DESeq2 Transcriptome Analysis

[Quat]

Hello,

I'm trying to put a pipeline together for someone which will result in a transcriptome analysis with deseq2 in the end and I'm having some issues.

There isn't a reference genome for the organism. What we've done is assemble the transcriptome sequencing (illumina). We then aligned the reads back to the contigs using bowtie2.

We'd like to analyze the resulting bam/sam files with DESeq2, but most of the functions to read bam alignments require a DTF or other annotation.

Is DESeq2 appropriate for this sort of analysis and is there a function to create a summarizedexperiment object from bam files without an annotation?

Thanks

[Michael Love]

hi,

DESeq is not designed for analysis at the transcript level, because there are many reads which are shared between transcripts of a gene. Therefore you cannot generate raw counts of reads which *uniquely* align to the features. DESeq requires that one can identify all the exons of a gene, and can generate the count of reads which can be uniquely aligned to that gene. Simon explains this in more depth here: https://stat.ethz.ch/pipermail/bioco...ry/043410.html

Software such as BitSeq or Cuffdiff will perform analysis at the transcript level.

http://www.bioconductor.org/packages/release/bioc/html/BitSeq.html

http://cole-trapnell-lab.github.io/cufflinks/cuffdiff/

[Quat]

Thanks, I'm giving bitseq a try.

I've seen people use bowtie to report only the best match and then run DESeq, but Im guessing that isn't appropriate.

[ickou]

http://www.ncbi.nlm.nih.gov/pubmed/25063469