Alignment of de novo assembled transcripts to reference transcriptome?

Question

[Shraddha]

Hi all,

I started with 40 samples of raw (but trimmed) RNASeq samples. I used these as inputs for SPAdes-rna and Trinity, to assemble them without a reference. I now have 80 assembled transcriptomes (?), and I tried to follow the workflow of

.fq file > .bam file (+ alignment stats) > featurecounts > deseq2

This worked until the alignment step (I used bowtie2 and the transcriptome of a related plant, but not the target plant), but featurecounts yields all 0s and no alignments assigned.

Right now, I'm mapping the raw reads to the same transcriptome and getting much better alignment scores (here I'm using bowtie2, hisat2 and STAR). I also have a reference genome of the target plant, which I plan to map the raw reads to.

Questions:

1. Should assembled transcripts be aligned back to a transcriptome? or a genome? or at all?
2. I read a paper that said that bowtie2 shouldn't be used for RNASeq data, but their manual says it can be. Is there a concrete answer to this?
3. Should assembled transcripts just be used directly (and only) for functional annotation?
4. Is there literature that has a workflow for this? Or an explanation? I can only find fragments (which is everyone's problem, I know, but maybe someone already addressed it in this case?)

Thanks!

[Friederike]

The extensive wiki of the Trinity github repo should address most of your questions.

In short: use RSEM on the Trinity output for read quantification and then either use the Trinity wrapper for DE analysis or directly go to DESeq2 with the RSEM results

[Shraddha]

Yes I see, that's very useful, thank you!