InferCNV clones identification

[Lloyd Bod]

Dear all,

I have scRNAseq of tumor cells and used InferCNV to analyze the mutational load and now I would like to identify the different tumor clones. 

I believe that a certain clone would be cells sharing CNVs on the same genes, right? Is there a way to get this information among the CNV files? 

Another question, is there a way to know the genes undergoing CNV per chromosomes? 

Thank you for your help,

Lloyd

[cgeo...@broadinstitute.org]

Hi Lloyd,

Clones should indeed share the same CNVs. Depending on what options you used to run inferCNV, there a couple of ways you can get that information.

If you did not run the HMM predictions, you can:

- process the residual expression matrix ( "infercnv.observations.txt" ) using cutoffs or criteria you set yourself.

- use the NGCHM ( https://github.com/broadinstitute/infercnv/wiki/Next-Generation-Clustered-Heat-Map ) to explore the results and interactively select regions of interest/groups of cells.

If you ran the HMM predictions, first, compare the residual expression view ( "infercnv.png" ) and the HMM predictions view ( "infercnv.19_HMM_predHMMi6.*.png" ) to make sure results are consistent. It is possible that results in the HMM view are not accurate if you have different clones within a single patient/sample and ran the HMM with the analysis mode set to sample. If that is the case, you should rerun your analysis in subclusters mode, which will subdivide your patient/sample in smaller groups and make predictions for each of them independently ( https://github.com/broadinstitute/infercnv/wiki/infercnv-tumor-subclusters ).

Once the results are consistent, you can access the information about each CNV region in the following way:

- which genes they contain ( "HMM_CNV_predictions.HMMi6.*.pred_cnv_regions.dat"  +   "HMM_CNV_predictions.HMMi6.*.pred_cnv_genes.dat" ) matched to a cell_group identifier

- the cell_group identifier are listed in "17_HMM_predHMMi6.*.cell_groupings" with a list of the cells they contain.

Regards,

Christophe.