search

dynverse / Waterfall

Unofficial Waterfall R package
0 stars 0 forks source link

Clarify pseudotime calculating #2

Open NicolasHipp opened 5 years ago

NicolasHipp commented 5 years ago

Hi everyone,

First, I would like to thanks you for the work on the pre-print paper and the shinny app which is really useful. I'm here about a method that you used in the benchmark : Waterfall. I know that you do not code the algorithm but as you used it, maybe you encounter my problem during your analysis.

I have a 560 cells and 23000 genes dataset (smart-seqV2), and I would like to infer the pseudotime during this differentiation process. I ran the first step of Waterfall using Seurat normalized values (Log-Normalize). When I plot the MST of this dataset I obtained this figure : mst.of.classification(seu,k=6,color=all.col) image

So the process seems linear and I expected that the pseudotime organise the cells based on that but I get that (both extremities represent start and end of my process based on specific markers) pseudotime.df <-pseudotimeprog.foo(seu,k=6,color=all.col,seed=1) image

Does-anybody get similar things during analyses ? Thanks a lot, Nicolas

rcannood commented 5 years ago

I'm not entirely sure what your problem is. I assume that in your last plot, you assume the red line to follow the data instead of go right through it? This is because the ps object is not really meant to be plotted. You can use ps$pseudotime, though, which is the ordering of the cells along the red line in the first plot you showed. Does this provide an answer to your question?

If you're specifically interested in linear TI methods, I would recommend looking at the overview figure (WIP; I still need to add legends..) and our guidelines (as a figure or a Shiny app). You can run the different methods all with the same interface using the dyno package :)

NicolasHipp commented 5 years ago

Hi, Thanks a lot for this answer, effectively I was hoping that the pseudo-time will follow my MST path. When I look for coordinates in my ps$pseudotime, ps<-mst.of.classification(seu,k=6,color=all.col) ; ps$pseudotime Cells are ordered by the line in the last plot, meaning that the brown cells (which are my T0 cells) are close to the blue and grey cells :/. which is impossible in the biology of the system.

I will have a look on the dyne package to evaluate if I find that every time !!!

Thanks again ;) nicolas