Using Transcriptome data to improve genome assembly

[sphil]

Hey guys,

I need to assemble a genome de novo. Unfortunately I got Illumina SE50bp data (I know, worst case...). Coverage is approx. 20 - 25 so not as bad as it could be. However further on i got some transcritome data from the same organism which is also Illumina PE95bp reads. Now I am thinking of using this transcriptome data to improve the de novo genome assembly.

As I am thinking of it a first shot workflow would look like:

Assemble SE50bp DNA reads with SoapDenovo into contigs.
Assemble PE95bp cDNA reads with SoapDenovo-trans (to stay in a closed form, say) into contigs.

Try to improve contigs from DNA-seq with RNA-seq conitigs via CAP3, for instance.

But the problem i got is, does it make sense at all to try. Sure more data isn't bad at all but the RNA-seq is not mentioned to be support data at all.

What's your opinion?



Thanks in advance for your post.


Best,


Phil

[colindaven]

Sounds like a reasonable idea - I would try it.

Why not use a scaffolder like SSPACE to attempt to bridge DNA contigs with the transcriptome derived PE reads ?

Good luck.

[sphil]

Didn't have that in mind... I will definitely give it a try!
Thank you!

[natstreet]

Your transcript contigs will contain many cases of spliced exons, including alternative splicing. How will you handle such cases if you throw both transcripts and genomic contigs into CAP3 together?

[sphil]

CAP3 isnt the right choice i guess. SSPACE seems to be more promising since it tries to elongate the given DNA contigs with RNA reads, contigs i.e..
However, using CAP3 with the given quality trimming, strand specific and similarity score adjustments should be able to improve the contigs at least somewhat.
Through the minimal overlap score one should be able to define splice sites in a way to filter those regions. Adding the similarity score of the contig itself should than be enough to verifiy wether the 'rna-contig' fits the 'dna-contig' or not. The only problem I see is the way how to recalculate alternative splice sites. BUT since it is a 'very simple' organism this event should not take to much effort into account!
Fortunatelly it is not human...

I will compare the two approaches if anyone is interested...

[HESmith]

RNA-Seq has been used successfully to scaffold a nematode genome (see http://www.ncbi.nlm.nih.gov/pubmed/20980554). You might want to consider adapting their pipeline for your data.

[sphil]

Thanks a lot for the hint!