Question: Should I skip the eBayes step when using Limma for Affymetrix miRNA v1 chip?
gravatar for Guest User
5.2 years ago by
Guest User12k
Guest User12k wrote:
Dear List, I am working with Affymetrix's miRNA V1 chip, which uses very different probe sets for different molecule types, e.g. 4 identical probes for one miR, or 11 different probes for a snoRNA. I have read that the eBayes step assumes equal error variance between probe sets so it is not suitable for this kind of mixed set of probe set designs. To further complicate matters I am thinking about generating a custom CDF where the miR probe sets would have varied number of probes. Should I look at everything through Limma without the eBayes step (making it equivelant to a normal t-test?), or separate into several different analyses for different molecule types and only drop the eBayes step for the miRs (which will have varying sizes of probe sets)? Many thanks, Scott -- output of sessionInfo(): > sessionInfo() R version 3.0.2 (2013-09-25) Platform: x86_64-w64-mingw32/x64 (64-bit) locale: [1] LC_COLLATE=English_United Kingdom.1252 [2] LC_CTYPE=English_United Kingdom.1252 [3] LC_MONETARY=English_United Kingdom.1252 [4] LC_NUMERIC=C [5] LC_TIME=English_United Kingdom.1252 attached base packages: [1] stats graphics grDevices utils datasets methods base -- Sent via the guest posting facility at
mirna cdf probe limma • 1.0k views
ADD COMMENTlink modified 5.1 years ago by Gordon Smyth39k • written 5.2 years ago by Guest User12k
Answer: Should I skip the eBayes step when using Limma for Affymetrix miRNA v1 chip?
gravatar for Gordon Smyth
5.1 years ago by
Gordon Smyth39k
Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia
Gordon Smyth39k wrote:

Dear Scott,

You are misinterpreting the empirical Bayes assumption of eBayes. eBayes doesn't assume equal variances, it only assumes that the variances can be considered as a random sample from the same distribution.

Omitting the eBayes step would would throw the baby out with the bath water.

I doubt that the error variance depends quite as directly on the number of probes in a probe-set as you might think.  When we have analysed the miRNA Affymetrix chip, we have found that it has major problems from the point of view of normalization, while the issue that you raise is relatively minor.

I could suggest ways to take into account the number of probes per probe-set in the eBayes calculations, but I don't think this will be important.

Best wishes

PS. If you have the choice, RNA-seq is cheaper and better.

[This answer was originally posted to the Bioconductor mailing list on 22 August 2014.]

ADD COMMENTlink written 5.1 years ago by Gordon Smyth39k

Just for the record, here is how you can incorporate probe-set size into the eBayes step to see whether it is important.  I will simulate a little toy example where we know there should be a trend:

  n <- 1:100 # number of probes per probe-set
  ID <- rep(1:100,n) # probe-set IDs
  nprobes <- length(ID)
  x <- matrix(rnorm(nprobes*3),nprobes,3)
  y <- avereps(x,ID=ID)

# y has 100 rows, each row is an average of n probes

  design <- matrix(1,3,1)
  fit <- lmFit(y,design)
  fit$Amean <- log(n)
  fit <- eBayes(fit,trend=TRUE)

Limma will estimate a decreasing trend of variance vs n as well as doing empirical Bayes squeezing around the trend.  The x-axis label of the plot will say "average expression" but it is actually log(n).

Best wishes

[This answer was originally posted to the Bioconductor mailing list on 24 August 2014.]

ADD REPLYlink written 5.1 years ago by Gordon Smyth39k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 297 users visited in the last hour