Question

cross platform normalization in R

0

Entering edit mode

nazaninhoseinkhan • 0

@nazaninhoseinkhan-9194

Last seen 8.8 years ago

Dear all, I am trying to perform cross platform normalization between affymetrix, illumina and agilent microarray data. I know I can do it using Combat package which is now available in SVA package in R. My problem is I have not recognized yet the structure of input files in SVA. The manual works with bladderbatch and phenodata that their structures is not clear. I will appreciate any advice Nazanin

R Combat SVA • 2.7k views

ADD COMMENT • link updated 8.8 years ago by Steve Lianoglou ★ 13k • written 8.8 years ago by nazaninhoseinkhan • 0

1

Entering edit mode

Just to further extend a bit Steve's answer---without being an expert i think it is not appropriate to combine data from different technology, except if there is a very strong reason or a specific down-stream procedure----but if you had data from i.e. same technology, same or very simila experiment design but different chip, you could check the vignette from the inSilicoMerging R package:

https://www.bioconductor.org/packages/release/bioc/html/inSilicoMerging.html

I have used it two combine data between 2 different affymetrix platforms(hgu133a & plus2), and it has various merging techniques, like Combat you have mentioned---but in the end you merge and keep only the common probesets/genes

Anyway, you can also check the following review, which adress the matter in detail, and describes many methodologies and R packages, like the GeneMeta:

http://www.mdpi.com/2076-3905/4/3/389

Best,

Efstathios

ADD REPLY • link 8.8 years ago svlachavas ▴ 830

score 0 · Answer 1 · 2016-01-23

Combat is used to attempt to remove unknown batch effects from single experiments, not to combine multiple experiments from different platforms. Combining data in the way you are after is not trivial, and people usually do this using meta-analysis. Just yesterday (I think) Jim MacDonald said he likes to use the GeneMeta package for this, so you might want to start there.