Skip to content

04_Interactive_Exploration_Shiny.md

Latest commit

 

History

History
280 lines (186 loc) · 7.03 KB

04_Interactive_Exploration_Shiny.md

File metadata and controls

280 lines (186 loc) · 7.03 KB

Using Shiny to Prioritize Genes of Interest

Overview

Shiny_Hotgenes() can be used not only to explore results interactively, but also to prioritize genes of interest from multiple complementary perspectives. A typical workflow starts with differential expression results, then refines candidates with multivariate structure from PCA, overlap across contrasts from Venn diagrams, and pathway-level context from the GSEA tab.

library(Hotgenes)

App Overview

The image below shows the full Shiny interface across the main tabs.

Hotgenes Shiny app overview

Hotgenes Shiny app overview

1. Launching the App with a Single Object

The simplest call passes one Hotgenes object:

# Load a pre-built example
fit_Hotgenes <- readRDS(
  system.file("extdata", "fit_Hotgenes.RDS",
              package = "Hotgenes",
              mustWork = TRUE)
)

Shiny_Hotgenes(fit_Hotgenes)

The app opens in your default browser (or the RStudio viewer pane). Stop it by pressing Escape or closing the browser tab.

2. Start with DE Stats

The DE Stats tab is often the first step for identifying candidate genes. Filters on adjusted p-value, fold change, and labels provide a fast first-pass shortlist, and the heatmap subtab helps review global expression patterns across samples. In strong contrasts, this step can still produce long candidate lists.

DE statistics and subtab views captured from the interactive app.

 DE Stats / Statistics subtab DE Stats / Statistics subtab DE Stats / Volcano Plots subtab DE Stats / Volcano Plots subtab DE Stats / Heatmap subtab DE Stats / Heatmap subtab

3. Refine Candidates with PCA

The PCA tab helps refine candidate genes by revealing structure linked to key biological conditions.

  • PCA shows sample separation patterns.
  • Clustered Quali Vars links metadata variables to dominant axes.
  • Clustered Features highlights features driving those patterns (and can be shared to the GSEA tab).
  • PCA Heatmap provides a compact annotated view of clustered features.

The PCA tab provides sample-level principal component analysis.

PCA scatter plot Clustered Quali Vars subtab Clustered Features subtab PCA Heatmap subtab

4. Compare Key Contrasts with Venn Diagram

The Venn Diagram tab helps narrow broad DE lists by identifying:

  • genes shared across selected contrasts (with or without directionality),
  • genes specific to one contrast,
  • smaller intersection sets that are easier to prioritize.

After identifying shared or contrast-specific sets, review those candidates in the VennD Heatmap subtab.

Overlap analysis across selected contrasts.

 Venn Diagram tab Venn Diagram tab

5. Link Genes to Biology with GSEA

The GSEA tab connects DE patterns to biological pathways. This step helps move from statistically significant genes to biologically grounded candidates by inspecting enriched pathways and leading-edge genes. Genes prioritized from enriched pathways or leading-edge results can then be reviewed in the Heatmap subtab (within GSEA) and the Expression Plots tab.

GSEA tab for pathway-level interpretation and leading-edge exploration.

 GSEA tab GSEA tab

6. Visualize Selected Genes with Expression Plots

Once genes are prioritized, the Expression Plots tab is useful to inspect expression trajectories for selected genes across samples and conditions.

Expression trajectory plots for selected genes and contrasts.

 Expression Plots tab Expression Plots tab

7. Comparing Multiple Objects in the Same App

Pass a named list of Hotgenes objects to switch between experiments in the app. The names become labels in the dataset selector.

# Load a second object (DESeq2-based example)
dds_Hotgenes <- readRDS(
  system.file("extdata", "dds_Hotgenes.RDS",
              package = "Hotgenes",
              mustWork = TRUE)
) |> update_object()

# Combine into a named list
Hotgenes_list <- list(
  limma_Ewing  = fit_Hotgenes,
  DESeq2_Ewing = dds_Hotgenes
)

Shiny_Hotgenes(Hotgenes_list)

8. Tips for Large Datasets

Tip Rationale
Store objects as .RDS files and load them with readRDS() Avoids re-running expensive analyses every session
Use update_object() when loading saved objects Ensures slot structure matches the current package version
Subset contrasts with contrasts = c("A_vs_B") Reduces memory usage in app tables and plots
Set padj_cut and .log2FoldChange tightly Speeds up Venn and heatmap calculations for large feature spaces
Pre-compute GSEA results before launching the app Reduces waiting time during pathway exploration

Summary

Tab_or_Feature Role_in_prioritization
DE Stats First-pass filtering by DE statistics and thresholds
PCA Refines candidates using multivariate sample/feature structure
Venn Diagram Finds shared vs contrast-specific candidates
GSEA Links candidate genes to enriched biological pathways
Expression Plots Validates expression behavior of selected genes
Multi-object selector Compares candidate behavior across Hotgenes objects