baseMean in ggmaplot is not real A

[QiangShiPKU]

Hi, I noticed that the baseMean is (R+G)/2 (thus log2( (R+G)/2 ) rather than real A belows
https://en.wikipedia.org/wiki/MA_plot

[kassambara]

ok, thanks. the ggmaplot() expect a column containing mean and log2foldchange for creating the plot. No computation is made inside the function.

[QiangShiPKU]

ok, thanks. the ggmaplot() expect a column containing mean and log2foldchange for creating the plot. No computation is made inside the function.

The real A in MA plot is the mean of two log-scale values (log first, then mean), instead of log scale of the mean of two values (mean first, then log). However, if I am not mistaken, ggmaplot() uses the log scale of the baseMean column as X axis, that is not real A. If users supply only the mean of two values but not two original values, the real A cannot be calculated. So, I think this function should expect two original/log-scale columns to calculate their mean (real A) as X axis.

[kassambara]

thank you @QiangShiPKU for the clarification, I can see now the problem.

Let's keep this issue open

[kassambara]

The design of this function was influenced by the DESeq2 pipeline (https://www.bioconductor.org/packages/devel/workflows/vignettes/rnaseqGene/inst/doc/rnaseqGene.html#ma-plot), were log2 fold change is plotted against the basemean in log scale..

Obviously, I think that an option should be available in ggmaplot() to create a real MA-plot as you mentionned.

[QiangShiPKU]

Great

[kassambara]

I think that we should allow two possible formats for the input data:

1/
baseMean | log2FoldChange|padj

Here, we'll use log2(baseMean) as the x-axis variable (like in the current version of the function)

2/
baseMeanLog2 | log2FoldChange|padj

here, baseMeanLog2 is assumed to be the mean of logged values; so we'll use it as x-axis variable without any transformation

[QiangShiPKU]

That's a good idea

[kassambara]

fixed now, thanks