Identification of modules of genes that vary over pseudotime (scRNA-seq pseudotime analysis)

Question

[Lucy]

Hi,

I am very interested to get the community's view on the best methods for identifying genes that vary over pseudotime and also modules of genes that covary over pseudotime.

I have used both the Slingshot and SCORPIUS approaches to look at T cell activation trajectories. However, I am interested in identifying the genes that vary over pseudotime, and in particular in identifying modules of genes that vary over pseudotime. I see some genes are that are activated early in the pseudotime pathway and others later, and I would like to group these into relevant modules.

I am aware of the module identification approaches available in Monocle3 and SCORPIUS and have tested both of these. However I find that they give very different results. I was wondering whether anyone has done a comparison of different gene module identification methods for time courses, and whether there is any literature out there on this?

Many thanks in advance for your thoughts and input.

Best wishes,

Lucy

Answer 1

[jared.andrews07]

You will likely be interested in dynverse, which did a really excellent comparison of different trajectory inference/pseudotime ordering methods.

I tend to use slingshot + tradeSeq with viz by dittoSeq to do this, though I usually pick specific pseudotime points corresponding to the middle of each developmental stage/point in time that I want to compare to each other. And just do that for each possible comparison to get a list of genes for each, then slap them together, and make a heatmap where cells are ordered by pseudotime and the expression for all of the identified genes are shown.

[Lucy]

Thank you - yes I have read the comparison paper and found it very useful. That's what led me to choosing Slingshot and SCORPIUS as good methods to try. Have you tested any methods for gene module analysis to identify genes that show similar patterns of expression over pseudotime - this is what I am really interested in.

[jared.andrews07]

I either do this manually via the method listed in my answer or tradeSeq, depending on how many lineages are of interest.