Bashing on monstrous sequencing collections

So, there's this fairly large collection of about 700 RNAseq samples, from 300 species in 40 or so phyla. It's called the Marine Microbial Eukaryotic Transcriptome Sequencing Project (MMETSP), and was funded by the Moore Foundation as a truly field-wide collaboration to improve our reference collection for genes (and more). Back When, it was sequenced and assembled by the National Center for Genome Resources, and published in PLOS Biology (Keeling et al., 2014).

Partly because we think assembly has improved in the last few years, partly as an educational exercise, partly as an infrastructure exercise, partly as a demo, and partly just because we can, Lisa Cohen in my lab is starting to reassemble all of the data - starting with about 10%. She has some of the basic evaluations (mostly via transrate) posted, and before we pull the trigger on the rest of the assemblies, we're pausing to reflect and to think about what metrics to use, and what kinds of resources we plan to produce. (We are not lacking in ideas, but we might be lacking in good ideas, if you know what I mean.)

In particular, this exercise raises some interesting questions that we hope to dig into:

what does a good transcriptome look like, and how could having 700 assemblies help us figure that out? (hint: distributions)

what is a good canonical set of analyses for characterizing transcriptome assemblies?

what products should we be making available for each assembly?

what kind of data formatting makes it easiest for other bioinformaticians to build off of the compute we're doing?

how should we distribute the workflow components? (Lisa really likes shell scripts but I've been lobbying for something more structured. 'make' doesn't really fit the bill here, though.)

how do we "alert" the community if and when we come up with better assemblies? How do we merge assemblies between programs and efforts, and properly credit everyone involved?

Anyway, feedback welcome, here or on Lisa's post! We are happy to share methods, data, analyses, results, etc. etc.

--titus

p.s. Yes, that's right. I ask new grad students to start by

assemblying 700 transcriptomes. So? :)

A workshop report from the May 2015 non-model RNAseq workshop at UC Davis

We just finished teaching the second of my RNAseq workshops at UC Davis -- the fifth workshop I've hosted since I took a faculty position here in VetMed. In order, we've done a Train the Trainers, a Data Carpentry, a reference-guided RNAseq assembly workshop, a mothur (microbial ecology) workshop, and a …

There are comments.

Page 1 / 1

Living in an Ivory Basement Stochastic thoughts on science, testing, and programming.

Other articles

A workshop report from the May 2015 non-model RNAseq workshop at UC Davis