From Data to Insights: A Hands-On Workshop with {teal}

Interactive clinical trial data exploration framework

Shiny

Clinical Data

Intermediate

Authors

Nina Qi (Principal Data Scientist, Genentech)

Dony Unardi (Data Scientist, Genentech)

Overview

Intermediate Shiny Interactive Apps

{teal} is an innovative open-source R-Shiny framework that has transformed clinical trial data analysis and visualization. This hands-on workshop, built on the latest {teal} 1.0 release, will progressively cover practical topics for building interactive applications that deliver insights faster while promoting efficiency, transparency, and reproducibility.

What You’ll Learn

🎯 {teal} fundamentals - Core concepts and architecture
🧩 Building blocks - Modules and components
📊 Clinical applications - Real-world use cases
🔧 Hands-on exercises - Step-by-step building
🚀 Best practices - Production-ready apps

Prerequisites

Required Knowledge:

Basic R programming
Understanding of clinical trial data (CDISC helpful)
No prior Shiny or {teal} experience required!

All R users welcome - from beginners to experienced developers

Key Tools

{teal}

{teal.modules.clinical}

{teal.data}

Shiny

Workshop Materials

Resources

GitHub Workshop: github.com/pharmaverse/tealworkshop-rinpharma2025

{teal} Website: insightsengineering.github.io/teal

What is {teal}?

Framework Overview

{teal} provides a modular framework for building interactive data exploration applications:

🧩 Modular - Plug-and-play components
🔄 Reactive - Built on Shiny
📊 Clinical-focused - Pharma use cases
✅ Validated - GxP-ready
🌐 Pharmaverse - Part of ecosystem

Key Features

Data Management:

Multiple dataset support
Filtering and subsetting
Data relationships
Reproducible states

Modules:

Pre-built clinical modules
Custom module creation
Module composition
Shared state management

Reproducibility:

Code generation
Session state export
Report generation
Audit trails

The {teal} 1.0 Release

What’s New

✨ Stable API
📦 Improved module system
🎨 Better UI/UX
📚 Enhanced documentation
🔧 Performance improvements

Ecosystem

Core packages:

{teal} - Framework core
{teal.data} - Data management
{teal.code} - Code tracking
{teal.modules.clinical} - Clinical modules
{teal.modules.general} - General modules

Building Your First {teal} App

Minimal Example

library(teal)

# Create app
app <- init(
  data = teal_data(ADSL = pharmaverseadam::adsl),
  modules = modules(
    tm_data_table("Data Table"),
    tm_variable_browser("Variable Browser")
  )
)

# Run
shinyApp(app$ui, app$server)

With Clinical Modules

library(teal)
library(teal.modules.clinical)

# Prepare data
data <- teal_data(
  ADSL = pharmaverseadam::adsl,
  ADAE = pharmaverseadam::adae,
  ADTTE = pharmaverseadam::adtte
)

# Build app
app <- init(
  data = data,
  modules = modules(
    # Demographics
    tm_t_summary(
      label = "Demographics Table",
      dataname = "ADSL",
      arm_var = choices_selected("ARM", "ARM"),
      summarize_vars = choices_selected(
        c("AGE", "SEX", "RACE"),
        c("AGE", "SEX")
      )
    ),
    
    # Adverse Events
    tm_t_events_summary(
      label = "Adverse Events",
      dataname = "ADAE",
      arm_var = choices_selected("ARM", "ARM"),
      flag_var_anl = choices_selected("ANL01FL", "ANL01FL")
    ),
    
    # Kaplan-Meier
    tm_g_km(
      label = "Survival Analysis",
      dataname = "ADTTE",
      arm_var = choices_selected("ARM", "ARM"),
      paramcd = choices_selected("PARAMCD", "OS")
    )
  ),
  title = "Clinical Trial Explorer"
)

shinyApp(app$ui, app$server)

Workshop Exercises

Exercise 1: Basic App (30 min)

Goal: Create simple data exploration app

Tasks:

Load ADSL dataset
Add data table module
Add variable browser
Run and explore

Exercise 2: Demographics Module (45 min)

Goal: Build demographics summary

Tasks:

Configure demographics module
Select variables to summarize
Add treatment arm comparison
Generate table

