Minutes 2024-03-08

Author

Eric Nantz

Published

March 8, 2024

  • Ben Straub (GSK)
  • Eric Nantz (Eli Lilly)
  • Nate Mockler (Biogen)
  • Sam Parmar (Pfizer)

Meeting Minutes

  • Eric shared the status of his initial tests of the web-assembly version of the application available at https://github.com/RConsortium/submissions-pilot4-webR
    • Application works successfully when using work computer browser (Edge), however on his Linux machine certain browsers (Edge) were not able to view the application, while other chrome-variant browsers had no issue
    • Sam mentioned he encountered a few issues running the app in his work browser, but was successful after discussions with IT
    • Additional testing will be paramount prior to transfer to FDA
    • Eric will prepare the repository with a few minor enhancements in preparation for wider testing amongst the group. Ben mentioned that with the Pilot 3 effort winding down, there could be additional resources available for testing soon.
  • When the web-R binary packages are being pulled down from the webR CDN, there isn’t a clear way to pin those to a specific version, unlike renv when we can indeed control versions of packages installed. A comment on a separate issue https://github.com/posit-dev/r-shinylive/issues/56 indicates there will be a future state when pinning package versions is possible.
  • Ben mentioned an issue they encountered with the emmeans package being updated such that it requires R version 4.3 or greater. We leverage this package in the pilot 4 app. Eric’s testing did not encounter any issues with this package, but this note has been logged in the issue tracker for awareness.
  • It is unclear whether that will change, but we will be interested in a future keynote presentation from George Stagg at the upcoming ShinyConf 2024 which is going to address reproducibility in the context of web-assembly and web-R.
  • With the latest CRAN releases of shinylive and httpuv, the pilot 4 setup instructions will ensure the CRAN releases for this packages are installed by the user prior to running the web-assembly version of the application
  • When the web-assembly version of the application is created, the bundle is essentially a massive collection of JavaScript files
  • It is not clear yet whether compiling the web-assembly application results in differences between operating systems.
  • While the current pilot 4 web-assembly repository contains a pre-made bundle of the application (built using the suppleid build.R script), we need to verify with FDA if having JavaScript files in the .zip archive of the repository is acceptable for transfer in the eCTD gateway.
  • The ADRG for pilot 4 will leverage existing material from the pilot 2 ADRG, with the majority of updates landing in the installation instructions.
  • The group agreed to have a separate repository created that is tailored for the actual transfer, following a similar mechanism as Pilot 2. Ben mentioned the approach worked well for Pilot 3.
  • The group acknowledges that asking the FDA review to inspect the JavaScript bundle to find the application R code would be very difficult. Therefore the ADRG will include instructions for running the application in the traditional method of launching an R session (bootstrapped with renv) and executing shiny::runApp() much like Pilot 2.
  • Next steps for the pilot include:
    • Elicit additional testers for the web-assembly application in the Slack channel
    • Eric to bootstrap the Windows virtual machine testing environment in a similar fashion as pilot 2’s custom testing environment.
    • Create a new repository for the FDA-ready version of the web-assembly application
    • Initialize a draft of the ADRG