Data matrix with single-cell CNV calls

[elisafazz]

Hello,

This issue is similar to the one found in #184.

My ultimate goal is to know the CNV calls for each individual cell. If this is not possible, then I can use a matrix that holds the subcluster assignment for each individual cell. I imagined that this might be in the slotted object or infercnv.observation_groupings.txt, but I could not find the information that I needed. Can you please help with this? Please let me know if I can provide further clarification.

Best, Elisa

[GeorgescuC]

Hi @elisafazz ,

There are a few options to obtain that information. If you are interested in the HMM predictions you can find the following:

• "infercnv.17/19_HMM*.observations.txt" from steps 17/19 are text file matrices of cells by genes that contain the states predicted by the HMM before (step 17) and after filtering using the Bayesian network (step 19).
• "infercnv.20HMM*.observations.txt" is similar to above, but rather than HMM states, the values in the matrix are the proportion of copies compared to the normal diploid state (so 0.5 means 1 copy only, 1 means both copies are present, see link above).
• "*.inferncnv_obj" are backup R objects (use readRDS() to load them) that contain the same matrices as above in the infercnv_obj@expr.data slot for steps 17/19/20, but also exist for the residual expression results. You can find the subclustering information in the infercnv_obj@tumor_subclusters slot.
• "*.cell_groupings" are tab formatted tables that show the subclusters content with each cell.
• "*.pred_cnv_genes.dat" are tab formatted tables that show for each subcluster the regions identified as CNVs, the state (steps 17/19) or proportion of that CNV (step 20), and the genes contained in the CNV region.
• "*.pred_cnv_regions.dat" are tab formatted tables that show for each subcluster the regions identified as CNVs, the state (steps 17/19) or proportion of that CNV (step 20), and the chromosome and positions the region spans.

I hope this helps you get the information you are interested in, but if you have more questions, please let me know.

Regards, Christophe.