Denovo transcriptome assembly

[tellsparck]

Hi Forum,
We have some RNA-Seq data generated using Nugen's Ovation RNA-Seq System V2 kit.I plan to do denovo transcript assembly using cufflinks with this data. But when I look at the BAM files, I see many genes which have quite a lot of reads in introns. My feeling is that most of it comes from pre-mRNA (the kit uses both random primers and oligo-dT primers) although a fraction of it may be biologically meaningful (retained introns/alternate exons etc). Does anyone have any experience with denovo assembly using libraries made with this or similar kits? Thanks!

[tellsparck]

Hi, just to give more details about my concerns, Since cufflinks depends on junction reads for denovo assembly, having lots of premRNA will affect these types of reads and I have no idea how this will affect the results...
Thanks!

[Mocca]

Try using another assembler that handles isoform resolving differently. I work with Trinity and find that good (2n vertebrate species) however it struggles a bit with paralogs and I also find retained introns (might be artefacts, might not). Any genome resources available to you (same or closely related species)? If yes, you could try a mixed approach with de novo first and then map the transcripts to the genome reference. Best of luck :-)

[tellsparck]

Thanks for your input Mocca.
My samples are from mouse (my bad not mentioning it the first time around) and originally, the idea was to use cufflinks because it is a genome guided assembler. I saw that Trinity has a genome guided version now- have you used it and do you have any advice on that?
Thanks!

[Mocca]

Having a model species gives you more alternatives. Trinity genome guided worked OK for me, but its difficult to compare as I work on a non-model species. Trinity would cluster your RNAseq reads according to chromosome (and any scaffolds not connected to a chromosome) and then assemble each cluster de novo. You could then for example take the trinity transcripts and map them to the genome and see how they behave. There are also some good papers on mixed approaches: de novo first and then mapping towards reference, concatenation of several transcriptome assemblies based on various parameteres etc. depending on what you are interested in.

Hope you find an approach that fits your project :-)