RPKM and RNAseq data normalization

[drghosh1139]

Hi Everyone,
I am working on a set of bacterial RNAseq data (generated using Illumina Hiseq 2000 sequencer) for differential gene expression between control and treated samples. I am using CLCBio genomic workbench for the expression analysis. I have mapped the reads using RPKM method. If I am not wrong, I understand that RPKM is a type of normalisation within sample group. But if I need to compare between samples (control vs treated), how should I should normalize to get the differential expression.
I also see one option, log-transformation, before normalization. Do I need to do log-transformation with the RNAseq expression data generated through RPKM method.
Additionally, for RNAseq data which method is better for normalization; scaling, quantile or by total?

Please suggest and share your thought.

[Simon Anders]

You are looking at the issue the wrong way round. You should first figure out how (with which method) you want to detect differential expression, and then, you use whatever normalization you need for this method.

[drghosh1139]

You are looking at the issue the wrong way round. You should first figure out how (with which method) you want to detect differential expression, and then, you use whatever normalization you need for this method.

Hi Simon,
Thank you very much for your response. I am relatively new to the RNAseq analysis. Could you please elaborate on these methods with emphasis on RPKM? I want to detect differential expression by RPKM

[steinmann]

RPKM is not a method but a metric and it wont allow you to detect differentially expressed genes. Maybe you should try and figure out what you really want to do.