Hello,

I'd like to combine two microarrays (same platform and chip) - two different experiments.

Currently I have results of two experiments (Affy U133 Plus 2.0) performed in two different labs. I'd like to combine them into one with batch effect correction. I'd like to perform MDS, PCA and other analysis.

What I did so far?

I took .CEL files read them with affy package and did the normalization using classic RMA. As a result I have two ExpressionSets.

Then I found inSilicoMerging package which allows you to merge experiments even from different platforms, I read the publication and actually I was looking promising.

I tried to go through all steps described in the article:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3568420/

to learn how to do this and how the data should look like regarding the preprocessing.

Unfortunately the InSilicoDb database doesn't work, so I couldn't obtain the training data from tutorial. Nevertheless I loaded my data, then normalization using affy package.

After I got my ExpressionSets, time for some merging:

library(inSilicoMerging)
esets <- list(exp1,exp2)
eset_NONE <- inSilicoMerging::merge(esets, method="NONE")

Then I got the following error:

  INSILICOMERGING: Run with no additional merging technique...
Error in .validate_assayDataElementReplace(obj, value) :
  object and replacement value have different dimensions

My requests to you:

1. Is it possible to fix it somehow, to perform merging?

My ExpressionSets:

> exp1
ExpressionSet (storageMode: lockedEnvironment)
assayData: 54675 features, 15 samples 
  element names: exprs 
protocolData
  sampleNames: sample1.CEL sample2.CEL ... sample15.CEL (15 total)
  varLabels: ScanDate
  varMetadata: labelDescription
phenoData
  sampleNames: sample1.CEL sample2.CEL ... sample15.CEL (15 total)
  varLabels: sample
  varMetadata: labelDescription
featureData: none
experimentData: use 'experimentData(object)'
Annotation: hgu133plus2

 
> exp2
ExpressionSet (storageMode: lockedEnvironment)
assayData: 54675 features, 32 samples 
  element names: exprs 
protocolData
  sampleNames: sample2_1.CEL sample2_2.CEL ... sample2_32.CEL
    (32 total)
  varLabels: ScanDate
  varMetadata: labelDescription
phenoData
  sampleNames: sample2_1.CEL sample2_2.CEL ... sample2_32).CEL
    (32 total)
  varLabels: sample
  varMetadata: labelDescription
featureData: none
experimentData: use 'experimentData(object)'
Annotation: hgu133plus2 

2. Do you know any other packages (better than this one, later published) that are designed to do such merging?

I'll be very grateful for any clues and Ideas.

Best Regards,

Adam

InSilicoMerging was deprecated in BioC3.3 and removed in Bioc3.4 (we are on BioC3.8, so that's like two years ago). Obviously it's suboptimal to use software that has been abandoned.

You may be able to use the frma or SCAN.UPC to make the data similar enough to combine and analyze as on set, or if you aren't concerned with between-dataset comparisons (which are confounded), you could either combine and fit a batch effect (again, if it's orthogonal to the comparisons you care about), or alternatively use GeneMeta.