289W Poster - Population Genetics
Wednesday June 08, 9:15 PM - 10:00 PM

Authors:
Chibuikem Nwizu ¹; Ava Soleimany ²; Lorin Crawford ^1,2

Affiliations:
1) Brown University Providence, RI; 2) Microsoft Research New England, Cambridge, MA

Keywords:
Theory & Method Development

The recent innovations in single-cell sequencing technology have allowed researchers to uncover biological mechanisms that drive phenotypic variation between both healthy and disease populations. These data can be collected at different time points, enabling the profiling of pertinent disease and developmental processes under various micro-environmental perturbations. Many statistical methods have been developed for modeling time-series data with the key assumption that repeated measurements are taken from the same unit over different time points. However, this assumption breaks down in single-cell analyses due to the practical inability of sequencing the same unit (cell) twice. In this project, we explore a novel strategy to overcome this limitation by taking a population genetic perspective on longitudinal single-cell RNA-sequencing (scRNA-seq) studies. Methodologically, we present a novel dynamic, Bayesian hierarchical Dirichlet process model (SCoOP) where, rather than making statistical inferences about individual cell trajectories, we seek to model the trajectories of phenotypic clusters that are present among the profiled cellular populations. By adopting this approach, we can both identify evolving cell population-specific signatures of disease and capture significant mechanistic shifts. To demonstrate the utility of our approach, we study a real data set of time-series scRNA-seq collected from patient-derived organoids of pancreatic ductal adenocarcinoma. Here, we highlight our ability to observe and estimate the strength of phenotypic selection on cellular clusters over time as a response to micro-environmental perturbations.