GC content and length bias in RNA-seq data

Question

[parvin.shariaty]

Hi I have seven RNA-seq data in two groups( control,treat) that produced from two companies (controls=BGI,treats=novogen). I used NOISeq to quality control of count data, and the results recommended normalization for GC content and length. But I have studied some papers and have founded that GC content and length normalization in gene expression comparison between two groups introduced bias. So I would like to use normalization factors instead of size factors in DESeq2(as described in differential analysis of count data - the DESeq2 package).

normFactors<-matrix(runif(nrow(dds)*ncol(dds),0.5,1.5),ncol=ncol(dds),nrow=nrow(dds),dimnames=list(1:nrow(dds),1:ncol(dds)))
normFactors<- normFactors / exp(rowMeans(log(normFactors)))
normalizationFactors(dds)<- normFactors
sizeFactors(dds)<-normFactors

Is it a correct manner to normalization and introduce the normalization factors in DESeq2? Thank you for cooperation in advance. Best regards Parvin

[WouterDeCoster]

It is not correct to use groups for which the data was generated differently, by a different company in your case. Your groups are confounded by technical differences in addition to the biological effects.

[Kevin Blighe]

In addition to what Wouter has said, it looks like you are making your situation overly complex. You have 2 different datasets from 2 different platforms. You can analyse these together and include batch/platform as a covariate, but you must exercise caution. I have done this in the past but we eventually decided not to proceed for fear of criticism.