Remove Docker dependency

[rchurt]

Hello,

I'd like to use Corason on a Google Colab virtual machine, but sadly they don't support Docker and have no plans to do so (googlecolab/colabtools#299). Unlike most packages that only use Docker for installation (e.g., BiG-SCAPE), I see that Corason actually uses Docker in its execution (e.g., docker run ... in run_corason). Would it be very difficult to modify the scripts so that they can run without Docker?

Thanks!

[nselem]

Hello
Other packages has a cover script that writes de docker run instruction for you, but it canot be just the instalation, because docker is a container to run software and its dependencies inside it. CORASON is a set of perl scripts, with dependencies on blast, muscle, gblocks newick utilities and fasttree. It does not need docker in its execution, docker is just the container. You would need to install those dependencies in the paths indicated in the dockerfile, it could be troublesome. I think easiest way is to get accesion to a LINUX machine, for example in Amazon Web Services. Sorry, these are not the best news for you.

[rchurt]

I see, then perhaps I have some confusion. In the installation instructions for BiG-SCAPE under "Manual installation," there's no dependency on Docker, and I install the other dependencies via Conda, which works great in Google Colab. Since you said that CORASON also doesn't depend on Docker, is a similar "manual installation" possible for CORASON, and how could I do it? Currently when I clone the GutHub repo and run run_corason, I get the error: /usr/local/envs/CORASON/corason/run_corason: line 42: docker: command not found

CCing @jorgecnavarrom in case he has some tips

[nselem]

Its because BiG-SCAPE has two possible instalation forms, one is with conda and another one with docker, since CORASON is not python it has not the option to use conda. It will be hard to run a manual installation, for the dependencies, that is why conda and docker are valuable resources. For CORASON you need a linux or mac machine in order tat docker runs. Nevertheless I can run your data for you, If you whish.

[jorgecnavarrom]

CORASON is a set of perl scripts, with dependencies on blast, muscle, gblocks newick utilities and fasttree

So in principle if those dependencies are installed (perhaps even through conda) it should be possible to run the CORASON perl scripts right? (e.g. https://github.com/nselem/corason/blob/master/CORASON/corason.pl)

[rchurt]

Yup that's what I'm thinking, and fortunately Conda does support Perl. I'll post some code for how to install those dependencies in a Perl virtual environment via Conda, but there will need to be a few changes to the master script that calls the others, right?

[nselem]

Yes, we can try, the paths would be what we must change, but it would be great to have CORASON in CONDA too.

[rchurt]

This sounds great, and I don't think it should be too hard.

After installing Conda, you can run the following commands in terminal:

# Add channels to look for packages in
conda config --add channels bioconda
conda config --add channels conda-forge

# Create virtual environment for CORASON
conda create -y --prefix /usr/local/envs/CORASON

# Install dependencies into virtual environment
conda install -y --name CORASON perl perl-bioperl perl-app-cpanminus fasttree quicktree blast muscle newick_utils biopython gblocks vim

# Clone CORASON from Git and install in virtual environment
cd /usr/local/envs/CORASON
git clone https://github.com/nselem/corason.git
chmod a+x /usr/local/envs/CORASON/corason/run_corason

# Install cpan modules in virtual environment
conda activate CORASON
cpanm SVG
## I'm not sure if you need to install these from cpan or if they're included in the perl-bioperl module above
cpanm Bio::SeqIO
cpanm Bio::SeqFeature::Generic
cpanm Bio::Seq
conda deactivate

# Check that everything was installed correctly
conda activate CORASON
bash /usr/local/envs/CORASON/corason/run_corason --help
conda deactivate

This should get you started, but let me know if you run into any Conda problems. Do you also need to install any cpan packages? If so I can give you code for that.

[nselem]

Ok, I will try, give me some days. :) Yes we need svg from cpan, basically, all needs are in the Dockerfile of the github repo.

[rchurt]

Ok, I just edited the above to include that. The code above should work, but if you run into any problems with cpan in conda, you can look here for troubleshooting

You may need to put these two lines at the top of Perl scripts you want to run, but I'm not sure--it might just be enough to do conda activate CORASON in terminal before calling the Perl script, as above.

module load anaconda
source activate CORASON

Please do let me know if I can help more with the Conda stuff. I know others use Conda + Perl a lot, so I'm sure we can get it working, and I think that would be great for users.

[rchurt]

Hi @nselem,
Just wanted to check in to see how this is going and if there's anything I can do to help--just let me know.

[jcpasion]

Hello @rchurt,

This sounds great, and I don't think it should be too hard.

After installing Conda, you can run the following commands in terminal:

# Add channels to look for packages in
conda config --add channels bioconda
conda config --add channels conda-forge

# Create virtual environment for CORASON
conda create -y --prefix /usr/local/envs/CORASON

# Install dependencies into virtual environment
conda install -y --name CORASON perl perl-bioperl perl-app-cpanminus fasttree quicktree blast muscle newick_utils biopython gblocks vim

# Clone CORASON from Git and install in virtual environment
cd /usr/local/envs/CORASON
git clone https://github.com/nselem/corason.git
chmod a+x /usr/local/envs/CORASON/corason/run_corason

# Install cpan modules in virtual environment
conda activate CORASON
cpanm SVG
## I'm not sure if you need to install these from cpan or if they're included in the perl-bioperl module above
cpanm Bio::SeqIO
cpanm Bio::SeqFeature::Generic
cpanm Bio::Seq
conda deactivate

# Check that everything was installed correctly
conda activate CORASON
bash /usr/local/envs/CORASON/corason/run_corason --help
conda deactivate

This should get you started, but let me know if you run into any Conda problems. Do you also need to install any cpan packages? If so I can give you code for that

I can try this as well, would love to have this tool in Conda.

[jcpasion]

I followed these steps am at the point where I can actually run CORASON. Of note, the program run_corason points to docker, so, according to some documentation I found, instead we should be using corason.pl.

Unfortunately, I am working with proprietary bacterial genomes and am not able to upload them to RAST for annotation. For my own research, I may have to build phylogenetic trees from the Big-Scape network data instead.

[jcpasion]

Doesn't seem like I'll be able to test CORASON because the paremeters for "run_corason" and "corason.pl" are different, and I can't run corason.pl without RAST IDs. Is there a way to run it without RAST IDs? I'm trying to use the example dataset from the website.
@nselem

[nselem]

Hello jc, sorry, Im late with conda version, let me keep trying, you can use gbks. We can set a meeting next week and try do it togheter

[rchurt]

Hey @nselem, just wanted to see how this is going--did you manage to get it to work in Conda without Docker?