Duplication and loss signals in the reference cells

[hxj5]

Thanks for developing this great tool.

I am analyzing a public 10x scRNA-seq dataset of a human gastric cancer sample. When I test inferCNV (v1.18.1) using "B cells" as reference cells, there are both duplication and loss signals in them.

The inferCNV wiki page points out that it could occur in a cluster of human MHC genes on chr6.

My question is if it happens elsewhere (not on chr6), as in the test results, what are the reasons, e.g., marker genes, heterogenicity of the ref cells, or highly variable genes? Is there any method to remove the pattern from ref cells? Thank you!