[BioC] EdgeR LogCpm and LogFC values calculation

[Gordon K Smyth]

On Sun, 2 Feb 2014, Koen Van den Berge wrote:

> Dear Gordon,
>
> Thank you for your reply.

> Since the LRT table only gives a LogCPM for the average of each gene, my 
> desire was to calculate this for each sample separately, for all genes. 
> In this way, I wanted to create a dot plot - as added in the appendix - 
> which could visually show the changes in expression values for the genes 
> we have found to be significantly differentially expressed (note I had 
> three samples in the control group and four in the treatment group). 
> This is my motivation in reaching out to the individual LogCPM values 
> and since these did not coincide with the ones provided by the table, I 
> was worried.

> Since the fitted model reproduces a single coefficient for each 
> treatment group, I guess it will not be possible to calculate 
> sample-wise log2 cpms that would be similar to the ones provided by the 
> table, and maybe I should stick to my descriptive log2 cpm plots as 
> provided in the appendix?

Yes.  As you say, the fitted model doesn't estimate sample-wise values. 
You need to plot the data, not the model.

Gordon

> Thank you in advance,
> Koen
>
>

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}