Reduce complexity of a de novo assembly prior to annotation

[FGponce]

Hi everyone.

We have a trinity assembly that presumably contains lots of isoforms/hybrids/paralogs etc.

We have a second organism in there at a low level so don't want to filter out isoforms with low read evidence as others may.

I'm aware trinity creates a component that it splits to isoforms but this component file doesn't appear to be generated as a fasta at any time.

How can we filter out the best transcript models before we do some in depth annotation? Can we run cuffmerge on a trinity assembly eg put two copies of the assembly in as two transcriptomes and get it to combine with itself?

Any other ideas?

Cheers,

FGPonce

[Apexy]

Have you tried using any of the de novo clustering algorithms to curb redundancy?

[FGponce]

No havn't heard of those. Are they incorporated as a funtion of a particular tool or suite of tools?

[Apexy]

Hi,
My understanding is that you have assembled a set of transcripts and you would like to annotate them. As it is with transcriptome assemblers, there is usually a lot of redundancy. It has become popular that these transcripts can be group together before using the popular BLASTX for annotation. However, biggest challenge to identifying the threshold at which further clustering leads to loss of information because sequence similarity does not invariably mean functional similarity. Programs such as WCD : http://www.sanbi.ac.za/resources/sof...downloads/wcd/, Cdhit: http://cd-hit.org/, Vmatch: http://www.vmatch.de/ and CAP3 http://seq.cs.iastate.edu/ have been used of cluster assembled transcripts, EST and 454 type sequences.

HTH