Leveraging SingleCell Data to Define Cell Differentiation Transitions
Vanderbilt University, Nashville TN
Investigators
Linked publications, trials & patents
Abstract
Phenotypic cell transitions are integral to tissue development, regeneration, and homeostasis, where a single potent cell type can give rise to more than one functionally distinct cell type. To better understand the dynamics of cell fate decisions, a lineage topology, from common progenitor to each differentiated cell state, must be known. While signaling profiles of cellular end states, stem and differentiated states, are generally well characterized, the signaling dynamics along each trajectory and at branch points, splits in a common lineage, are less defined. Recent computational approaches have focused on defining stemdifferentiated cell transitions from single cell data. Defining the temporal progression of cell transitions is challenging due to the continuum of cell states that exist between stable states. While previous efforts have successfully mapped linear differentiation lineages, no approach exists that can map branched differentiation lineages with statistically testable results. Our method borrows concepts from graph theory, where a spanning tree is constructed from a densitydependent down sampled set of datapoints. Lineage branches and cellular end states (stem cells and differentiated cell types) are identified by closeness, a graph attribute. The resulting transition maps are statistically scored both globally, on the overall topology, and locally, on each branch point, using a modified rootmean squared deviation algorithm. A final transitional map comprises a set of top scoring lineages which can be validated against the spatialtemporal cell progression of the crypt. Our hypothesis is that a cell makes its fate decision based upon a probabilistic distribution conditioned on its current state, its age, and the state of its surrounding neighbors. From hyperplexed imaging technology, we will model probabilistic cell fate decisions using a Bayesian framework to construct topologies of signaling dependencies. The mammalian intestinal epithelium is our model system, where its short turnover time necessitates continuous regeneration and differentiation of epithelial cells. Once established, our model will provide an unique insight into the effect of disease on stemdifferentiated cell transitions, paving the way for new systems based therapeutics.
View original record on NIH RePORTER →