RNA-seq - Recovering antisense RNA counts

[a.kmg]

Hello to all,

I have RNA-seq data generated with the mRNA stranded protocol of Illumina (which is a dUTP protocol).

Generally, I used TopHat/Bowtie and htseq-count to align and quantify my data. So, I have done that, with the "fr-firststrand" for TopHat and "stranded=reverse" for HTSeq.

Now, I want to separate the sense and antisense RNAs from my data. I have seen two possibilities :

With Htseq-count => We can use the --stranded=yes option, to keep all the anti-sense expression counts and the --stranded=reverse option, to keep all the sense counts.

samtools view -F 16 Reads.bam => to obtain reads that map on forward strand
samtools view -f 16 Reads.bam => to obtain reads that map on forward strand

I think there is a biological principle that I do not quite understand.

For the second option, which uses samtools, it is assumed that all reads that align to the reverse strand correspond to antisense transcripts ? and all reads that align to the forward strand correspond to sense transcripts ?

In fact, simply, I do not understand the antisense selection method.

Can anyone explain this to me? And what is the best method for you? The two methods are equivalent?

Thanks in advance to all of you.

Best,
a.kmg