Using the AMR Shiny dashboard

Source: vignettes/using-amr-dashboard.Rmd

Overview

The AMR Shiny dashboard (amRshiny) is a visualization tool for exploring antimicrobial resistance patterns. It is designed to ingest genome metadata from amRdata and custom ML outputs generated by the amRml package and provide interactive exploration of model performance, predictive features, and genome isolate metadata. This allows researchers to compare results across species, drugs/drug classes, molecular scales, and other biologically relevant attributes (e.g., country, time).

Running the dashboard

Install the package, then launch the app:

devtools::install()
library(amRshiny)
launchAMRDashboard

Once the app opens, you should see the AMR Dashboard in your browser.

Using the dashboard

The dashboard is organized into tabs. Each tab shows a different view of the AMR data (e.g., Metadata summaries, Model Performance, or Bug/Drug feature comparisons). Use the filters on each page to update the corresponding plots accordingly.

Common filters include: - Bug/Species - Drug or Drug class - Model scale (i.e. genes, domains, proteins) - Data type (i.e. binary, count)

Questions you can explore

  • Which molecular scale performs best for a given bug–drug combination?
  • Which features are conserved across drugs or species?
  • Do model features shift across geography or time?

Dashboard tabs

The dashboard is organized into five tabs.

Metadata

Explore AMR surveillance metadata for a selected species. Visualizations summarize phenotypes, countries of isolation, hosts, sources, and temporal trends. Data cards provide quick counts of genomes, drugs, drug classes, and other key summaries.

Model performance

Compare model performance across species and drug classes. Select a drug class to filter available drugs, choose a drug and data type (binary or count), and view side‑by‑side performance summaries. Boxplots show performance distributions for the class, with overlaid points for the selected drug.

Bug/Drug feature comparison

Investigate important predictive features across species or drugs.
- Across bugs: compare ranked COGs for a given drug/class across selected species, with links to COG and NCBI HMM annotations.
- Across drugs: within a species, compare top COGs across drugs or classes.

Model holdouts

View performance and important features for country or year holdout models, and compare results across datasets to identify shared or exclusive features.

Query data

Interactively query and extract subsets of the underlying AMR data for downstream analysis. Use filters to define a cohort, then view or download the results for further work.