STAR split reads into two strands

Question

[Alexander]

Hi all,

I'm trying to find a way to split reads mapped with STAR into two separate bam-files, one for each strand. STAR is able to do that when creating bedgraph or bigwig outputs (--outWigType and --outWigStrand), however I can not find any option to split raw reads into two separate files correponding to the strand they were transcribed from. I do not want to get counts per gene, I know how to do that, instead I need to get actual reads for each strand. Any ideas?

Thanks a lot in advance

[swbarnes2]

Based on strand of genome, or transcript?

[Alexander]

Based on strand of genome

[swbarnes2]

Why can't you split on the binary flags with samtools?

[Alexander]

Because this will give me 50/50 split. 50% of all reads will map to plus strand and another 50% will map to minus strand, but what I care about is from which strand of the genome the transcript came from. May be I was not clear?

[swbarnes2]

You can filter for both alignment direction and which read of the pair you are looking at.

[Alexander]

Hmmm.. do you mean flags for first in pair and second in pair? Yea, I thought about scripting this, but decided to check if there is a ready solution. Is it such a rare thing to do that no available tool offer such a function?

[swbarnes2]

I'm not sure why you need a script. samtools view -f 80 will, for instance, get all the R1 reads that map to your genome in reverse. Isn't that one of the categories you want?

[Alexander]

Well... technically yes, but I probably want to check that R1 is indeed downstream of R2 by the coordinates with this filtering, just to be on the safe side. May be I'm overcomplicating things :)

[swbarnes2]

Okay, won't f -82 do that for you, by demanding proper pairing? Is this a non-model organism, so you can't just look up the orientation of each gene, instead of deducing it from reads?

[Alexander]

Sounds good! Thanks a lot for the ideas! Good discussion always helps :) I will do first filtering for -f 2 and then do further filtering steps on the outcome file and put it together.

Answer 1

[Divon]

You can use my Genozip program:

genozip myfile.bam
genocat myfile.bam.genozip --FLAG -0x10 --output myfile.forward-strand-reads.bam
genocat myfile.bam.genozip --FLAG +0x10 --output myfile.reverse-strand-reads.bam

See here: https://genozip.com

Paper: https://www.researchgate.net/publication/349347156_Genozip_-_A_Universal_Extensible_Genomic_Data_Compressor

[Alexander]

Thanks. Hm... but how different is it from usng samtools tags? Would not it give me 50/50 split for all reads in the bam file?

[Divon]

It will give you the strand of the reference genome that your read maps to. For example, if your entire data is from a single RNA molecule, then it will all map to the same strand. Whether that is actually the strand your RNA came from in the DNA of the individual you sampled from, depends on whether their DNA matches the reference genome in that particular region.

[Alexander]

It will give you the strand of the reference genome that your read maps to

Yes, but in paired end sequencing one read will always map to the opposite strand, regardless of the transcript origin, so I will get 50/50 split. What I care about is that both mates for read pairs that came from transcript produced in sense orientation would be labeled as originating from plus strand, and both mates from read pairs that came from anti-sense transcript would be labeled as originating from minus strand.