Background expression detected by DGE model: e.g. HBA genes in liver, hemoglobin genes. Picked up by the "organ" term, but still identified as organ*celltype specific for certain celltypes. Possible causes:
Certain samples have high background and those celltypes are over represented in these cell types: (1) check whether the expression of these genes is indeed sample specific, if yes (2) do soup correction w cellbender/soupX (Ask Ling, she has been running for many of these samples)
Certain celltypes are closer to tissue structures that might contaminate e.g. closer to portal vein in liver, higher chance to have blood contamination
Alternative solution: filter based on organ specific coefficient
Is soup affecting signature-level analysis? try computing a "soup" expression signature, from organ term
Background expression detected by DGE model: e.g. HBA genes in liver, hemoglobin genes. Picked up by the "organ" term, but still identified as organ*celltype specific for certain celltypes. Possible causes:
Is soup affecting signature-level analysis? try computing a "soup" expression signature, from organ term