High-throughput single-cell gene expression experiments can be used to uncover branching dynamics in cell populations undergoing differentiation through pseudotime methods. Researchers from the University of Manchester have develop the branching Gaussian process (BGP), a non-parametric model that is able to identify branching dynamics for individual genes and provide an estimate of branching times for each gene with an associated credible region. They demonstrate the effectiveness of the method on simulated data, a single-cell RNA-seq haematopoiesis study and mouse embryonic stem cells generated using droplet barcoding. The method is robust to high levels of technical variation and dropout, which are common in single-cell data.
Haematopoiesis gene expression, showing the BGP fit for the MPO gene
a The Wishbone branching assignment is shown for each cell along with the global branching time (black dashed line), the most likely branching time (blue solid line) and posterior branching time uncertainty (magenta background). The sample of cells used to fit the BGP model is shown with larger markers. b The posterior cell assignment is shown in the top subpanel. In the bottom subpanel, the posterior branching time is shown. Pseudotime is shown on the horizontal axis of all plots. a, b Gene expression is depicted on the vertical axis. c The posterior branching probability BGP branching Gaussian process
Availability – An open source Python implementation of BGP is available at https://github.com/ManchesterBioinference/BranchedGP