Hi Everyone
If I have data from four miRNA libraries. One of the libraries have 4 times more reads than other three libraries. How will the data be normalized among different libraries? People use EDGER and DESeq for normalization. Although, I do understand the basic concept of total count methods. But I do not understand how these softwares do the normalization (I mean the scaling factors) and how effective is that normalization on cases like I described above.
Any explanation would be highly appreciated.
Thanks
Hi Albino
I don't understand the concept of normalization factors used by EdgeR or any other normalization software.
Using RPKM (Mapped Reads/KB/Million Reads) or FPKM (Fragments/KB/Million Reads) are two common ways of normalizing sequencing data.
http://jefworks.com/rpkm-and-fpkm-explained/
It seems to me that both John and Jim have good suggestions. You may try both and then decide.
Hi Swati,
I've been working under the assumption that there still isn't a consensus on the best normalization method. The method you choose probably depends on what you're looking for and what you already know about your samples. This paper is maybe the best summary of all of the popular options:
http://bib.oxfordjournals.org/content/early/2012/09/15/bib.bbs046.long
The 4:1 size factor difference may be a worry and it wouldn't surprise me if weird things happen because of it. Did you do physically more sequencing on this sample or did you just end up with more mapped reads for some reason?
Hi,
I was in a similar situation with my data (3 times as many reads for some libraries). I used DESeq and it handled it just fine. The tool uses variance stabilizing transformation if you need normalized data. For differential expression you don't need that. Just use your data as is and see what happens. I think it is general consensus now that RPKM normalization is flawed and should not be used.
On my data there was no difference between using edgeR/DESeq/miRNA library size normalization.
You can try all of them and plot them against each other. I guess you will see a straight line.
Just be careful on type of your experiment. If it's for example Dicer knockout vs. normal then in the Dicer knockout normalization to a miRNA library size overestimates the expression (since the overall miRNA expression is lower for all miRNAs).
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Biomolecules
- Nucleic Acids
- RNA
- Small RNA
- microRNA