Questions regarding input data processing and DDRTree output

[asmagen]

Hello,

Regarding the input and processing; if I use importCDS to import a Seurat object, does it matter what normalization and regression are performed via Seurat (regress out nUMI etc)? Or does your method take only the raw data?

Regarding the DDRTree output:

1. What does the max_components parameter in reduceDimension control? The number of clusters or splits?
2. What does Component1/Component2 in plot_cell_trajectory output correspond to after using the DDRTree method? I believe they are not the top 2 PCs because the structure of the points follows and matches the trajectory very well. How is it that the cells match the tree lines?
3. How can one interpret the distances between trajectories in plot_cell_trajectory output? Do bigger distances between branch splits correspond to bigger dissimilarity? and similarly longer branches correspond to possibly more differentiation process?

Thanks

[McStber]

Hi asmagen ;)

Concerning your question for the max_components parameter in reduceDimension control:

As you might noticed after increasing the number for "max_components" Monocle will generate more branch points, more states and trajectories. This will group your cells into better splitted trajectories. You have to choose a max_comp = X value which fits the best for the differentation of your cells (that means what makes sense?!...checking also literature). I would recommend to choose a value which generates a trajectory plot where the differents states do not lie too much over each other (this happens when you choose a high value)!!

I would be also very interested if someone can answer the other questions!

Best, Michael

[asmagen]

Thanks. Any ideas here anyone?

[ruqianl]

I had a quick look at the code for function importCDS, it grabs the raw.data from the Seurat object.

[yingyonghui]

Hi @MichaelStber ,

I now have some questions with the max_components parameter, and I wonder if you could give me some advice.

I run the reduceDimension() with max_components=2 and get 2 branch points. And what I want to know is how can I adjust the parameters to get a trajectory plot with less (0 or 1) branch points?

Any suggestions would be appreciated! Thanks!

Best, Yingyonghui

[McStber]

Hi @yingyonghui ,

Sorry for late response! To be honest I don't know if this wasn't a coincidence that with max.comp = 2 you also ended up with 2 branch points. The number of branch points depends from my observation on many factors like the max. components, the number of dimension you used, the qvalue and of course also in general the number of cells. If you want to have few branch points, I would go down first with the max.comp.

Question: Which cells are you studying since you aim for even no branch points? This would mean there would be only one direct biological process for your cells, with only two states or even only one state???

I hope I could help you a bit.

Best, Michael

[yingyonghui]

Hi @MichaelStber ，

Thanks for your reply. Actually I'm focusing on the intestine stem cells from the mouse, and they may derive from different clusters as defined in Seurat. There is only a small number of cells in each cluster and they are sequenced at a time series. I just want to see if there is a clear trajectory among these cells across the time points.

Maybe I should try other parameters you mentioned.

Thanks!

Best, Yingyinghui

[chenzhuohui588]

Hi @asmagen, I encounted the same problem as you did. I want to extract the component 1/2 data of each point after the DDRTree dimensional reduction. The problem is that I can't find the corresponding data in the CDS. Have you figured out how to export the Component 1/Component 2 data of each cell?

Any suggestions would be appreciated! Thanks!

Best, Zhuohui Chen