Custom cell typing as input

[Jwong684]

Hi clustSIGNAL devs,

I have been trying to test your tool out on my CosMx spatial transcriptomics dataset. The initial results I get from your pipeline seem pretty promising but I was wondering if there is a way for me to integrate my own cell-typing as input.

I noticed from your step-by-step pipe that clustSIGNAL::p1_clustering(spe, dimRed = "PCA") returns a column that will give initialClusters. What do these values necessarily mean? If I just replace these values with discrete names, ex. T-cell, malignant B cell, etc., would that be sufficient to run the rest of the pipeline?

Thank you!

[PratibhaPanwar]

Hi @Jwong684, thanks for trying out the method and for the positive feedback!

Yes, it is possible to use your own cell-typing with ClustSIGNAL through the step-by-step run. The custom clusters will need to be saved in the metadata of your SpatialExperiment object (colData()) and the column name would need to be 'initCluster'. Then the following steps, like neighbour detection, entropy calculation, etc., should run with no problems - essentially, you would be skipping the clustSIGNAL::p1_clustering() step.

The initClusters are obtained after we cluster all cells in the dataset, and then sub-cluster the clusters - we want to get down to small groups of cells that likely have very similar gene expression. These are just cluster labels so even if you replace them with custom cell type names, the following method steps should run fine.

The initClusters affect the entropy measurement and what space could be defined as homogeneous/heterogeneous. So, when using custom clusters as input, it would be good to ensure that the cells in each cluster have similar gene expression pattern, and that the clusters are reliable.