Methods for 'exploratory analysis' of sequenced non-model organism transcriptomes

[ShellfishGene]

With the price of next gen sequencing being as low as it is now, often whole runs are used for single experiments to answer specific questions. After this is done, the quesiton of 'what else can we do with the data?' comes up. (Or, and unfortunately this is becoming more common, there is no specific question at all before sequencing, it's done 'just to see what comes out of it'.) As a bioinformatician you then get a load of transcriptome reads, and that question.
What do you do in that case? Especially for non-model organisms, where you don't have a genome to map to?

The most obvious steps are assembly and blastx, maybe with GO annoation using something like Blast2GO. RNA-Seq is also possible, but of limited use without a comparison. Maybe one can find highly expressed genes that don't have a blast match anywhere and are thus interesting.
What are other ideas for 'exploratory analysis' of non-model organism transcriptomes?

[colindaven]

Hi,

we have plenty of de novo assembled contigs from bacterial genome resequencing projects which don t have matches (on blastn and blastx levels).
We have treated these as dodgy low confidence contigs to date and have not used them further.

With transcriptomes I would be wary of novel contigs. Maybe oligonucleotide usage would be interesting if the contigs are more than 1000 bp to see if the contigs are possibly novel transcripts or assembly / sequencing errors.

I would definitely recommend BLAST against metagenomic DBs such as those at CAMERA.

[jordi]

Hi,
maybe that post is too old, but I think it's very interesting what to do with the thousands of reads from a next-generation platform without a reference genome. Given, for instance, the 454 read length, why not try to blast them directly?? Besides, usually, those not included in a contig (singletons) are treated as a unigenes. We were looking for a specific proteins in a 454 transcriptome from a non-model specie. According to this we tried to align the 6 frames translation of each read to an aminoacid reference sequence from the closer phylogenetically specie.
What do you think about this approach??
Regards