Values in the data.frame are negative by RMA normalization, how deseq2 package work for that data?

Question

[dqz377977905]

Hi，guys： I got a dataset from NCBI GEO , but the data was processed by the method under mentioned1. Signal normalization was performed by RMA method. RMA expression measure was computed using Bioconductor Affy package in R language (Bioconductor 2.5; R version 2.11.0) Some values in the data.frame are negative, and i got confused. And we know RMA method, the data was normalization by log2 processed. Should I deal with multiplier and power the values like (gse75333 <- 8^(gse75333) gse75333 <- gse75333*100)? And then i could process the data by Deseq2 package in R. I want to know the process is or not corret for data analysis.

gse75333 <- getGEO("GSE75333", GSEMatrix =TRUE, AnnotGPL=TRUE)
if (length(gse75333) > 1) idx <- grep("GPL570", attr(gse75333, "names")) else idx <- 1
gse75333 <- gse75333[[idx]]
gse75333 <- exprs(gse75333)
gse75333 <- 8^(gse75333)
gse75333 <- gse75333*100

[ATpoint]

Please use the formatting bar with the code option (10101) to highlight code and structure your question.

[dqz377977905]

I'm sorry about that. That is my first post for help. Next, I would structure it.

[ATpoint]

No need to apologize. If you structure your question well you notably increase your change to get good answers, so it is purely in your interest ;-)

Answer 1

[Benn]

Do you know you have microarray data here, and not RNA-seq? DEseq2 is for RNA-seq data, which is generated by reads that align to the genome, and are then quantified per gene (raw read counts). But you have old school microarrays, these arrays measure intensity of fluorescent labeled RNA/DNA which hybridize to probe sequences attached to the surface of the array. Please read more about it if you are unfamiliar with them. For microarray data you can try limma instead of DEseq2.

[dqz377977905]

Thank you for your attention. The suggestion is very helpful for our group. And I learn the process method in a bioinformatic lesson. But I also want to know about that the process that I mention up, could I do it like that. Because of the RMA normalization is logarithm.

[Benn]

The RMA normalization is for microarrays (you have microarray data), so please use tools specifically designed for microarrays. Limma is the most popular tool for microarray analysis. There is no reason to try converting your microarray data into RNA-seq data, that doesn't make sense! It's like doing PCR and check for product with a western-blot...