RNA-seq with UniGene sequences as target

[hrseq]

Hi all

What is the best approach when I want to get expression estimates from RNA-seq data for a species where the full genome does not exist but UniGene has a lot of sequences for it? In my case it is wheat.

Is it more appropriate to map the reads against the unique sequences in UniGene or to map them against all sequences?

Or is there a completely different approach?

regards
hubert

[keithforest]

RSEM and eXpress are two tools that can estimate expression for unigenes without a genome reference.

[hrseq]

Thanks, I know those, but with what should I feed them?

The unique or all sequences from unigene??

[Wallysb01]

Thanks, I know those, but with what should I feed them?

The unique or all sequences from unigene??

I would say it depends on how many sequences you’re talking about, as things will get kinda funny if there are too many sequences that are near identical. Then you’ll start seeing the same gene in your expression lists over and over, which might just be a pain but no real trouble.

Anyway, what kind of sequencing did you do? Might it just make sense to use Trinity for assembly and their own packages using RSEM/DESeq for expression tests?

If you have 1x50bp reads and lowish depth, that probably isn’t a good idea. But if depth is good and its 2x100bp reads (especially strand specific and ribodepleted/polyA enriched) the assembly should be good enough and you may avoid issues where your genes of interest aren’t in unigene but are in your RNA-seq.

If you want to reduce the list of unigenes, but maybe not rid yourself of too many important isoform differences, you could use CD-HIT for that....