[BioC] How to compute the stabilized variance in DESeq?

[Simon Anders]

Dear Peng

On 04/09/2012 11:46 PM, Peng Yu wrote:
> Hi Simon,
>
> I see the resVarA&  resVarB columns, which is the ratio between the
> stabilized variance and the actual variance. To get the stabilized
> variance, is the following R code correct?
>
> var(values_A)*resVarA

No, for several reasons:

- assuming that values_A is a matrix, 'var' gives you a single value 
computed over all values. You want 'rowVars'

- resVarA/B was the ratio of per-gene variance estimate and fitted 
variance. Both are estimates, neither is an 'actual' value in the sense 
of a population value. You might be confusing things with the 
'varianze-stabilizing transformation', because there is no quantity in 
DESeq that we call the 'stabilized variance'. Hence, I don't know what 
you are actually looking for.

- You are using an outdated version of DESeq. In the current version, 
the resVarA/B columns have been removed. See the new vignette for 
details and reasons for this.

See also the description of the new 'fitInfo' function in the vignette, 
which might give you what you want.

   Simon