Estimation of expression levels

[kenosaki]

Hi,

I use RPKM for the estimation of expression levels from RNA-Seq data usually.
Does anyone know the alternative method?
In what cases, the alternative way would be better than RPKM?

Thanks

[Simon Anders]

That depends very much on what you want to do with your estimated expression levels.

[kenosaki]

I'd like to look at,
1) Global expression changes of genes between two groups of sample from different tissues. (like microarray experiment)
2) The exons that specifically spliced in the differentially expressed genes of interest identified from 1)
3) Possibly, putative exons of genes in either tissues.

[Simon Anders]

1) Global expression changes of genes between two groups of sample from different tissues. (like microarray experiment)

The crucial part is to use an appropriate method to assess the statistical significance of any differential expression you may see. See our tool, DESeq, or Robinson and Smyth's edgeR for statistical methods.

Both tools want unnormalized, raw counts. they don;t work if you normalize to RPKM, and methods that use RPKM as input have less statistical power.

2) The exons that specifically spliced in the differentially expressed genes of interest identified from 1)

Several people (including us) are working on a proper statistical test for this, but nothing is out yet. Cufflinks (Trapnell et al.) gets you at least half the way.

3) Possibly, putative exons of genes in either tissues.

For discovery of new features, normalization is quite unimportant. Either it is there or it isn't. There is no background, after all.

Simon

[kenosaki]

Thank you Simon,

I found the very informative thread about RPKM.

http://seqanswers.com/forums/showthread.php?t=3254

Ken