How can I overcome the error of "No residual degrees of freedom in linear model fits"?
1
0
Entering edit mode
@jurat-shahidin-9488
Last seen 18 months ago
Chicago, IL, USA

Hi:

I am experimenting preprocessed Affymetrix microarrays expression data matrix (Affymetrix probe-sets in rows (32830 probesets), and RNA samples in columns (735 samples)) for my downstream analysis. However, I have pheno metadata for the experiment observations.

data that I used:

> dim(eset_HTA20)
[1] 32830   735
> eset_HTA20[1:3, 1:3]
       Tarca_001_P1A01 Tarca_003_P1A03 Tarca_004_P1A04
1_at          6.062215        6.125023        5.875502
10_at         3.796484        3.805305        3.450245
100_at        5.849338        6.191562        6.550525

> dim(phenoDat)
[1] 735   7
> phenoDat[1:3,]
                       SampleID   GA Batch     Set Train Platform
Tarca_001_P1A01 Tarca_001_P1A01 11.0     1 PRB_HTA     1    HTA20
Tarca_003_P1A03 Tarca_013_P1B01 15.3     1 PRB_HTA     1    HTA20
Tarca_004_P1A04 Tarca_025_P1C01 21.7     1 PRB_HTA     1    HTA20
                      T
Tarca_001_P1A01 Preterm
Tarca_003_P1A03 Preterm
Tarca_004_P1A04 Preterm

my attempt:

I tried differential expression analysis on eset_HTA20 data, here is my workflow by using limma package:

ano <- phenoDat
esetHTA <- HTA20_rma_filt

ano$ID <- factor(ano$SampleID)
designs <- model.matrix(~0+T+SampleID, ano)
esets <- eset_HTA20[, rownames(ano)]
colnames(designs)<-substr(colnames(designs),2,100)
fit <- lmFit(esets, designs)
cont.matrix <- makeContrasts( contrasts="Term-Preterm",levels=designs)
fit2 <- contrasts.fit(fit, cont.matrix)
fit2 <- eBayes(fit2)

but when I ran fit2 <- eBayes(fit2), I got the following error:

> fit2 <- eBayes(fit2)
Error in .ebayes(fit = fit, proportion = proportion, stdev.coef.lim = stdev.coef.lim,  : 
  No residual degrees of freedom in linear model fits

I looked into support site here and Gordon explained several threads that this happened if data doesn't have replicate. I am not quite sure how to proceed with my analysis now. How can I know data have replicates or not? If there is not replicate, what's the possible solution to go through with this issue? can anyone point me out any possible strategy I could try here? any thought?

limma microarray error • 391 views
ADD COMMENT
2
Entering edit mode
Dario Strbenac ★ 1.5k
@dario-strbenac-5916
Last seen 7 days ago
Australia

I suppose that there is a distinct sample ID for each of your 735 samples. You don't want to have SampleID as a term in your model formula. You ought to include Batch, though.

ADD COMMENT
0
Entering edit mode

bingo, I just fixed this by taking your opinion, but why I should include Batch instead of SampleID? I am a newbie for differential expression analysis, so this is not crystal clear to me yet. Could you give extended comments on your proposal? I want to learn? what else I have to try for designs matrix? any idea?

ADD REPLY
0
Entering edit mode

To be blunt, this is not a set to use learn on. It's way too large and complex. If you don't understand why batch should be a category and sample ID should not, you need to practice with tutorials and much smaller sets so that you understand the basics of what you are doing.

ADD REPLY

Login before adding your answer.

Traffic: 457 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6