Cancer is a disorder of unrestrained cell proliferation, but increasingly it seems that not all proliferating cells in a tumor matter equally. As with cells in normal tissues, tumor cells appear to progress through lineage stages, in which the capacity for unlimited self-renewal is, at some point, lost. The cancer stem cell hypothesis states that cancer diagnostic, prognostic and therapeutic efforts need to be focused on that population of cells-often a small minority-that undergoes long-term self-renewal. While this hypothesis acknowledges the existence of lineage progression in cancers, it is silent on the function that lineages normally serve. We recently found, through experimental and theoretical work, that a likely raison-d'etre for lineages is to provide a framework for powerful feedback control of growth and regeneration, through mechanisms that target the differentiation decisions of individual cells. For cancer to develop, such feedback control must be disrupted, and the natural history of most tumors suggests that it becomes disrupted progressively over time. Our studies indicate that what happens in a tissue when feedback is compromised can be very complex, yet still understandable and predictable. We argue, therefore, that from the details of how a tumor develops over time-size, shape, growth rate, stem cell fraction, etc.-one ought to be able to infer specific information about the kinds of control processes that operate (or recently operated) within the tumor and its surrounding environment. Such information can both provide insight into how different types of tumors develop, as well as patient-specific information about prognosis and the effects of therapy. The proposed project focuses on learning how to obtain such information from the observable properties of tumors. Three-dimensional mathematical models that incorporate various types of lineage progression, feedback, evolutionary processes, and therapeutic interventions will first be created, analyzed, and used to generate large numbers of simulations of solid tumor growth and progression. From these results, mappings from tumor properties to feedback and lineage architectures will be found through state-of-the art machine-learning algorithms. The ability of these mappings to reproduce and predict the behaviors of real tumors will be assessed using established animal models of breast cancer, in which luminescent and fluorescent imaging techniques are used to follow tumors, and their stem cells, over time. This will enable the validation of particular model architectures, or suggest methods for their refinement, and allow the determination of control strategies at work in tumors that can be exploited to provide a leap forward in both personalized medicine and cancer care. What makes this project a """"""""grand opportunity"""""""" is the pursuit of rapid progress through a highly multidisciplinary team that will draw on new advances in the areas of cell lineage behaviors, cancer stem cells, three-dimensional mathematical and computational modeling, and machine-learning.
Tumors arise when the feedback control of cell growth breaks down. We hypothesize that, within the details of how a tumor grows lie important clues about the nature of feedback processes-including those that still remain or may be re-activated. By describing how such clues can be found, we will be defining a new approach for predicting how individual tumors behave in cancer patients, and how they respond to different kinds of therapy.
