class: center, middle, inverse, title-slide # <span style="color:#ffffff"><strong>Transitioning to <span style="color:orange">R </strong></span> ## <span style="color:#FFFFFF">for Clinical Submissions</span> ### <span style="color:#FFFFFF">twitter:<span class="citation">@rinpharma</span> github:philbowsher</span> --- class: subtitle-slide, white, middle, center # Focus early on package management. Consider hiring an R admin to help manage the package environment for clinical reporting. --- class: panel-narrow-slide, left # Step 1 - **Add RStudio as a Trusted Vendor** - This will speed things up considerable - This will add RStudio software and packages as trusted by IT & QA - [Helpful validation documents are here](https://www.rstudio.com/solutions/pharma/) - Verification docs: [RSC](https://docs.rstudio.com/rsc/post-setup-assistant/#verification-tasks), [RSP](https://docs.rstudio.com/rpm/installation/#step-3-verify-installation), [RSPM](https://docs.rstudio.com/rpm/installation/#step-3-verify-installation), [RSPM2](https://docs.rstudio.com/rspm/admin/getting-started/configuration/#quickstart-cran) --- class: panel-narrow-slide, left # Step 2 - **Get users off of desktop** - If you must use desktops, still follow the advice in the steps below - Desktop can be equally challenging for IT and R users --- class: panel-narrow-slide, left # Step 3 - **Get a Server (cloud or onsite) for Development** - IT will manage the server for users - Have users use RStudio Projects, renv and version control - RStudio IDE has integrated support for version control --- class: panel-narrow-slide, left # Step 4 - **Create an R version and Package plan** - Which version(s) of R will be supported? - How often will you update packages? - Which packages will be validated? - [Package Strategy Maps](https://environments.rstudio.com/reproduce.html) --- class: panel-narrow-slide, left # Step 5 - **Created a Validated repository of packages based on Plan in Step 4** - [RStudio Validation Docs](https://www.rstudio.com/solutions/pharma/) - Add other packages according to your [RISK-BASED APPROACH](https://www.pharmar.org/white-paper/) - This is your Shared Baseline - Freeze repo when you upgrade packages or add R versions - IT can set repo as default via Renviron.site for users --- class: panel-narrow-slide, left # Step 5 Info - IT can surface out the validated packages in RStudio via the System Library - This will prevent the need for users to install packages - Alternatively, Users can install packages from the Validated repo - renv can be used to isolate the packages for the project users are working on - An R package library is tied to a specific version of R, so different R versions have distinct package libraries --- class: panel-narrow-slide, left # Step 5.5 - **As time goes on, Don't upgrade R** - Rather, Install multiple versions of R on the same server - [R Installs](https://docs.rstudio.com/resources/install-r/) - Point older versions of R to frozen repo states - Don't install packages from CRAN - Organizations will have multiple repositories (CRAN, validation etc.) aligning to different use cases --- class: subtitle-slide, white, middle, center # Steps 1-5 are about managing packages and R. Steps 6-10 are about internal processes for your own code, packages and infrastruture. --- class: panel-narrow-slide, left # Step 6 - **Validate your internal packages & UDFs** - Ellis Hughes - https://rinpharma.com/publication/rinpharma_145/ --- class: panel-narrow-slide, left # Step 7 - **Determine your method for data management & TLFs** - RStudio projects, tidyverse, htmlwidgets, RMarkdown, GT - https://db.rstudio.com/ - Roche way, rtables, loon - GSK way, timber, xportr, metacore - Merck way, r2rtf, pkglite, cleanslate - Others, officer, pharmartf? --- class: panel-narrow-slide, left # Step 8 - **Harden Production and not RStudio** - When reporting is done, migrate the processes to a production environment - Learn basic software development practices like testing, CI/CD, automation tools, versioning (git) etc. - Create Shiny apps to standardize reporting and enable new capabilities - [The future of clinical tfls](https://williamnoble.shinyapps.io/the_future_of_clinical_tfls/) --- class: panel-narrow-slide, left # Step 9 - **Create a R cloud strategy for clinical reporting** - [RStudio Cloud tools](https://github.com/rstudio/rstudio-cloud-tools) for VPCs [cloudformation](https://docs.rstudio.com/rstudio-team/cloudformation/) & [ARM](https://portal.azure.com/#create/microsoft.template/uri/https%3A%2F%2Fraw.githubusercontent.com%2Frstudio%2Frstudio-cloud-tools%2Fmaster%2Fazure%2Farm-templates%2Fazuredeploy.json) - Review Kubernetes as a container-orchestration system - Consider Validating Docker Images - If you are installing packages as part of a Dockerfile, it is important to make sure you are setting the correct frozen repo in the install.packages command --- class: panel-narrow-slide, left # Step 10 - **Listen at R in Pharma** - Learn what other pharmas are doing in the DS space - https://rinpharma.com/ --- class: subtitle-slide, white, middle, center # rstudio conference & rinpharma