A simple question about Gene expression and CRISPR Gene effect

Hi everyone! I’m an undergraduate doing research based on gene expression and CRISPR gene effect data in Depmap. When I combine the expression and gene effect data together, I found various genes that are zero expressed in most tumor cell lines, but have a wide range of gene effect. For example, gene USP17L10 is not expressed in most tumor cell lines, but have a gene effect range from -1 to 0.6. How can we explain this? Thank you ~

1 Like

Hi Jialin, like any screen, Achilles has a false positive rate. You can control false discovery by using the probability of gene dependency (CRISPRGeneDependency - sometimes labeled with “PR” in portal tools). However, in the case of USP17L10 specifically, I think the expression profile is wrong based on its expected gene functions, which imply ubiquitous expression.

Thank you for your explanation! I wonder if you have removed batch effect in Gene expression TPM data? Because these days I am aggregating gene expression data from different sources, the PCA results lead me to remove batch effects using ComBat() in sva package. Then I got TPM below zero in some genes. It is confusing ~

I agree with Jialin. It appears that this gene isn’t expressed in 99.9% of human cells. There’s a high probability that the gene effect of USP17L10 shown in the database is a false positive result.


This is an interesting case. I checked the gene expression of USP17L10 in TCGA and GTEX, and in the vast majority of tumors and normal tissues it is zero.


So the expression of 0 observed in cell lines is consistent with that observed in tissues and tumors.

It might be interesting to look at how the individual guides behave for this gene, to see if one is much stronger than the others, leading to the observed more negative dependency scores than would be expected.