This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Limma Differential Expression with Duplicate Correlation

Dear all,

After some attempts with DESEQ2, I have now moved to Limma to solve my issue with 3 samples per individual in my Differential Expression Analysis. I have to adjust my model for age, sample site and diagnosis. I am interested in differentially expressed genes in sex (male versus female).

I followed this tutorial under "Limma Analysis": https://bioc.ism.ac.jp/packages/3.14/bioc/vignettes/variancePartition/inst/doc/dream.html

I wonder if this code makes sense now, I dont quite understand why sometimes people put "~0 + AGE.GROUP + ..." What is the +0 doing, is it fine without?

# apply duplicateCorrelation is two rounds
design = model.matrix( ~ AGE.GROUP + SAMPLE.SITE + DIAGNOSIS + SEX, metadata)

vobj_tmp = voom( geneExpr, design, plot=FALSE)
dupcor <- duplicateCorrelation(vobj_tmp,design,block=metadata$INDIVIDUAL)

# run voom considering the duplicateCorrelation results
# in order to compute more accurate precision weights
# Otherwise, use the results from the first voom run
vobj = voom( geneExpr, design, plot=FALSE, block=metadata$INDIVIDUAL, correlation =dupcor$consensus)

# Estimate linear mixed model with a single variance component
# Fit the model for each gene, 
#dupcor <- duplicateCorrelation(vobj, design, block=metadata$INDIVIDUAL)

# But this step uses only the genome-wide average for the random effect
fitDupCor <- lmFit(vobj, design, block=metadata$INDIVIDUAL, correlation=dupcor$consensus)

# Fit Empirical Bayes for moderated t-statistics
fitDupCor <- eBayes( fitDupCor )

Thank you so much, Bine

duplicate differential analysis limma expression correlation
• 5,729 views • link
written by Nona • Nona

1 answer

The 0 in the design formula specifies whether to include an intercept term in the module or not. If included the interscept term generally represents the expression in the baseline condition, with other coefficients representing logfoldchanges from this baseline condition. With no intercept there is no baseline condition, and log2foldchanges are calcuated as the difference between coefficients for the two conditions.

So, as far as I'm aware, at least in standard fixed effect setting, to include a intercept or not is a question of how contrasts are encoded, not what the final results are.

• 0 views • link
written by i.sudbery i.sudbery

Thank you very much for this explanation.

• 0 views • link
written by Nona • Nona

Log in to answer this question.

More posts like this