Dr Cook:data Visualizer

data-visualizer

1. Overview

data-visualizer produces publication-quality figures for academic manuscripts, generating complete, runnable code in ggplot2 or matplotlib with correct resolution, font sizes, and color palettes. Three figure categories are supported.

Omics figures: Volcano plots, heatmaps, PCA/UMAP scatter plots, and enrichment bubble charts — produced from bioinformatics analysis outputs such as DESeq2 results or clusterProfiler enrichment tables. Typically downstream from bioinformatics-assistant.

Clinical and statistical figures: Boxplots, violin plots, Kaplan-Meier survival curves, and forest plots for meta-analyses. These communicate group comparisons, patient outcomes, and pooled effect sizes.

Network figures: PPI network visualization and Compound-Target-Pathway figures for TCM network pharmacology. Output is Cytoscape-ready edge/node tables or R igraph code.

2. Parameters

Required

Optional

3. Workflow

Step 1 — Check upstream context_output. Inherit parameters.domain, parameters.target_journal, and parameters.language without asking. If raw_text contains tabular data (e.g., DESeq2 columns log2FoldChange, padj, gene), treat it as data_description.

Step 2 — Collect figure_type and data_description. If figure_type is not specified by the user, present the figure menu:

"Which type of figure would you like? (1) Volcano plot — DEG results (2) Heatmap — expression matrix (3) PCA / UMAP — dimensionality reduction (4) Boxplot / Violin plot — group comparison (5) Survival curve (Kaplan-Meier) — time-to-event data (6) Forest plot — meta-analysis effect sizes (7) Network figure — PPI or Compound-Target-Pathway (8) Enrichment bubble chart — GO/KEGG results (9) Describe your own figure"

After figure_type is known, confirm data_description if not already provided. Accept a pasted data sample, column name list, or reference to an upstream R object (e.g., "DESeq2 res with gene, log2FoldChange, padj columns").

Step 3 — Confirm tool and color scheme. Default tool to ggplot2. If target_journal is known, auto-suggest color scheme:

• Nature, Science, Cell → journal_bw; note "Color figures incur additional fees in print."
• Chinese journals → default (full color acceptable).
• All others → default.

Apply the suggested or default palette without asking if color_scheme is unspecified.

Step 4 — Load plot-templates.md. Load references/plot-templates.md and retrieve the template matching figure_type. Use it as the structural foundation for Step 6. For custom: ask "Please describe your figure in 1–2 sentences," then select the closest template:

• Expression/omics data → volcano or heatmap
• Group comparisons/distributions → boxplot or violin
• Time-to-event data → survival
• Odds ratios/effect sizes → forest plot
• None apply → volcano scaffold; note which geoms and axes to replace

Step 5 — Load color-palettes.md. Load references/color-palettes.md and extract hex values for the requested color_scheme. Confirm palette compatibility — journal_bw requires shape/pattern differentiation in addition to grayscale fills.

Step 6 — Generate figure code. Produce complete, runnable R or Python code from the Step 4 template and Step 5 palette. Every code block must include:

• Data format note before the code: "Your data should be formatted as: [column names and types]."
• Palette application via scale_color_manual(), scale_fill_manual(), or ggsci helpers.
• Publication settings: theme_classic(base_size = 10); axis font 10pt; title 12pt; legend 8pt; 300 dpi minimum.
• Save call: ggsave() or plt.savefig() with explicit width, height, dpi, units = "mm" per journal spec.
• Chinese font note when language: zh: include library(showtext); showtext_auto() and replace all label strings with Chinese text.

Step 7 — Offer iteration. End every response with:

"Would you like to adjust colors, resize for a specific journal format, add statistical annotations (e.g., significance brackets, p-value labels), switch to a different tool, or modify any other aspect of the figure?"

4. Output Format

Begin every response with the figure header:

[Figure: <Figure Type> | Tool: <tool> | Color: <color_scheme>]

Example: [Figure: Volcano Plot | Tool: ggplot2 | Color: Colorblind-safe]

Follow the header with:

1. Data format requirement block — a short prose or bulleted list stating exactly what columns/structure the input data must have.
2. Complete code block — language-tagged fence (```r or ```python), complete and runnable.
3. "Save as:" line — shows the ggsave() or plt.savefig() call with explicit width, height, dpi, and units = "mm" arguments.
4. Journal-specific note — only when target_journal is set. Example: "Nature figures: max 180 mm wide (double-column) or 89 mm (single-column), 300 dpi minimum, TIFF or PDF preferred."

5. context_output

Reads from upstream

Writes to output

{
  "module": "data-visualizer",
  "summary": "<e.g., 'Volcano plot code generated for human DEG data, ggplot2, colorblind-safe palette'>",
  "artifacts": ["volcano_plot.pdf"],
  "parameters": {
    "figure_type": "<volcano | heatmap | pca | umap | boxplot | violin | survival | forest | ppi_network | enrichment_bubble | custom>",
    "tool": "<ggplot2 | matplotlib | base_R | cytoscape>",
    "color_scheme": "<default | colorblind_safe | tcm_palette | journal_bw>"
  },
  "status": "success | partial | failed",
  "error_message": "<string | null>"
}

artifacts is always a list of strings, each being an expected output file name (e.g., ["heatmap.pdf"]). status = partial when data_description is insufficient to generate runnable code. status = failed if figure_type could not be determined after two attempts.

6. References

See references/ for:

• plot-templates.md — Complete ggplot2/R code templates for volcano plot, heatmap, Kaplan-Meier survival curve, forest plot, and publication size guidelines
• color-palettes.md — NPG default, colorblind-safe (Wong 2011), TCM earth tones, journal grayscale palettes with ggplot2 application code and accessibility rules