Differential expression in edgeR with cuffdiff count output

Question

[mjg]

Hello,

I have RNA-seq data from paired samples and want to do isoform-level differential expression.

My understanding is that is not possible to explicitly specify paired design in cuffdiff, so I considered the option of using the isoforms.read_group_tracking file and extract the raw_frags column, which I think corresponds to counts. Then use this as input in edgeR where I can specify paired-sample design.

Does this sound sensible?

If not, my question is: which package can I use to do isoform-level DE which takes into consideration paired-sample designs.

Thanks,
Maria

Answer 1

[mark.ziemann]

DEX-Seq Estimates exon usage coefficients from the fitted terms of the GLM and may be just what you're looking for.

http://www.bioconductor.org/packages/release/bioc/manuals/DEXSeq/man/DEXSeq.pdf

[mjg]

Thank you for your suggestion! I did not know about DEX-Seq, I am running it now and will compare it with my previous results.

Maria

Answer 2

[Lior Pachter]

You can use sleuth.

[super]

I have played sleuth. But if I am interested with DE genes but not DE transcripts in downstream analysis, how could I do it with sleuth?

e.g. I got the DE transcripts table with Q<0.05 with associated gene name. But as you know, one gene could have multiple transcripts.

like

target_id          Q                        ext_gene
ENST00000257570    0                        OASL
ENST00000339275    1.08823325997074e-138    OASL
ENST00000543677    4.50865532164411e-20     OASL

Could I use it for finding DE genes like edgeR/DESeq? how?

Thank you!