mapping strand-specific RNA-seq reads using TopHat

[caswater]

I am using ScriptSeq Complete Gold Kit to generate pair-end RNAseq data, and wished to map the reads in a strand-specific manner. I referred TopHat FAQ, which suggested to try 'library-type' - 'fr-firststrand' and 'fr-secondstrand' separately, and the one resulting in many more splicing junctions in junctions.bed should be the right library type.

I checked the forum, and people said that the kit should go with 'fr-secondstrand', but I still wished to double-check it. So I ran 'unstranded','firststrand' and 'secondstrand' separately on my read pairs, but found the results are quite unexpectedly similar. Specifically the junctions.bed and acceptedhits.bam file are pretty much of the same size for all of them. As I am using pair end, presumably if I force strand-specific mapping, at least one option above should give really poor alignment. But the alignment was quite similar, as you can see from the attached figure (from 1st to 3rd panels are first-strand, second-strand and non-stranded, respectively).

Can you help see what's wrong here? Thanks!

tophat --library-type fr-secondstrand -o $3 -r $4 -p 4 -g 1 -G $GTF $Ref $read_A $read_B

[bharati]

Hi,
Unstranded library of RNASeq can have strand bias, If yes then how can it be ruled out?
do we really look into it in case of fusion identification using tophat fusion?
Regards
Bharati

[fanli]

@caswater:
Tophat by default aligns to the genome and retains discordant mappings as well. So that would explain why your three alignments are fairly similar. A better way to check the strand specificity and library-type is using RSeqC.

see:

http://seqanswers.com/forums/showpost.php?p=109498&postcount=50

http://seqanswers.com/forums/showthread.php?t=50459