Cancer begins as a disease of the genome, with DNA mutations initiating a cascade of events that lead to cancer progression. As single or small collection of cells undergo state transitions to become cancer cells and ultimately evolve into a malignant neoplasm, the immune system is activated and new vasculature is formed, involving non-cancerous cells in the system. This process involves the flow and transfer of information across multiple scales in time and space. Information is encoded within and transferred between cells and across multiple genomic scales may be detected at the system?s level. Our hypothesis is that information contained in one or multiple genomic landscapes can be used to detect oncogenic perturbations and predict response to therapy. It has been shown that mutations associated with AML can be detected years before the onset of disease, however, they do not predict when the disease will manifest or response to treatment. Nevertheless, these sets of mutations can be characterized by distinct gene expression signatures collectively representing perturbations underlying the observed clinical phenotypes. Thus, there is an urgent need for novel and insightful interrogations and predictions of high-dimensional genomic data sets on a system level. Our approach aims to 1) make use of the maximum amount of relevant information in the system 2) be simple and parsimonious with the data, and 3) provide insight and predictions. We propose to validate a mathematical model and approach that considers genome-wide gene activity as state transition from a healthy state to a cancer state from the perspectives of messenger RNAs (mRNAs; transcriptome), non-coding microRNA (miRNAs; the miRome), and DNA methylation (epigenome). The theory and mathematics of state transitions is well known in the systems biology community and is a powerful tool for interpreting and predicting the behavior of complex systems such as genomics and cancer biology. The central hypothesis of this proposal is that information produced during a biological process such as cancer, can be detected from different viewpoints (i.e., transcriptome, miRome, epigenome) such that information contained in one viewpoint of the genomic landscape can be mapped into another, and that disease development and progression can be interpreted and predicted with mathematical models of information flow in a multidimensional genomic space. We propose the following aims:
Specific Aim 1. Parameterize a mathematical model of multi-dimensional state transition.
Specific Aim 2. Quantify the impact of treatment on state transition dynamics and develop a model of therapy response and relapse. We will quantify and model therapy response in controlled AML mouse model.
Specific Aim 3. Characterize the information contained in the transcriptome, miRome, and epigenome state-spaces. Impact. Through an iterative dialog between biological experiments and mathematical modeling, this work will provide insight into perturbations contributing to leukemia initiation and progression, which will guide the design of new therapies targeting pathways at critical transition points.
Personalized medicine based on patients' genomic information holds incredible promise for selecting less toxic and more effective cancer treatment. However, there are significant limitations with existing approaches to analyzing the large volume of complex and changing genomic data over the course of disease progression. The goal of this research is to develop and validate a mathematical theory and geometrical method to interpret changes in multidimensional genomic data over time that may eventually be used to predict cancer evolution and response to therapy in the clinic.