Which of these steps in RNA-Seq pipeline do you do?

[PFS]

In your analysis of RNA-Seq pipeline, which of the following steps do you normally do?

(Illumina)

PRE-PROCESSING
1) removal of reads with consecutive stretches of "B" qualities
2) trimming 3' end (what criteria do you use?)
3) trimming 5' end (what criteria do you use?)
4) removal of duplicate reads (reads that start at the same exact position, both in first and second mate)

Diff Exp ANALYSIS
5) Removal of strand assignment before estimating abindances
6) tophat/cufflinks/cuffdiff
7) bowtie and DESeq/EdgeR/DEGSeq

ENRICHMENT
8) correction length bias (goseq)

9) ....

Thanks!

[kopi-o]

Just some general comments -

6)tophat/cufflinks/cuffdiff

Not sure it makes sense to bundle these as a "package" - I always use TopHat but not necessarily Cufflinks (and very rarely CuffDiff).

7) bowtie and DESeq/EdgeR/DEGSeq

Why do you have Bowtie bundled with the DE packages here?

[PFS]

Hi,

You are right. It would make more sense to separate TopHat from Cufflinks. I guess my intention was to pinpoint to some kind of alignment tools with or without transcriptome assembly tool.

As far as bowtie and the R packages, I meant to indicate an alignment tool (here bowtie or bwa) and packages that can do differential expression analysis with count data.

[pbluescript]

I trim based on average quality alone.
As long as I optimize my PCR cycles, I don't worry about removing duplicate reads. With RNA-Seq, it's actually pretty normal to get reads that start and end at the same place that are not PCR duplicates.
I use Tophat for mapping, but then after that it really depends on the question I'm asking about the data.

[kopi-o]

I usually remove duplicates if it's a paired end run (reads where both mates start at the same positions, as you wrote), then TopHat, then Cufflinks and/or HTSeq and DESeq. Pre-processing of the reads depends a bit on the platform (Illumina/SOLiD) and on whether the run "looked good" in general.