search

akdess / CaSpER

76 stars 27 forks source link

Control sample for scRNAseq data #40

Open laillern opened 3 years ago

laillern commented 3 years ago

Hello,

I am new to scRNAseq data, so please forgive my ignorance..

I was wondering how I should deal with controls when using CaSpER . I understand that a bulk RNAseq sample would have a matching control sample, but how does it work with single cells? How do you select control cells from the 2000 cells present in a sample? I have several drug resistant cell lines and a parental one. I am supposed to merge the counts for parental+ drug-resistant-sample1, and then indicate the controls as all the cells coming from the parental cell-line?

Thanks for any help you can provide!

akdess commented 3 years ago

Dear @laillern , I am very sorry for my late reply.

That is a very important question and not very easy to answer. You can first cluster the cells using Seurat and select control (might be non-malignant cells if it is a cancer study) cells using known cell marker genes. You expect to see the non-malignant cells to cluster together in TSNE or UMAP plot. Additionally the CaSpER large scale CNV calls should be neutral in non-malignant cells.

You can generate TSNE plot using RunTSNE function in Seurat and using FeaturePlot function in Seurat you can see the expression of non-malignant cell marker.

If you don't know any non-malignant cell marker, you can predict those cells by first clustering the cells using Seurat and annotating the TSNE or UMAP cluster as non-malignant if large scale CNV calls generated by Casper are all (or almost all, considering the false calls) neutral.

Hope it helps. Please let me know if you have any further questions. Best, Akdes