Visualization

lobster-visualization

FreeBeginner

Publication-quality visualizations: UMAP, violin plots, heatmaps, and volcano plots

Input

AnnDataDE ResultsCluster LabelsGene Lists

Output

UMAP PlotsViolin PlotsHeatmapsVolcano PlotsDot Plots

Agents (1)

└── visualization_expert — Publication-quality plotting and figure generation

pip install lobster-visualization

Agents

visualization_expert

Specialized agent for creating visualizations from analyzed omics data.

Capabilities:

UMAP/t-SNE plots with gene overlays
Violin plots and dot plots
Heatmaps with clustering
Volcano plots and MA plots
Custom plot styling

Supported Plot Types

Plot Type	Use Case	Best For
UMAP	Dimensionality reduction	Cell clustering, embeddings
t-SNE	Dimensionality reduction	Small datasets, local structure
Violin	Gene expression distribution	Comparing expression across groups
Dot plot	Gene markers across clusters	Marker gene visualization
Heatmap	Expression patterns	Pathway/gene set analysis
Volcano	Differential expression	Identifying significant genes
MA plot	Fold change vs abundance	Quality check for DE analysis

Services

lobster-visualization includes visualization services bundled with the package:

Service	Purpose
SingleCellVisualizationService	UMAP, t-SNE, violin plots, dot plots, heatmaps

The service is installed automatically with the agent package.

Example Workflows

UMAP with Gene Overlay

User: Create a UMAP colored by cell type with CD3E expression overlay

[visualization_expert]
- Generates UMAP with categorical coloring
- Overlays CD3E expression as continuous color
- Saves to plots/
- Reports how many files were saved and up to three of their names

Marker Gene Dot Plot

User: Show me the expression of marker genes across clusters

[visualization_expert]
- Retrieves marker genes from analysis state
- Creates dot plot with cluster on x-axis
- Size = percent expressing, color = mean expression
- Saves publication-ready figure

Volcano Plot for DE Results

User: Make a volcano plot from my differential expression results

[visualization_expert]
- Loads DE results from transcriptomics analysis
- Plots log2 fold change vs -log10 p-value
- Highlights significant genes (|FC| > 1, padj < 0.05)
- Labels top significant genes

Dependencies

lobster-visualization requires visualization libraries:

Library	Purpose
matplotlib	Core plotting library
seaborn	Statistical visualizations
scanpy	Single-cell plotting integration
plotly	Interactive plots

These are installed automatically with the package.

Configuration

# .lobster_workspace/config.toml
enabled = ["visualization_expert"]

Plot Storage

Visualization tools save plots to the workspace plots/ folder as interactive HTML files, plus a PNG when image export succeeds and is not skipped. Saving can be skipped or can fail:

.lobster_workspace/plots/
|-- plot_1_UMAP_cell_type.html
|-- plot_1_UMAP_cell_type.png
|-- plot_2_Volcano_treated_vs_control.html
|-- plot_2_Volcano_treated_vs_control.png

Since Lobster AI 1.1.433, the create_interactive_plot tool reports the workspace-relative path of each file it saved, or the reason a save was skipped or failed. Other visualization tools report a count of saved files and may not include these details. PNG export is skipped when a plot's recorded cell count exceeds 50,000, and it fails when the image export engine is not available; a PNG skip or failure does not remove an HTML file that was saved. Saved plots also appear in the workspace file listing that agents see during hand-offs.

Plot Customization

visualization_expert supports plot customization through natural language:

User: Make the UMAP larger with a white background and save as PDF

[visualization_expert adjusts]
- Figure size: 12x10 inches
- Background: white
- Format: PDF vector graphics

Customizable Parameters

Parameter	Options	Default
Figure size	Any dimensions	8x6 inches
DPI	72-600	150
Format	PNG, PDF, SVG	PNG
Color palette	Any matplotlib/seaborn	Default
Font size	Any size	12pt
Background	White, transparent	White

Integration with Analysis Agents

visualization_expert receives data from analysis agents:

transcriptomics_expert
  -> Performs clustering
  -> Stores results in state
  -> Calls visualization_expert
    -> Reads cluster assignments
    -> Generates UMAP plot

Next Steps

Single-Cell Tutorial

Complete workflow with visualization examples

Transcriptomics Agent

Generate data for visualization

Export Your Work

Export analyses as PDF reports, notebooks, or slide decks

Visualization

Next Steps

On this page