Exercise 3: Adverse Events (45 min)

Goal: Create AE summary app

Tasks:

Load ADAE dataset
Configure AE module
Filter by severity
Add system organ class view

Exercise 4: Survival Analysis (45 min)

Goal: Add Kaplan-Meier plots

Tasks:

Load ADTTE dataset
Configure KM module
Add risk tables
Customize plot appearance

Exercise 5: Complete Clinical App (60 min)

Goal: Combine everything

Tasks:

Multiple modules
Data filtering
Custom styling
Code generation
Export capabilities

Advanced Topics

Custom Modules

Create your own {teal} modules:

tm_my_custom_module <- function(label, dataname) {
  module(
    label = label,
    server = function(input, output, session, data) {
      # Your server logic
    },
    ui = function(id) {
      # Your UI
    }
  )
}

Data Filtering

Global filters across all modules:

app <- init(
  data = data,
  modules = modules(...),
  filter = teal_slices(
    teal_slice(dataname = "ADSL", varname = "SAFFL", selected = "Y"),
    teal_slice(dataname = "ADAE", varname = "AESER", selected = "Y")
  )
)

Code Generation

Users can extract R code for reproducibility:

# Built-in "Show R Code" button
# Generates exact code to reproduce current view
# Can be run independently

Real-World Applications

Clinical Study Reports

Interactive data review
Ad-hoc analyses
Quality checks
Stakeholder presentations

Safety Monitoring

Real-time AE tracking
Signal detection
Dose-escalation decisions
DSMB reports

Efficacy Analysis

Endpoint visualization
Subgroup exploration
Sensitivity analyses
Forest plots

Exploratory Analysis

Hypothesis generation
Biomarker discovery
Population characterization
Covariate relationships

Benefits of {teal}

For Statisticians

✅ Fast exploration without coding
✅ Reproducible analyses
✅ Code generation
✅ Publication-ready outputs

For Programmers

✅ Modular development
✅ Reusable components
✅ Validated framework
✅ Extensible architecture

For Organizations

✅ Consistency across studies
✅ Reduced development time
✅ GxP compliance ready
✅ Lower training burden

Deployment

Local

# Run locally
shinyApp(app$ui, app$server)

Posit Connect

# Deploy to Connect
rsconnect::deployApp(
  appDir = "path/to/app",
  appTitle = "Clinical Trial Explorer"
)

Docker

Package as Docker container for enterprise deployment.

Best Practices

Design

Keep modules focused
Use meaningful labels
Provide helpful tooltips
Test with real users

Performance

Pre-process data when possible
Use reactive expressions wisely
Implement caching
Monitor memory usage

Validation

Document module behavior
Test edge cases
Version control
Maintain change log

Learning Outcomes

✅ Understand {teal} architecture
✅ Build basic {teal} applications
✅ Use clinical modules effectively
✅ Create custom modules
✅ Implement data filtering
✅ Deploy production-ready apps
✅ Ensure reproducibility

Resources

Documentation:

{teal} website with vignettes
Module reference guides
API documentation

Community:

Pharmaverse Slack (#teal channel)
GitHub discussions
Monthly office hours

Examples:

Gallery of {teal} apps
Template repositories
Use case studies

Next Steps

After the workshop:

Practice - Build an app for your data
Explore - Try different modules
Customize - Create custom modules
Deploy - Share with your team
Contribute - Join pharmaverse community

Similar Workshops

Cardinal: Clinical Reporting - Related to TFLs
HPC Cluster with Shiny - Advanced Shiny

Next Steps

For validation: Validating Shiny Apps presentation
Career skills: Shiny Development

