Necrosis has recently emerged as a programmed cell-death alternative to cellular apoptosis. Programmed necrosis is important in the pathology of a number of human diseases including myocardial infarct, inflammatory bowel diseases, stroke, and neurodegeneration. Death receptor mediated signaling can induce either apoptotic or necrotic cell death and thus represents an ideal system to understand the mechanistic origins of cellular decisions processes. This work uses novel mathematical modeling approaches and experimental data, to explore mechanistic hypotheses about cell death decisions between apoptosis or necrosis. We explore on results from the Zinkel lab that demonstrate that the pro-apoptotic protein Bid, inhibits and modulates Rip1-driven programmed necrosis. Given that Bid and RIP1 function within a complex network of biological signals, we explore the systems-level behavior of protein interactions that could regulate necrosis or apoptosis signals. We generate mathematical models, calibrated to experimental data, which describe biochemical interactions that can modulate apoptosis or necrosis outcome. Our apoptosis-necrosis reaction model (ANRM) extends previous apoptosis mathematical models with new necrosis pathway interactions to describe a comprehensive cell death mechanistic framework. ANRM can simulate cellular processes that lead to either apoptosis or necrosis cell fates based only on initial protein concentrations. The ANRM biochemical interaction topology has been calibrated to experimental data using parameters from multiple cells including Jurkat, and Hematopoietic Progenitor cells. We find that the best fits to experimental data occur when a regulatory step for Bid is accentuated. In our work, anti-necrotic activity of Bid required regulation with kinetic rate four orders of magnitude different than generic values thus implying, from a modeling perspective, that this interaction is fundamental to cell commitment to either form of cell death.

A critical goal in cancer research is to stratify patient sub-populations likely to respond to molecularly targeted therapies. A number of recent technological developments hold the promise to address this question more systematically. For example, advances in low-cost molecular profiling technologies enable characterization of thousands of tumor genomes. Moreover, pharmacogenomics projects have characterized sensitivity profiles of hundreds of cancer cell lines to genetic and pharmacologic perturbations.
In this talk I will describe novel advances in machine learning approaches to identify patient-specific therapeutic vulnerabilities through integration of heterogeneous genetic and functional genomics datasets. Approaches include advanced high-dimensional regression-based approaches; multi-task learning frameworks to infer common predictors across multiple functional screening datasets; and semi-supervised approaches for incorporating patient genetic profiles into pharmacogenomic predictive models.
I will also describe advances in collaborative technologies allowing distributed teams of researchers to evolve best-in-class modeling approaches. Examples include the recent collaborative analysis project involving over 400 researchers in TCGA consortium, as well as world-wide crowd-sourced “collaborative challenges” to assess the best performing models for problems such as breast cancer molecular prognostic modeling, somatic mutation identification from raw sequencing data, and prediction of gene essentiality from genome wide RNAi screens.

Organoid cultures of normal tissues, recapitulating 3D structure and multilineage differentiation, represent a promising method for the de novo initiation and study of cancer. Such organoid models combine the experimental tractability of 2D monolayer transformed cell lines with the accurate organ architecture of in vivo models. Here, we describe the propagation of primary organoids from diverse gastrointestinal tissues, and the successful in vitro conversion of these organoids to colon, gastric and pancreatic adenocarcinoma. Applications of organoid modeling to numerous questions in cancer biology will also be discussed.

As single-cell technologies expand, it is becoming clear that many cellular signaling events are very dynamic, necessitating a time-lapse approach. I will present our work on the single-cell kinetics of two kinases - ERK and AMPK - that play key roles in the response to targeted cancer therapies aimed at disrupting cellular growth, proliferation, and homeostasis. Induced by growth factor stimulation, ERK activity is a central controller of transcription factors involved in oncogenesis, including Myc, Fra-1, and Egr-1. We show that at the single-cell level, each of these factors interprets ERK dynamics differently, leading to a diversity of cellular states within a genetically homogeneous population. ERK pathway inhibitors, now being evaluated for use in multiple cancers, modulate ERK dynamics differentially and redefine the repertoire of cellular states in unique ways. AMPK responds to cellular energy deprivation, and we show that direct inhibition of glycolysis results in a strikingly regular modulation of AMPK activity and metabolic state. In contrast, PI3K and mTOR inhibitors, another key class of targeted therapies, lead to highly disordered disruption of metabolic dynamics. Together these findings underscore the concept that, despite the chemically specific of modern targeted cancer therapies, their usefulness may be limited by the highly variable kinetics that they induce within cellular populations, resulting in sub-optimal, heterogeneous responses.

In this presentation, I will discuss the use of genomics to determine optimal strategies for cancer prevention. The research I will discuss includes the interrogation of genomic data for women who have a family history of breast cancer and who face considerable uncertainty about which aggressive prevention strategies to pursue. Development of more accurate individualized risk-estimation tools may help women choose between standard screening, intensive screening, and prophylactic surgery. In addition, a better understanding of the biological processes that lead to familial breast cancer development may lead to better treatment strategies. Our multi-omic study and functional experiments support the concept that novel oncogenic processes are disrupted in women with a family history of breast cancer and play a role in FBC development. Further, our results represent a novel approach to produce individualized estimates of cancer risk and to identify disease-susceptibility mechanisms.

The development of personalized treatment regimes is an active area of current research in genomics. The focus of our research is to investigate core biological components that contribute to disease prognosis and development, and to develop latent variable models to accurately determine optimal therapeutic regimens for individual patients. To accomplish this aim, we have developed an adaptive Bayesian factor analysis model that integrates in vitro experimental data into our models while still allowing for the refinement and adaptation of drug or pathway profiles within each patient cohort and individual, efficiently accounting for cell-type specific pathway differences or any “rewiring” do to cancer deregulation. Our modeling approach serves an essential role in our attempts to develop a comprehensive and integrated set of relevant, biologically interpretable computational tools for genomic studies in personalized medicine. We are currently working on a variety of applications using data from cancer and pulmonary disease with the potential to be extremely important in treating patients with these diseases.

Non-Hodgkin lymphomas (NHLs) are a collection of over 30 cancers deriving from lymphoid cells. Diffuse large B-cell lymphoma (DLBCL) is a the most common NHL type and is clinically and genetically heterogeneous. Exome, whole genome and transcriptome sequencing in this tumour type has uncovered a plethora of common mutation targets and a long tail of infrequently mutated genes with potential relevance. Genome sequencing has also revealed myriad structural rearrangements and focal deletions affecting specific genes relevant to disease. Using these data, we have computationally dissected individual DLBCLs to identify multiple sub-clones and evidence for ongoing acquisition of driver mutations during tumour evolution. In this talk, I will provide an overview of the discoveries we have made into the molecular nature of DLBCL and other common NHLs using next-generation sequencing. I will also discuss the ongoing application of this knowledge to develop a suite of non-invasive assays for monitoring tumour dynamics in NHL using circulating tumour DNA.