How to interpret the nine-square plot from mageckFlute

Question

[Assa Yeroslaviz]

I am working on a CRISPR data set and have run the mageckFlute workflow with the MLE algorithm.

But I am not sure how to interpret the nine-square plot created by the SquareView function.

From reading the paper it somehow means, the genes in color shows a difference between the control and the treamment. But what does it mean here?

I have control samples from day 0, control samples from day 7 and treated samples from day 7.

my design matrix was as such:

sampleName  baseline    d7ctrl  d7sen
d0ctrl1     1   0   0
d0ctrl2     1   0   0
d0ctrl3     1   0   0
d7ctrl1     1   1   0
d7ctrl2     1   1   0
d7ctrl3     1   1   0
d7sen1      1   0   1
d7sen2      1   0   1
d7sen3      1   0   1

I compared my control and treatment after seven days against my day 0 samples.

Now, how can I interpret the purple genes? Are these genes the ones which show the highest difference between my treatment and my day 0 samples?

BTW, can this be done without mageck-VISPR? Can I use the normal mageck mle output for this analysis?

thanks

Assa

Answer 1

[jared.andrews07]

Light purple means unchanged in treatment samples and decreased in control samples. Dark purple means unchanged in treatment samples and increased in control samples.

Orange means decreased in treatment but unchanged in control. Green means increased in treatment but unchanged in control.

[ATpoint]

That having said, MAGeCK MLE (which FLUTE runs) calcualtes a "representative" β value for each gene as an aggregate of all shRNA/sgRNAs per gene. This plot is simply the ratio of these values comparing two conditions, so it's a way to classify the genes. You can now go and pick genes with larger/smaller β in ctrl or sen, as it might be some interesting biology that the knockdown affects one condition more/less than the other one. The colors and thresholds are arbitrary, it comes down to the ratio of the β values. So yes, you can run MLE directly, in my understanding FLUTE is mainly a wrapper to automate some plots, it does not fundamentally bring in anything new other than maybe the function to normalize the β values between conditions to compensate for different proliferation rates between cell lines (or whatever you used for the experiment) (see paper for details), despite that one could manually do that with something like quantile normalization as well. In my understanding there is no elaborate statistics in this plot, it is really just a way to visualize the differences between conditions.

[Assa Yeroslaviz]

thanks for mentioning it again. Yes, I understood, that FLUTE is just a wrapper for the MLE results, giving some nice visualization possibilities. I know one can play with the thresholds (like in any other differential expression analysis workflow ;-) ), but I wanted to be sure about how to interpret the results i do see.

[Assa Yeroslaviz]

thanks for clarifying this.