What is the best Linux distro for biocomputing?

[merilius]

Hi Guys,

So need to remove old Centos 5 on a compute node with some more recent distro... The least-effort choice for me is CentOS 7, but maybe you have better ideas?

I have considered

• CentOS
• Scientific Linux (also RPM-based)
• Bio-Linux (Debian/Ubuntu based)
• Arch ?

What we are mostly running: R+loads of Bioconductor, bowtie, tophat, bedtools, samtools, bioperl, slurm (as a queueing system).

For now I have been compiling software and building RPMs from the install dir for an easy removal/upgrade/downgrade using fpm tool. Which isn't great, but works.
CRAN + Bioconductor packages were installed using R + biocLite updaters as root.

Ideally, OS support for a given version should be at least 2-3 years (one post-doc time ), or it should be a rolling release. And of course it should have relatively modern libraries, so that there is no problem with compiling software.
Bio-Linux looks good, but lots of tools are outdated, so not much gain.

On my laptop I used to have Fedora and now I am using Arch. They do great as workstations, but they are very frequently updated, so I not sure if it is a good idea to run them on a "production" server.

Any thoughts?
Cheers,
M

[Richard Finney]

If it's a PC, use MINT or Ubuntu.
Install the "biology" packages yourself; probably from source.

It it's a server, Red Hat or Centos.

Same drill with installing tools : do it yourself.


Make sure your distro is popular :http://distrowatch.com/ so you can get/get for support.

[GenoMax]

If you don't have a lot of time to spend on this then stick with a distro you know (though RedHat keeps changing things in major ways across versions).

If you are up for a challenge then you could select something different (Ubuntu LTS server may be something to look at as that probably has good community support).

[maubp]

Debian-based Bio-Linux is a good option in terms of better than average package support for Bioinformatics tools (see also DebianMed).

[sarvidsson]

I just reinstalled our cluster with Ubuntu 14.04 LTS. Works great for us and has modern libs and great community support.

Some tweaks were necessary to get bcl2fastq (Illumina) and Newbler (Roche) running, however (only RedHat-like distros are officially supported for those). YMMV.

[rhinoceros]

For a server/workstation, IMO Debian or some derivative. Our cluster runs Rocks, which is based on CentOS..

[merilius]

So Debian it is

Thank you all for answers.

I checked Debian Med as maubp suggested. I haven't been using Debian for the last 13 years
It looks very good (the "testing" branch) as it has almost all software I need. It is not a derivative, but what Debian community calls a "pure blend" meaning that all the software is fully integrated into the main repo.