Bowtie, BWA, strand information

[Zimbobo]

Hello there,

I use Bowtie to map RNA-Seq reads to an algae genome. When I look at the read coverage for + and - strand separately I notice that there is expression on both strands in about the same places. That would be pretty efficient use of the genome. So I was wondering whether anybody has any experience with this and can tell me how reliable the +/- strand information from Bowtie is.

I cross checked my results with BWA, the results are very similar.

There is still the possibility that I mis-interpret the strand information in the sam output files.

Thanks in advance.

[Adrian_H]

re: Bowtie, BWA, strand information

There shouldn't be any problems with the strand information. Have you tried aligning the reads in question with another tool (like BLAT) that lets you see all possible alignments? Might help you see better what is going on.

[Simon Anders]

Hi Zimbobo

Just to make sure: You do know that Illumina's standard RNA-Seq protocol is not strand specific? (The strand that the aligner reports only informs you which end of the fragment you have sequenced, not, which strand it originally came from.)

Assuming that you used one of the newer strand-specific RNA-Seq protocols: What is the ratio of the expression strength on the sense vs the antisense strand? There have been several recent publications indicating for various species that many loci are transcribed in both directions, with the antisense one (in case of ORFs) maybe having regulatory function. (See e.g. Xu et al., "Bidirectional promoters generate pervasive transcription in yeast", Nature, 2009, PMID 19169243). Maybe you observe something similar with your algae.

Simon

[damiankao]

On a related question. I have ABI SOLiD data where the strandness is kept. If I do map with bowtie, I would have to use the parameter --nrc to make it not map to the reverse complement?

The sequence of the reads should be the sense strand, so I would want to have it mapped to the forward strand, and not reverse complement?

[Simon Anders]

On a related question. I have ABI SOLiD data where the strandness is kept. If I do map with bowtie, I would have to use the parameter --nrc to make it not map to the reverse complement?

The sequence of the reads should be the sense strand, so I would want to have it mapped to the forward strand, and not reverse complement?

No. If the read is from a feature on the '+' strand, you want it mapped to the '+' strand, and if it is from the '-' strand, you want it mapped to the '-' strand. Hence, you do want Bowtie to look at the reverse-complement ('-' strand), too, i.e. you should not use the '--nrc' option.

Simon

[damiankao]

I see. So Bowtie will keep strandness and map the read as it is? It won't try to map its reverse complement?

**edit
I just realised the error in my logic. I guess if it tries to also map the reverse complement of the read, the sam output would have duplicates of every mapping, one on the + stand, and one on the - strand.

[Simon Anders]

Of course. The read stays as it is. Th e'--nrc' option is about whether the reference should also be looked at in reverse.