Rethinking scRNA-seq Trajectories in Phylogenetic Paradigms: Overcoming Challenges of Missing Ancestral Information

In recent decades, many bioinformatics tools have been developed to reconstruct trajectories of biological processes, e.g., cell differentiation, using single-cell RNA-sequencing (scRNA-seq) data. Most tools tacitly assume that a cell’s ancestral transcriptomic profile can be approximated by means of its neighboring cells in an embedded gene expression space. However, many scRNA-seq datasets lack ancestral information due to missing early or transient states at the time of sequencing. We introduce CellREST, a bioinformatics tool that reformulates trajectory reconstruction as a phylogenetic inference problem. It infers trees linking cells that are assumed to share a common ancestral expression state. Using maximum likelihood tree inference, CellREST uncovers multiple different aspects of the transcriptomic landscape underlying a single scRNA-seq dataset, which can be visualized and combined into a single-cell network. We showcase CellREST’s performance on simulated and experimental scRNA-seq data and recover circular processes as well as cell type converging differentiation scenarios. By introducing and adapting phylogenetic concepts, CellREST provides a framework for interpreting transcriptomic relationships between cells within scRNA-seq data.