search

ImmuneDynamics / Spectre

A computational toolkit in R for the integration, exploration, and analysis of high-dimensional single-cell cytometry and imaging data.
https://immunedynamics.github.io/spectre/
MIT License
56 stars 21 forks source link

No appropriate meta clustring with run.flowsom #154

Open Tenrolf opened 1 year ago

Tenrolf commented 1 year ago

Dear all,

I am running into an issue with the metaclustering from the run.flowsom function. Whether using an automatic number or setting a number, I get one big metacluster with 99% of cells, and a few other metaclusters with a few cells.

However, the function prep.cytonorm lead to an appropriate metaclustering and the same data are well separated when using umap (with the same clustering channels).

Would anyone have an idea what could go wrong?

Thanks in advance!

R version: 4.2.2 Spectre: 1.0.0 CytoNorm: 0.0.15 flowSOM: 2.2.0

Tenrolf commented 1 year ago

An update to say that I could mainly solve the problem: I had an issue in the normalization distorting the data for a few cells from a few files, so these very few cells were so different that they were messing with the meta clustering.

However, I'd be curious why the meta clustering from prep.cytonorm was not impacted in the same way? (I also used it on the normalized files as a test).

tomashhurst commented 1 year ago

Hi @Tenrolf , that is a weird finding -- thanks for pointing it out! Especially the difference between run.flowsom and prep.cytonorm. Will have a dig around and see where the difference is coming from.

Tom

denvercal1234GitHub commented 1 year ago

Hi @Tenrolf -- How did you assess to detect that the issue was because of the normalisation distorting the data? Would you mind showing the bit of codes you used to examine this?