Proposal to use branches for this repo to fix issues in PharmacoDI #23
Description
We have recently run into the issue that annotation files are being updated before those changes can be propagated into the PharmacoSet objects via ORCESTRA.
This is causing significant issues in PharmacoDI when trying to generate the PharmacoDB 2.0 tables, since the PharmacoSet data is used as the ground truth and therefore in the ETL pipeline compound and sample identifiers are used to look-up additional metadata from this repository. If the identifiers don't match, this results in (1) errors for me to debug and/or (2) corrupted, missing or incorrect metadata.
Thus far, I have been patching the issues manually. This is neither efficient nor sustainable.
To solve this issue, I suggest we implement a new branch to this repository. I am not particular about the naming conventions. Typically I will have a development and master branch for most of my repos, so that could work. Other suggestions welcome.
Assuming the proposed naming conventions, the master branch would be stable. It would not receive any changes until ORCESTRA is triggered to rebuild the PharmacoSets. At that time, PharmacoSets could be built from the development branch and a PR opened to merge development into master.
This way we can have active improvements to our annotations, corrected as they are discovered, while ensuring that PharmacoDI works correctly.
RE: Naming conventions, we could keep master as the development branch, and create a release branch or something if that is preferable.
Thoughts on this would be appreciated. We could probably automate the PR using web hooks and GitHub Actions.