Potential Usage of Rix in clinical trial flow

[pzhang-cims]

Hi,

Thank you so much for the contributions of the package. I have been watching the corresponding presentations (including today's shiny conference topic from Eric) and find this might be a more convenient tool for the consistent environment. I have some quick questions about potential usage in pharma area:

1. I believe currently more and more organizations start to build their RStudio/Shiny server (e.g. Posit Workbench/Connect). Is package something that can be compatible with them, or more for personal usage?

2. Based on recent communications with personnel involved in R regulatory submission, current alignment between Sponsor and regulatory is to communicate about the environment requirements (R version, OS). To facilitate a reproducible environment at regulatory end, as they might not have a server-level RStudio, current practices ask data scientist to prepare/test from previous server level to desktop/local level for reproducibility, so that the desktop version from regulatory can possibly work. I'm wondering if Rix can be a potential tool to

   • "Copy/export" the specs/environment from server level to local level to create the same environment, and
   • Enable the regulatory to create same environment as well through rix configuration.

Thanks in advance and appreciate for the effort!

[b-rodrigues]

Thanks for your interest and encouraging words!

1. I believe currently more and more organizations start to build their RStudio/Shiny server (e.g. Posit Workbench/Connect). Is package something that can be compatible with them, or more for personal usage?

Could you elaborate? Do you mean if you can use rix to generate an expression to set up a shiny server? That should indeed be possible.

2. Based on recent communications with personnel involved in R regulatory submission, current alignment between Sponsor and regulatory is to communicate about the environment requirements (R version, OS). To facilitate a reproducible environment at regulatory end, as they might not have a server-level RStudio, current practices ask data scientist to prepare/test from previous server level to desktop/local level for reproducibility, so that the desktop version from regulatory can possibly work. I'm wondering if Rix can be a potential tool to>
   • "Copy/export" the specs/environment from server level to local level to create the same environment, and
   • Enable the regulatory to create same environment as well through rix configuration.

Well, rix generates a single default.nix that fully defines the environment an analysis was produced in. This default.nix can be used on a server, locally, on continuous integration...
So you could share this file with a regulatory team, and, if they install Nix, they can exactly recreate the same environment to review an analysis.

[pzhang-cims]

@b-rodrigues Thank you for the quick turnaround! I think for the point 2, if there can be a pilot submission to prove that, and if health authority can accept other system dependency installation (e.g., Nix), this can be a mor efficient way that renv::restore() through detailed process (which can be long). For the point 1, sorry after iterating this question, I feel like I should dive more into Rix first and get you back. Thank you so much for your perspective!

[b-rodrigues]

Well, that would certainly save everyone a lot of time! Nix is actually built around reproducibility, and it goes to great lengths to ensure that a build process always results in the exact same artifacts.

There's this paper that uses rix to generate the default.nix file: https://github.com/jgeller112/L2_VWP_Webcam
you could re-use the same file to get exactly the same environment to reproduce the results of this paper.

There's also this academic reference that may interest you: https://www.usenix.org/legacy/event/lisa04/tech/full_papers/dolstra/dolstra.pdf

[pzhang-cims]

Thank you so much for the materials! I will pass along to our IT and Software Development team for more discussions.