Mageck RRA is listing negative controls as high fold change in the gene summary

Question

[liz.b]

I've been running MageckRRA and providing a list of negative control sgRNAs using --norm-method control and --control-sgrna negative_ctrls.txt

In my gene_summary file I see these negative controls appearing as the top high and low fold change genes. For one of the sets of samples I ran RRA on the log file includes a lot of 'Skipping gene ... for permutation ...' messages, but for another set of samples it apparently didn't do this.

My colleague said the negative controls should not appear in the gene summary. Are they correct? Is there something I need to change to get the negative controls out of the results? Is this likely to be a code issue or an issue of poor quality data?

Thank you!

[ATpoint]

Please provide code and output examples, textual descriptions are hard to debug. IIRC when I used MAGeCK RRA for shRNA screening it did not return controls in the output. But I wanted to also see the stats for the negative controls, so I duplicated negative controls and added a "_1" to the duplicates. The original names I provided to the normalization --norm-method control --control-sgrna parameters, so normalization and permutation distribution where based on them. Still the duplicated ones would come back in the output so I could double-check that most controls were not called as significant. But yes, "normally" they should not be in results.

[liz.b]

Sorry! Will include code next time!

Fyi, my current solution is to stop using RRA and just also use MLE on direct sample to sample comparisons.

Answer 1

[jared.andrews07]

Most likely, your negative control sgRNAs specified are using the wrong IDs or in the wrong format somehow.

If provided appropriately, you should see messages in the log like:

INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg09 for permutation ... 
INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg05 for permutation ... 
INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg10 for permutation ... 
INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg04 for permutation ... 
INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg01 for permutation ... 
INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg02 for permutation ... 
INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg08 for permutation ... 
INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg06 for permutation ... 
INFO  @ Wed, 28 Sep 2022 16:48:26:   Skipping gene neg07 for permutation ...

And:

INFO  @ Wed, 28 Sep 2022 16:48:26:   Total # control sgRNAs: 51

And yes, the control guides/genes should not be in the summary files if done properly. I've found it useful to compare the normalized counts for the negative control guides/genes between timepoints to assess how consistent they are.

[jared.andrews07]

You may also be interested in a recent package that I got accepted into Bioconductor to visualize and compare MAGeCK results - CRISPRball. At worse, it'll help you make pretty pictures.