Calculate exon level expression using RSEM

[Gao, James (NIH/NCI) [E]]

Hello Bo,

I have been using RSEM to get reads count at gene/isoforms levels, is it also possible to get counts at individual exon level? I am thinking about using the exon level data to estimate alternative splicing events of genes. Please advice.

Thanks,

James

[Bo Li]

Hi James,

Currently RSEM does not provide exon level expression values. However,
our lab has another project working on estimate alternative splicing
events. Colin, maybe you can comment on this?

Best,
Bo

[Colin Dewey]

Hi James,

As Bo mentioned, RSEM does not provide counts at the exon level. However, you could easily use the genomic-coordinate-based alignment (in BAM format) produced by RSEM as input to another program that counts the number of reads falling within prespecified intervals (such as exons). I believe there are a number of such tools available (although I have not used them myself). One thing to note if you do this - you will need to filter out the alignments that have low quality/probability. You can do this by either filtering based on the SAM quality field or the custom probability field added by RSEM.

And yes, we are soon to release a new method for analyzing alternative splicing of genes from RNA-Seq data. We will post an announcement to this list once we have made it publicly available.

Best,
Colin

[Gao, James (NIH/NCI) [E]]

Hi Colin and Bo,

Thanks for the quick reply and looking forward to seeing your new method to analyze alternative splicing events. If you are looking for real data to test on, we are looking for another method to validate our findings. Because the current RSEM generates read counts on genes and isoforms, I thought you may also have a way to add an exon id in the reference files to generate the counts at individual exon level. At the meantime, I will take your suggestion to try on other methods to generate the counts.

Regards,

James

________________________________________
From: Colin Dewey [CDE...@bioostat.wisc.edu]
Sent: Wednesday, October 31, 2012 10:03 PM
To: rsem-...@googlegroups.com
Subject: Re: [rsem-users] Calculate exon level expression using RSEM

[Colin Dewey]

Hi James,

Part of the reason RSEM doesn't provide exon-level counts is that it really doesn't have any knowledge of "exons." RSEM acts directly on read alignments to transcript sequences, not to the genome where the exons are primarily defined. In addition, for de novo transcriptome assemblies, we don't have knowledge of exon boundaries.

Thanks for the offer of another data set to test on. We'll keep that in mind.

Best,
Colin

[David Haan]

Hi Guys,

Is it possible to "trick" RSEM into giving exon level counts by feeding it a bed file or gtf file of exon regions?

Thanks

David

[Colin Dewey]

Hi David,

You can certainly do this and it will give you something somewhat reasonable, but it is far from ideal for a couple of reasons:

1. Junction reads will not map

2. There is more information to allocate multi-mapping reads when considering transcript structures and abundances (i.e., exons in the same transcript should have similar counts, ignoring biases).

I would recommend the following approach instead:

1. Run RSEM with a full transcript GTF file and with the --output-genome-bam and --sampling-for-bam options.  This will generate a genome-level BAM with each read mapped to a single location.

2. Use bedtools or something similar on the resulting genome-level BAM and a BED file of exon intervals to generate counts for those intervals.

With this approach you use the full power of RSEM.

Best,

Colin

 

--
RSEM website: http://deweylab.biostat.wisc.edu/rsem/
---
You received this message because you are subscribed to the Google Groups "RSEM Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rsem-users+...@googlegroups.com.
To post to this group, send email to rsem-...@googlegroups.com.
Visit this group at http://groups.google.com/group/rsem-users.