Using limma on a very low ammount of genes
1
1
Entering edit mode
9 days ago
ijarne ▴ 20

Some days ago I recived a results table from a multiplexed ELISA array. I decided to use limma to work with the data as I could see that the data at the end was quite similar at the scale level to a microarray after checking it with the colleague that has produced the data which has certain expertise in the technique . Nevertheless I do only have a total of 22 cytokines to test whether if they are or not deferentially expressed. At this point I am working with the same code I usually use to perform microarray analysis.

Could anyone tell me wether if there better ways to analyze data in limma when having matrices with that low amount of genes to test ?

 ## model3, compare the hearning level loss level accounting for the day as a covariate
 # design
 groups1 <- meta_data$condition
 groups2 <- meta_data$day
 design_model3 <- model.matrix(~ 0 + groups1 + groups2)
 colnames(design_model3) <- gsub("groups[1,2]{1}| days", replacement = "", x = colnames(design_model3))
 # model fit
 fit_model3 <- lmFit(expression_matrix, design = design_model3)
 # prepare contrasts
 cnt_model3 <- makeContrasts("Loss vs No loss" = loss - no_loss, levels = design_model3)
 # fit the contrasts
 fit_model3 <- contrasts.fit(fit = fit_model3, contrasts = cnt_model3)
 fit_model3 <- eBayes(fit = fit_model3)
 tt_model3 <- topTable(fit = fit_model3, number = Inf)

Just a little comment, the day variable is a two-level factor it is no numerical, just in case anybody has any question about it. Thanks!

DE-Analysis limma • 299 views
ADD COMMENT
3
Entering edit mode
1 day ago
Gordon Smyth ★ 7.8k

limma doesn't have a lower limit on the number of genes, see:

https://support.bioconductor.org/p/56001/

although the benefits over ordinary t-tests get smaller as the number of genes decreases.

We use limma for qPCR experiments with just half a dozen genes. You could run limma even on just one or two genes, but then it would return the same results as ordinary t-tests. The lower limit for outperforming t-tests is about 4 genes.

ADD COMMENT
0
Entering edit mode

Thanks ! Just what I ended up figuring out.

ADD REPLY

Login before adding your answer.

Traffic: 2402 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