Last updated: November 2025 | R/Pharma 2025 Conference

--- title: "From Data to Insights: A Hands-On Workshop with {teal}" subtitle: "Interactive clinical trial data exploration framework" author: - "Nina Qi (Principal Data Scientist, Genentech)" - "Dony Unardi (Data Scientist, Genentech)" categories: [Shiny, Clinical Data, Intermediate] --- ## Overview [Intermediate]{.badge .badge-intermediate} [Shiny]{.badge .badge-category} [Interactive Apps]{.badge .badge-category} **{teal}** is an innovative open-source R-Shiny framework that has transformed clinical trial data analysis and visualization. This hands-on workshop, built on the latest {teal} 1.0 release, will progressively cover practical topics for building interactive applications that deliver insights faster while promoting efficiency, transparency, and reproducibility. ### What You'll Learn - 🎯 **{teal} fundamentals** - Core concepts and architecture - 🧩 **Building blocks** - Modules and components - 📊 **Clinical applications** - Real-world use cases - 🔧 **Hands-on exercises** - Step-by-step building - 🚀 **Best practices** - Production-ready apps ## Prerequisites ::: requirements **Required Knowledge:** - Basic R programming - Understanding of clinical trial data (CDISC helpful) - No prior Shiny or {teal} experience required! **All R users welcome** - from beginners to experienced developers ::: ## Key Tools ::: tool-tag {teal} ::: ::: tool-tag {teal.modules.clinical} ::: ::: tool-tag {teal.data} ::: ::: tool-tag Shiny ::: ## Workshop Materials ::: callout-note ## Resources **GitHub Workshop:** [github.com/pharmaverse/tealworkshop-rinpharma2025](https://github.com/pharmaverse/tealworkshop-rinpharma2025/tree/main) **{teal} Website:** [insightsengineering.github.io/teal](https://insightsengineering.github.io/teal) ::: ## What is {teal}? ### Framework Overview **{teal}** provides a modular framework for building interactive data exploration applications: - 🧩 **Modular** - Plug-and-play components - 🔄 **Reactive** - Built on Shiny - 📊 **Clinical-focused** - Pharma use cases - ✅ **Validated** - GxP-ready - 🌐 **Pharmaverse** - Part of ecosystem ### Key Features **Data Management:** - Multiple dataset support - Filtering and subsetting - Data relationships - Reproducible states **Modules:** - Pre-built clinical modules - Custom module creation - Module composition - Shared state management **Reproducibility:** - Code generation - Session state export - Report generation - Audit trails ## The {teal} 1.0 Release ### What's New - ✨ Stable API - 📦 Improved module system - 🎨 Better UI/UX - 📚 Enhanced documentation - 🔧 Performance improvements ### Ecosystem **Core packages:** - `{teal}` - Framework core - `{teal.data}` - Data management - `{teal.code}` - Code tracking - `{teal.modules.clinical}` - Clinical modules - `{teal.modules.general}` - General modules ## Building Your First {teal} App ### Minimal Example ``` r library(teal) # Create app app <- init( data = teal_data(ADSL = pharmaverseadam::adsl), modules = modules( tm_data_table("Data Table"), tm_variable_browser("Variable Browser") ) ) # Run shinyApp(app$ui, app$server) ``` ### With Clinical Modules ``` r library(teal) library(teal.modules.clinical) # Prepare data data <- teal_data( ADSL = pharmaverseadam::adsl, ADAE = pharmaverseadam::adae, ADTTE = pharmaverseadam::adtte ) # Build app app <- init( data = data, modules = modules( # Demographics tm_t_summary( label = "Demographics Table", dataname = "ADSL", arm_var = choices_selected("ARM", "ARM"), summarize_vars = choices_selected( c("AGE", "SEX", "RACE"), c("AGE", "SEX") ) ), # Adverse Events tm_t_events_summary( label = "Adverse Events", dataname = "ADAE", arm_var = choices_selected("ARM", "ARM"), flag_var_anl = choices_selected("ANL01FL", "ANL01FL") ), # Kaplan-Meier tm_g_km( label = "Survival Analysis", dataname = "ADTTE", arm_var = choices_selected("ARM", "ARM"), paramcd = choices_selected("PARAMCD", "OS") ) ), title = "Clinical Trial Explorer" ) shinyApp(app$ui, app$server) ``` ## Workshop Exercises ### Exercise 1: Basic App (30 min) **Goal:** Create simple data exploration app **Tasks:** - Load ADSL dataset - Add data table module - Add variable browser - Run and explore ### Exercise 2: Demographics Module (45 min) **Goal:** Build demographics summary **Tasks:** - Configure demographics module - Select variables to summarize - Add treatment arm