Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

issue about calculation in tSPACE #7

Open
Ruismart opened this issue Mar 26, 2024 · 1 comment
Open

issue about calculation in tSPACE #7

Ruismart opened this issue Mar 26, 2024 · 1 comment

Comments

@Ruismart
Copy link

Hi Denis,

Thanks for the beautiful tool tSPACE, I have been trying to use it to build reasonable trajectories for a bunch of developmental single cell data.
Here I have got some issue:

  I first ran a small dataset with about 3k cells and 1.5k variable genes, it took 15h to run on a 64G local PC, the trajectory output seems pretty good.

  then I wanted to run a bigger one with about tens of thosands of cells and same parameters, but it was terminated by me after 100h without an end.

  then I chose to use the top PCs as input, though it could be completed in just a few hours, the tSPACE output result becomes very similar to my old UMAP calculated using the same PCs. It seems like the existing PCs have been determined a lot by custom pre-normalization/-integration. Additionally, if a few datasets have to run individually, it might be hard to keep the consistensy.

So my question is: if there is a way to extract the tPC formula, as getting PCA coefficient from seur.obj@reductions$PCA@feature.loadings ?
Then I could run tSPACE on a standard and relatively small dataset at first, then extract the formula for each tPC, after that, I could do the calculation using those pre-built tPC-formulas on any new and bigger datasets with similar celltypes and same pre-normalization.

Kind Wishes,

Shaorui

@Ruismart
Copy link
Author

Ruismart commented Apr 2, 2024

After getting more familiar with the method/code, I think it might be not easy (like, linearly) to label back source-genes/PCs on final tPCs through the distance matrix. I have been trying to think about another way to do the calculation considering pre- feature selection could really make a huge contribute to final trajectories.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant