search

deweylab / RSEM

RSEM: accurate quantification of gene and isoform expression from RNA-Seq data
http://deweylab.biostat.wisc.edu/rsem/
GNU General Public License v3.0
408 stars 118 forks source link

Strand Specific Mapping Performs Poorly #34

Closed DarioS closed 7 years ago

DarioS commented 7 years ago

I use the --paired-end and --strand-specific options to map to the GENCODE Genes 25 transcriptome with Bowtie.

For --forward-prob 0 I get 2.79 % of reads mapped, for --forward-prob 1 I again get 2.79 % and when all of the strand specific options are left out of the command, I get 79.99 % mapped reads.

I have also mapped with STAR to the genome, and the mapping percentage is high and the reads are clearly strand-specific when viewed in IGV and coloured by strand of the first read in the pair. Why is the mapping percentage so low with the stand-specific approach to RSEM and Bowtie ? Also, using --forward-prob 0 and --forward-prob 1 should not produce identical mapping percentages. Are they being ignored in the code ?

I am using version 1.2.31.

cndewey commented 7 years ago

Hi,

When you specify the --forward-prob option to RSEM are you also using the --strand-specific option? If so, try omitting the --strand-specific option, which I believe forces --forward-prob 1.

In summary, try just --forward-prob 0 without the --strand-specific option.

Colin

On Sep 23, 2016, at 2:00 AM, DarioS notifications@github.com<mailto:notifications@github.com> wrote:

I use the --paired-end and --strand-specific options to map to the GENCODE Genes 25 transcriptome with Bowtie.

For --forward-prob 0 I get 2.79 % of reads mapped, for --forward-prob 1 I again get 2.79 % and when all of the strand specific options are left out of the command, I get 79.99 % mapped reads.

I have also mapped with STAR to the genome, and the mapping percentage is high and the reads are clearly strand-specific when viewed in IGV and coloured by strand of the first read in the pair. Why is the mapping percentage so low with the stand-specific approach to RSEM and Bowtie ? Also, using --forward-prob 0 and --forward-prob 1 should not produce identical mapping percentages. Are they being ignored in the code ?

I am using version 1.2.31.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/deweylab/RSEM/issues/34, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AAkM0dvskm5XUWeJrC8Ccz6-GBAvKPTwks5qs3kIgaJpZM4KEqxg.

DarioS commented 7 years ago

Thanks; the workaround is sufficient. But, it's a bit unexpected from an end user's perspective that --strand-specific dominates the setting of --forward-prob. Perhaps this could be changed, or at least explicitly documented. If --forward-prob was deprecated and --strand-specific was replaced by --strandedness and had none, forward, and reverse as the three acceptable values, it would be unambiguous and concise. It's also currently unlcear what happens if the user provide a value like 0.30 for --forward-prob, which is meaningless.

bli25wisc commented 7 years ago

Hi @DarioS, thanks for your suggestions! I have added --strandedness 'none/forward/reverse' option and deprecated --strand-specific and --forward-prob options. This will be reflected in the next release of RSEM.

Thanks, Bo