I was a reviewer on Boisvert et al., Ray Meta: scalable de novo metagenome assembly and profiling, and (as with DSK: k-mer counting with very low memory usage) I thought I'd share my review.

(Sorry, it's really short. My first round review had some comments that they handily addressed in the second submission, so I had nothing to add to what they said in the paper. :)

I do have two quick comments at the bottom that I didn't send them at the time of the review.

---

The authors discuss the use of Ray Meta to assemble metagenomes de novo, and, separately, to assign them to taxnonomies using a colored de Bruijn graph approach. The importance is several-fold: I believe this is the most scalable metagenome assembler yet published; the asssembly results are impressive; the scalability is impressive; and this is the first time the colored de Bruijn graph approach has been applied to metagenomes.

Ovearll, this is an impressive accomplishment and I am very pleasantly surprised that they could get such good results with a message-passing implementation. The references are very good, and the discussion is expert and relevant.

The authors are to be commended for releasing their software as open source through SF and github. Excellent!

I have no reservations about publication and recommend immediate acceptance.

Signed, C. Titus Brown ctb@msu.edu

---

OK, now for the two post-review comments --

First, in a comment thread, I urged them to post to arXiv, and Sebastien said:

But as to pushing manuscripts onto preprint servers, I prefer to let the standard peer review process follow its course. And I think it is best to have all the karma focused on a single URI. So our paper should be out in the wild somewhere soon.

This indicates, in my view, a fundamental misunderstanding of the purpose of science. I will be contacting him shortly re pistols at dawn. :)

Second, we are working on our own system for assembling large, nasty metagenomes -- you can read about it here. I would love to know if Ray Meta works well on our own data, but I haven't had time to think about how best to try it out; I don't have the manpower or time to do it myself, but I really want to play with Ray Meta. Maybe I'll try it out for the summer course... hmm. Anyway, I hope to write a much more thorough post about Ray Meta based on running it on our own nasty soil data. But that will have to wait.

--titus