I have single-cell count data where rows are genes and columns are cell identifiers (e.g., Pig1ATCGCTGATG). I want to create a custom signature matrix using the scRNA-seq option in CIBERSORTx.
The interface states that "each cell must be assigned a cell phenotype," which I interpret as needing a metadata file to map cell identifiers to their respective cell types. However, the scRNA-seq option only allows uploading a "single cell reference matrix file" without an option to specify a metadata file.
When I tried passing my count matrix directly, the resulting signature matrix had columns labeled as numbers (0, 1, 2, ...) instead of cell types.
Questions:
- How do I properly annotate my count matrix with cell types so that the signature matrix output includes the expected cell types?
- Should I be using the RNA-seq option instead of the scRNA-seq option for this purpose?
This would create repeated columns as different cells can be in the same cell type. I'm guessing this is fine for CIBERSORTx?
I use scSort to annotate the gene expression matrix. The matrix from scSort has cell types in the first column (potentially duplicated), and gene names in the first row. When I pass this in the job result from CIBERSORTx still doesn't show the cell types. It should show something like this:
Again, I am using the scRNA-seq option and I am not sure if this is the right option, which is a part of my original question. Is this the correct option jared.andrews07 ?
If you want to use scRNA-seq data as a reference, yes, you should be using the scRNA-seq option. Replicates are expected, and allowed even for bulk reference samples if I remember right.
I don't know what scSort is or does, but that format sounds wrong. Look at the CIBERSORTx examples. Rows should be genes, columns should be cell type of each cell (repeats are fine). Values should be counts.