search

aertslab / AUCell

AUCell: score single cells with gene regulatory networks
110 stars 26 forks source link

should the gene expression matrix be normalised? #15

Closed helianthuszhu closed 3 years ago

helianthuszhu commented 3 years ago

Hi, Thanks for providing this wonderful tools. I am just wondering what format of the inputted gene expression would be better, raw count, or normalised one, such as log2(count+1), log2(TPM+1).

Thank you very much.

Best, Xiaoqiang

s-aibar commented 3 years ago

It doesn't really matter much. The AUC score is calculated for each cell individually.

lucygarner commented 3 years ago

@s-aibar, I have just put a related question in pySCENIC. Specifically, I am wondering about activation datasets where the composition of the RNA will vary greatly between the activated cells and the unstimulated cells. In this case, the rankings of non-activation-associated genes would be reduced in the activated cells, due to the upregulation of a whole host of other genes, however their absolute expression might not actually decrease. I am seeing a lot of downregulated TF regulons in my dataset, and I am wondering whether some of these might be an artifact of this effect?