Expression variability between 2 groups of genes

[kintany]

Hi, I'm analyzing GTEx data (read counts, ~30 patients, one tissue) to check if expression variability is different for 2 groups of genes (I'm interested in one special set on genes). With DESeq2 I estimate dispersion for each gene ("dispGeneEst" in DESeq2) that represents variability of expression between samples after removing the Poisson variability from sequencing.

I would compare dispersions betweens group1 and group2 but the problem is that these two groups have different mean levels of expression and dispersion is normally higher for low expressed genes. So I would see high dispersion of expression for group1 only because genes in this group have lower levels of expression.

Could you please suggest what test can be used here? I tried to sample genes from group2 so that they have "matched" expression levels but I wasn't able to get a subset of genes from group2 that would have the same distribution of levels of expression (tested by ks test). Any suggestions are appreciated! I work mainly in R.

I also attached scatterplot visualizing dependance between dispersions and mean levels of expression for 2 groups of genes.

[Dario1984]

You have two groups of genes and every gene has a different length to every other gene. Longer with the same rate of transcription of shorter genes will have more counts. Also, different genes have different GC content, which you have not normalised for. The comparison which you have made is not logical.

[kintany]

You have two groups of genes and every gene has a different length to every other gene. Longer with the same rate of transcription of shorter genes will have more counts. Also, different genes have different GC content, which you have not normalised for. The comparison which you have made is not logical.

I do the analysis for dispersions of counts, not counts themselves. Length is not correlated with dispersion and has the same distribution for both groups. The same about GC content