Add a CI through GitHub actions to check that curation changes don't break the execution

[pcm32]

Currently curation PRs are not being checked in terms of their potential to break Snakemake workflow execution. It would be good if the Snakemake workflow can be run through GitHub actions to ensure that the curation changes don't break it.

The steps needed to run on GitHub actions can be partly replicated from the Jenkins jobs, and the logic for the ./run_snakemake.sh seen there can be found in the internal Gitlab repo ebi-gene-expression/snakemake-codon-runs, in the file anatomical_systems_mappings/run_snakemake.sh. That specific file is written to run the workflow on LSF, you would need to modify it and leave that different version publicly available (maybe on this same git public repo) so that it can be consumed by GitHub actions.

For github actions, you will likely need to do:

• Make sure that the GH Actions VM has singularity (through https://github.com/marketplace/actions/setup-singularity for instance)
• Possibly make sure that conda is installed
• Install Snakemake
• Avoid any LSF references for Snakemake
• Possibly remove all prefixes (conda, singularity)

[pcm32]

Alternatively this could be done with Jenkins as well if it is easier, there is a way of connecting GitHub to Jenkins behind firewalls so that the status is reported as a check here.

[YalanBi]

Hi @pcm32! I wonder how to get all the bulk atlas experiments to run snakemake? Should I download via ftp?

[pcm32]

Well, all is maybe too much. I would suggest picking a few and downloading from the FTP.