Single-cell RNA sequencing (scRNA-Seq) combined with pseudotime trajectory inference can illuminate the cell differentiation process of pluripotent stem cells. Currently, competing methods exist for constructing pseudotime trajectories based on expected topology (i.e. linear, cyclic, branching) and prior parameters (i.e. start cell, number of expected clusters). The researcher is left to judge which approach to select and no strategy exists that incorporates a statistical confidence measure in the predicted cell pseudotimes.
We performed controlled neural induction of human induced pluripotent stem cells (iPSCs) using dual- SMAD inhibition over six different timepoints (day 0-7) and carried out scRNA-Seq using the ddSEQ platform. We ran multiple trajectory inference (TI) methods as recommended by the Dyno R package, which has benchmarked over fifty methods for accuracy against a gold standard, scaling, and quality control. A consensus pseudotime was created using machine learning methods to combine top-scoring TI methods, followed by Gene Set Enrichment Analysis (GSEA).