1
$\begingroup$

Is there a good definition of pseudotime? Some tools are clearly labeled as pseudotime and produce values along a trajectory, but there are more complex approaches that involve branching and disconnected populations. There are 70 of them in the dynverse, so there are many options.

UMAP will also order cells, but that is generally not considered pseudotime. You can even overlay Monocle or PAGA results onto a UMAP so the cells are still ordered the same way. SPRING is specifically designed to capture continuous trajectories, but is also not considered pseudotime. Why not?

$\endgroup$
1
$\begingroup$

Pseudotime is the order of cells along a trajectory which in the simplest case is a minimum-spanning tree based on a (low-dimensional) manifold such as PCA or UMAP. Without a defined trajectory you have no pseudotime, and therefore UMAP is not pseudotime, but in fact can be used to define the trajectory.

$\endgroup$
2
  • $\begingroup$ To simplify: a PCA/UMAP plot is not pseudotime, but if you draw a line over it, that would be pseudotime, right? All the cells are still in the same positions, so they still show the same relationships. Adding that extra coordinate makes it pseudotime? $\endgroup$
    – burger
    Feb 8 at 15:01
  • 1
    $\begingroup$ Data scientists would argue that the "line" would need to come from a very elaborate machine learning approach, but yes, basically a line through the data makes it pseudotime. An accurate prediction of the root is critical though to give it some biological meaning, so the cluster of cells that most likely are the origin of a trajectory, e.g. stem cells. $\endgroup$
    – ATpoint
    Feb 8 at 17:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.