CLC genomic workbench and constructing a transciptome

[michael-steffen]

I am trying to construct an semi annotated transciptome out of a organism which doesn't have a sequenced genome.

What I am doing is using CLC genomics workbench and assembling my samples using the default parameters. This is giving me a ton of contigs back and of these contigs I am only choosing contigs with over 500 hits. I then make consensus sequences out of these contigs and blast them. Now ideally, this would be all I had to do, but what I am finding, is that many of these contigs are not unique to one another and are producing redundant blasts. I can re-assemble the consensus sequences and find that many assemble onto them selves.

Is there a optimal way to make the original assembly more efficient so that doing serial assemblies isn't necessary? If anything I am saying doesn't make sense, please say so.

Thanks,
Mike

[Brian Bushnell]

If you have truly redundant contigs, you can remove them with Dedupe like this:

dedupe.sh in=contigs.fa out=nodupes.fa

However, transcriptome assemblies of organisms with alternative splicing are supposed to have redundancies since there are multiple isoforms. Dedupe can be given various limits, though, like "only remove contained contigs that are at least 90% of the length of the containing contig" to avoid removing true short transcripts. Depending on how good CLC's assembler is, and how many isoforms you expect per gene in that organism versus how many you have, it may or may not be a good idea to deduplicate - but it's certainly better than doing serial assemblies pretending the contigs are reads.