getting RPKM and VST using DEseq2 package

Question

[zizigolu]

Hello,

I read DEseq2 vignette but I could not find any function to RPKM and VST. Then how I can have these files?

Thank you

Answer 1

[Antonio R. Franco]

Fereshteh

Don't stick to trying to calculate RPKM.

You can use other normalization methods included in DESeq2

In confused, you have choices

There are other packages, such as NOISeq, that allow you to normalize data through various methods. The nice thing about NOISeq is that it just after normalization, it will provide you with a statistical summary of each of the normalization method you try indicating if normalization is correct or not. You can then keep using NOISeq, or go to DESeq knowing what is the correct normalization

I did this with some of my data, and RPKM did not pass the filter. I had to try some other methods, and the latest I tried was OK

Give it a try. The NOISeq vignette will help you

Answer 2

[andrew.j.skelton73]

There's VST and RLog described on page 16 of the manual.

For RPKM, see Devon's answer here

Manual is available here

[zizigolu]

thank you Andrew, But i read both already but no clue yet

if i open rstudio, which i should type to start the analysis?

[andrew.j.skelton73]

You stated you've already read the vignette? - The DESeq2 manual is fantastically detailed and covers everything from preparing your counts / pheno table, and running all aspects of differential gene expression analyses. I'd suggest you go back and read the manual again, and follow it using example data provided in the package before starting on your own data.

[zizigolu]

thank you but nothing there about RPKM, also that was too puzzling

[andrew.j.skelton73]

Why do you want RPKMs? If you look at the link I put in my answer, Devon explains how to calculate them along with a very apt warning about NOT using them with DESeq2.

[zizigolu]

thank you, but even i don't know how to normalize my raw counts

[andrew.j.skelton73]

Ok, rule 1, if you don't know what to do, read the manual. The manual contains all the information about the tool, and even contains a guide with sample data to get your started if you're unsure. Frankly, I'm not going to spoon feed you the information from the manual.

[zizigolu]

yes you all right

thank you

[brendon]

Hey @andrew.j.skelton73, sorry to bump an old thread but I've got a related question. My understanding from the vignette and source code is rather that the the VST doesn't account for the feature length, but rather the inter-sample count variance by feature and the library size factors. Is this not the case? In using the normalized counts (from VST) for downstream analysis, is it not appropriate to adjust for feature lengths using the rpkm?

Thanks for your insight.

[andrew.j.skelton73]

Check out this discussion on gene length in DESeq2.