comparison - Generate table ### Exercise 3: Adverse Events (45 min) **Goal:** Create AE summary app **Tasks:** - Load ADAE dataset - Configure AE module - Filter by severity - Add system organ class view ### Exercise 4: Survival Analysis (45 min) **Goal:** Add Kaplan-Meier plots **Tasks:** - Load ADTTE dataset - Configure KM module - Add risk tables - Customize plot appearance ### Exercise 5: Complete Clinical App (60 min) **Goal:** Combine everything **Tasks:** - Multiple modules - Data filtering - Custom styling - Code generation - Export capabilities ## Advanced Topics ### Custom Modules Create your own {teal} modules: ``` r tm_my_custom_module <- function(label, dataname) { module( label = label, server = function(input, output, session, data) { # Your server logic }, ui = function(id) { # Your UI } ) } ``` ### Data Filtering Global filters across all modules: ``` r app <- init( data = data, modules = modules(...), filter = teal_slices( teal_slice(dataname = "ADSL", varname = "SAFFL", selected = "Y"), teal_slice(dataname = "ADAE", varname = "AESER", selected = "Y") ) ) ``` ### Code Generation Users can extract R code for reproducibility: ``` r # Built-in "Show R Code" button # Generates exact code to reproduce current view # Can be run independently ``` ## Real-World Applications ### Clinical Study Reports - Interactive data review - Ad-hoc analyses - Quality checks - Stakeholder presentations ### Safety Monitoring - Real-time AE tracking - Signal detection - Dose-escalation decisions - DSMB reports ### Efficacy Analysis - Endpoint visualization - Subgroup exploration - Sensitivity analyses - Forest plots ### Exploratory Analysis - Hypothesis generation - Biomarker discovery - Population characterization - Covariate relationships ## Benefits of {teal} ### For Statisticians - ✅ Fast exploration without coding - ✅ Reproducible analyses - ✅ Code generation - ✅ Publication-ready outputs ### For Programmers - ✅ Modular development - ✅ Reusable components - ✅ Validated framework - ✅ Extensible architecture ### For Organizations - ✅ Consistency across studies - ✅ Reduced development time - ✅ GxP compliance ready - ✅ Lower training burden ## Deployment ### Local ``` r # Run locally shinyApp(app$ui, app$server) ``` ### Posit Connect ``` r # Deploy to Connect rsconnect::deployApp( appDir = "path/to/app", appTitle = "Clinical Trial Explorer" ) ``` ### Docker Package as Docker container for enterprise deployment. ## Best Practices ### Design - Keep modules focused - Use meaningful labels - Provide helpful tooltips - Test with real users ### Performance - Pre-process data when possible - Use reactive expressions wisely - Implement caching - Monitor memory usage ### Validation - Document module behavior - Test edge cases - Version control - Maintain change log ## Learning Outcomes ✅ Understand {teal} architecture\ ✅ Build basic {teal} applications\ ✅ Use clinical modules effectively\ ✅ Create custom modules\ ✅ Implement data filtering\ ✅ Deploy production-ready apps\ ✅ Ensure reproducibility ## Resources **Documentation:** - {teal} website with vignettes - Module reference guides - API documentation **Community:** - Pharmaverse Slack (#teal channel) - GitHub discussions - Monthly office hours **Examples:** - Gallery of {teal} apps - Template repositories - Use case studies ## Next Steps After the workshop: 1. **Practice** - Build an app for your data 2. **Explore** - Try different modules 3. **Customize** - Create custom modules 4. **Deploy** - Share with your team 5. **Contribute** - Join pharmaverse community ------------------------------------------------------------------------ ### Similar Workshops - [Cardinal: Clinical Reporting](../clinical-reporting/cardinal-tlgs.qmd) - Related to TFLs - [HPC Cluster with Shiny](hpc-cluster-shiny.qmd) - Advanced Shiny ### Related Presentations - [admiral's Journey to Stability](../presentations/europe-us-sessions.qmd#r-we-there-yet-admirals-journey-to-stability) - Related pharmaverse package ### Next Steps - **For validation:** [Validating Shiny Apps presentation](../presentations/europe-us-sessions.qmd#validating-shiny-apps-in-regulated-environments) - **Career skills:** [Shiny Development](../summary/career-insights.qmd#4-shiny-application-development) ------------------------------------------------------------------------ *Last updated: November 2025 \| R/Pharma 2025 Conference*