How to analyze these two figures

[capricy]

The figures here showed the gene expression level (log scale) of two female individuals (top) and two male ones (bottom).

Each spot represented a gene and each axis represented an individual.

The figures suggested that, generally, genes with lower expression levels were more scattered.

My question is:

If the bottom figure (male expression profile) was considered to be the expected trend, how do I find the genes in the female (top figure ) that demonstrated more variable expression between two individuals?

Thanks for any inputs!!

[TiborNagy]

I think Wilcoxon test can be used in this case. You apply this test in every gene and check if there are significant differences between groups. Of course do not forget about multiple test correction.

[dpryan]

@TiborNagy: Testing for rank changes doesn't make sense if you're interested looking at changes in variability. Presumably Capricy is looking for genes with the same average expression between genders but increased variance in females or males.

@Capricy: There are a number of methods out there for trying to measure changes in variability, but I don't think any of them will work well with N=2 per group.

[gringer]

I think it's more appropriate to demonstrate differences like this with a MA plot:

https://en.wikipedia.org/wiki/MA_plot

https://upload.wikimedia.org/wikipedia/commons/6/63/Pre-normalization.MAplot.png

In other words, plot mean expression on the X axis, and difference on the Y axis. It's generally the case that low expression values will have more variation. I prefer using smoothscatter for this because it makes density changes easier to see than a flat scatter plot:

http://sphaerula.com/legacy/R/smoothScatterPlot.html

With respect to selection, you could use your male graph to generate a fitting curve for the female graph based on a running 95% percentile (or similar), and select the genes that reside outside the curve.

[capricy]

@gringer: Could you explain more about how to generating the fitting curve for the graph? Or point me to any script/package I can use?

Thanks

And I re-graphed as you suggested: logFoldchange vs log(mean). The plot was attached here.

The left one is the male and the right one is the female.