Pseudotime analysis of B cells

Clustering analysis is useful for grouping cell types based on similar gene expression patterns but does not capture information related to developmental trajectories of cells. To assess developmental trajectories, we first embedded cells in a low-dimensional diffusion map (e.g. performed non-linear dimensionality reduction⁵⁸. We then used the R package slingshot⁵⁷ to infer a pseudotime for each cell along the developmental trajectory, and to infer individual trajectories. To evaluate whether genes were statistically associated with pseudotime, we performed LOESS regression using the R package gam, where we fit gene expression as a function of pseudotime along each trajectory. We focused on the trajectory that was characterized by progression from naïve B cells to germinal center B cells.

For pseudotime analysis of germinal center B cells, slingshot could not be used since it assumes a linear trajectory. Germinal center B cells are in a cycle between light and dark zones, and therefore require pseudotime inference based on a cyclical process. Therefore, a principal curve was fit along the circular trajectory to infer the pseudotime of each cell in this process. Genes were once again investigated for their relationship to pseudotime and were clustered based correlation of gene expression over pseudotime. An R package called “circletime” and accompanying vignette were created to encapsulate the code necessary to generate all aspects of the cyclical pseudotemporal analysis.