Question: Limma model for paired and pool together
0
gravatar for b.nota
18 months ago by
b.nota330
Netherlands
b.nota330 wrote:

I would like to use limma (voom or trend) for my RNAseq analysis, and I am wondering how to correctly model my data. I have 2 different cells (hi and lo) from 2 donors, so I want to use a paired design. However, there is also a pool of more donors (hi and lo). How can I include this correctly in my design?

This here is my targets file:

Sample    Donor    Cell
D1_lo    D1    lo
D1_hi    D1    hi
D2_lo    D2    lo
D2_hi    D2    hi
Pool_lo    Pool    lo
Pool_hi    Pool    hi

 

rnaseq limma design • 268 views
ADD COMMENTlink modified 18 months ago by Ryan C. Thompson7.4k • written 18 months ago by b.nota330
Answer: Limma model for paired and pool together
1
gravatar for Ryan C. Thompson
18 months ago by
The Scripps Research Institute, La Jolla, CA
Ryan C. Thompson7.4k wrote:

The main problem I see is that the pooled samples would be expected to have lower biological variance, since they are effectively the average of several donors. You can model this sample heteroskedasticity using a design of ~Cell + Donor and then using voomWithQualityWeights. Make sure to set plot=TRUE and verify that the Pool samples have higher weights (i.e. lower variance) than the others. If this is not the case, you need to investigate further, since the data is not matching your expectations based on the design.

ADD COMMENTlink modified 18 months ago • written 18 months ago by Ryan C. Thompson7.4k
1

Agreed, the paired model ~Cell + Donor is correct here.

There's no heteroskedasticity issue here though because any variation between Donors is being removed by the paired model. The analysis of variation is purely within Donor, so the degree of variability in baseline expression from one Donor to another doesn't enter into the analysis.

So it's just an ordinary paired analysis without any special considerations (assuming that the pool doesn't include D1 and D2).

ADD REPLYlink modified 18 months ago • written 18 months ago by Gordon Smyth38k

The pool is made of 3 other donors, so this will work for me! Thanks both.

ADD REPLYlink written 18 months ago by b.nota330

I'm not sure I follow the logic that there's no heteroskedasticity issue. I would expect the pooled samples to have smaller residuals on average, although perhaps multiple pools would be required to show this effect.

ADD REPLYlink written 18 months ago by Ryan C. Thompson7.4k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 393 users visited in the last hour