[BioC] weird model design for DE analysis

[Michael Love]

hi Lorena

Can you clarify on a few points below:

On Fri, Jul 26, 2013 at 1:29 PM, Lorena Pantano
<lorena.pantano at gmail.com> wrote:
> Hi,
>
> I have some doubts about my data.
> I would like to do DE analysis of RNAseq data.
>
> I have two set of experiments done in two different time points, so there
> is probably batch effect: let's say batch A and B.
>
> For A, I have 7 controls and 7 cases (clinical identified)
> For B, I have the same 7 controls and other 7 pre-cases (preclinical
> identified)
>
> I wanted to know if I can analyze all together, but I have questions like:
>
>
> 1-7 cases from A are only in batch A, and the others 7 are only in batch B,
> is it correct to setup a blocked variable indicating the batch effect
> although I only have one entire group for one batch, and another entire
> group for another batch?

You can still use blocking. Because the control group spans batch A
and batch B, you can still estimate the batch effect (the model matrix
should still be full rank).

> 2-and 7 control for A and B are same people, so it is not quite right to
> merge all together because it would be more like technical replicates.
>

I don't understand this last sentence. Do you mean that some, but not
all, of individuals in the control group from batch A are also in
batch B?

I wouldn't say they are technical replicates if the samples are drawn
from the same individual. RNA-Seq experiments of the same individual
will contain biological variability.

> My goal is:
>
> 1-get DE genes that follow additive effect. Something like increase a
> little in pre-cases, and increase more in cases.
>

It sounds to me like you might want to treat pre-cases and cases as
separate levels of a factor variable, rather than assume that the
trend will necessarily be: control < pre-case < case. But to receive
more specific advice, you should probably explain more about the share
of individuals across the 28 samples.

Mike