Trinity - de novo transcriptome assembler

[samanta]

A group from Broad institute published a new de novo transcriptome assembly program in Nature Biotech. The paper came out in electronic version in April, and a collaborator of mine pointed it out to me. Until then, I was getting too frustrated with Velvet+Oases (especially Oases), because Oases was too demanding for RAM requirement. We have been using Trinity for transcriptome assembly since then, and I am so far happy with the results. Trinity seems to do the job with much less RAM.

I will post a comparison in my blog (http://www.homolog.us/blogs/) over the next few days and let you know here.

[westerman]

I am finding Trinity to be dreadfully slow (days to run) on anything significant (say, a single Illumina hiScan lane). I suspect that this is because of the 'inchworm' part of the pipeline is single-CPU and perhaps the 'chrysalis' part is as well. I will be interested in reading what you discover.

Nice blog, BTW.

[samanta]

When I was a student at Purdue, my computer science teacher used to tell us that a slow program is better than one that does not run For some of my data sets, Velvet+Oases failed due to RAM limit, but Trinity at least produced some results, although after 20 hours.

>Nice blog, BTW.

Thank you. You may enjoy today's post on de Bruijn graphs, where I explained why transcriptome assemblers are different from genome assemblers.