More Parallel CONCOCT - Githubissues

BinPro / CONCOCT

Clustering cONtigs with COverage and ComposiTion

Other

120 stars 48 forks source link

More Parallel CONCOCT #201

Closed alneberg closed 5 years ago

alneberg commented 5 years ago

Quite a large pull request where the work by @chrisquince to make CONCOCT parallel and much faster is finally included.

Small changes include:

Fix for the random seed.
A new minimised Dockerfile which is more up to date
A new gitter channel on the main Readme instead of the old mailing list.

chrisquince commented 5 years ago

Hi Johannes,

I tried this on a clean install on my mac and it appears to run but with a lot of warnings:

"RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. "

Something to do with library versions and numpy. I will try it on my server now. Then we need to update the doc to add usage of the refine version of concoct once that is done we can make this a newer version.

The version after this should probably be python3, do you agree?

Best, Chris

alneberg commented 5 years ago

Hi Chris,

yes I saw those warnings as well. I will have a look at it. I think it is something raised by a newer version of numpy as you say.

I would like to restructure a large part of the documentation and tutorial. For example, I would like us to recommend the use of Kallisto by default and otherwise plain bowtie2 instead of the bash script we currently use. Is this something you would like to discuss?

I definitely agree that it is time to support python 3.

Great to hear from you again, Johannes

chrisquince commented 5 years ago

Hi Johannes,

Yes, I am happy with that. I actually use bwa mostly. If we give perhaps a couple of alternative ways to generate the coverages. Did you test kallisto on a mock community?

OK well it would not be a big conversion to python3 as most of our code is actually C.

Nice to hear from you too!

Best, Chris