the mathematical optimisation of the implants to be communicated effectively. The model and techniques we present could easily be translated to a range of medical applications.

Inference of perturbed immune repertoire dynamics using high-throughput sequencing

Max Puelma Touzel, Université de MontréAL

High-throughput sequencing provides access to expression-level detail of cell populations. Identifying signal in this data is nevertheless a challenge: intrinsic variability, vast subsampling, and indirect access together make difficult the reliable and accurate inference of the changes in population size due to environmental perturbations. Here, in the context of antigen-perturbed immune cell repertoires, we present a generative model of observed sequence count pairs. For pairs of replicates, our model captures the natural variability in the system, giving reproducible behaviour across donors. Using the replicate model as a baseline, we then learn the parameters of a prior distribution of the ratio of a clone's frequency pair for pairs of repertoires sampled at different time points. After validating the model on synthetic repertoire data, we infer the posterior distribution of this ratio for every observed clone. Finally, we use point estimates obtained from the posterior to identify candidate responding clones, which are confirmed in a subsequent functional assay. Our method can be used in the clinic to track disease-specific clones expanding or contracting after infection, vaccination, or therapy.