2:00 - 2:30 PM - Scaling Shiny: Experiences from building a large application in Pharma
The shiny app enables a metadata driven approach to creating mock shells and Clinical Study Reports outputs at Idorsia. It aims to create a single point of access platform for a process that requires integrating multiple external components, streamlining the workflow and enabling statistical programmer to work more efficiently.
The talk is about the design considerations of building out this large app, examining how far Shiny can be stretched, and details some of the technical implications in integrating the different components that the app relies on. E.g. SAS Webserver, Database, S3 Bucket, Pinnacle 21, and others.
It is currently used in production on a pivotal phase 3 study. By the time this talk is live it will be used on 5 studies.
The application is relied upon by multiple people in a production environment whilst new features are constantly being added. It is therefore important to have readable and glanceable code so one can easily spot and address issues quickly.
We learned pretty early on to separate concerns into different packages, e.g. one wrapping a specific API, another to handle updates on data, push long-running processes in the background, etc.
These challenges were interesting to solve and worth sharing with the community.