Normalizing by Library Amount

[TheSeqGeek]

I got my Seq data back and processed with bow tie and htseq and I have my mapped gene counts... in my library I see that this is how much DNA was used for sequencing by the staff

615 ng ---- Sample 1
423 ng ---- Sample 2
426 ng ----- Sample 3


Can I divide the total mapped reads by these numbers to get a kind of weighted normalization?

What would be the danger of doing that?

[crazyhottommy]

Normalized to the total mapped reads would be more reasonable.

[NicoBxl]

use DESeq or edgeR ( in R ) to normalize and check for DE genes

[TheSeqGeek]

use DESeq or edgeR ( in R ) to normalize and check for DE genes

I use DESeq2 and I print out the normalized values then do a ttest of my own but my pvalues are nothing like in DESeq2.

I use this code to print out normalized values.

normalizedCounts <- t( t(counts(dds)) / sizeFactors(dds) )

[NicoBxl]

why do you use a ttest and not DESeq test (based on negative binomial) ? RNA-Seq data do not follow a normal distribution.

for normalized count, you can use also : normalizedCounts <- counts(dds,normalized=TRUE)

[TheSeqGeek]

why do you use a ttest and not DESeq test (based on negative binomial) ? RNA-Seq data do not follow a normal distribution.

Well there we go... I guess DESeq2 doesn't perform ttest...

Can you reference some quality sources to get familiar with negative binomial theory

[NicoBxl]

read DESeq paper : http://genomebiology.com/2010/11/10/R106 and also DESeq vignette that is pretty well done.
or maybe edgeR paper also