GeorgescuC
commented
10 months ago Hi @mohammadghorbani28 ,
What infercnv does boils down to is looking for groups of colocated expressed genes whose expression is significantly more or less expressed as a group than is seen in a set of reference cells. This is also the reason why we strongly recommend using as a reference a set of healthy cells of the same type as your (potentially) cancerous cells, because cluster of genes can be co-expressed as part of the normal healthy behavior of certain cell types. So if you were to compare such a cell type to a different one as reference, you would see a signal that could be wrongly interpreted as a CNV. An example of strong signal that does not associate with actual CNVs can be seen in immune cells for the cluster of MHC genes on chromosome 6 in human cells, where it is very common to see 2 distinct levels of expression in healthy cells (potentially associated with activation, but this is a guess that hasn't been verified). So yes you could technically use infercnv to look for certain behaviors in things besides cancer, but they might not be CNVs, just co-regulated genes. I hope this helps.
Regards, Christophe.
Please forgive my ignorance but I have a fundamental question regard to appropriateness of using infercnv package.
Everywhere I looking at, infercnv has been used in cancer studies since structural changes are expected. But can we use infercnv in other cases besides cancer? For example on a pbmc dataset where we already know there is an inflammatory status and according to infercnv heatmap, certain chromosome(which contain inflammatory related genes) shows a robust signal. So there is a result but is this meaningful at all? Is this due to a particular Biological state or actual structural changes?