Who This Is For

This article is for readers who want to see what a complete beginner-friendly ggplot workflow looks like from start to finish. Instead of learning one isolated geom, you will move from a raw table to a final publishable figure using the same ideas from the earlier articles.

What You Will Do

• Start from a small synthetic bioinformatics-style table.
• Summarize and clean the data.
• Build a rough first plot.
• Improve it with grouping, faceting, labels, and styling.
• Export a final figure and compare it to the rough version.

Before You Start

• You should be comfortable with the first seven articles in this series.
• You need ggplot2, dplyr, tidyr, ggrepel, and patchwork.
• You should understand that plot design is a sequence of decisions, not a one-line trick.

The companion script for this article is:

R draw/scripts/08-ggplot-from-zero-full-workflow.R

Step 1: Start from a Raw Table

The script creates a synthetic table with:

• sample_id
• module
• condition
• batch
• replicate
• score

This dataset is small on purpose. The goal is not biological realism at full scale. The goal is to practice the workflow with a structure that feels closer to real scientific data than the earlier public teaching datasets.

raw_bioinfo <- tidyr::expand_grid(
  sample_id = paste0("Sample_", 1:24),
  module = c("Inflammation", "Cell cycle", "Interferon"),
  condition = c("Control", "Treatment"),
  batch = c("Batch_1", "Batch_2"),
  replicate = 1:3
)

Step 2: Summarize Before Plotting

Rather than plotting every raw row immediately, the script creates a summary table first:

workflow_table <- raw_bioinfo %>%
  group_by(module, condition, batch) %>%
  summarise(
    mean_score = mean(score),
    sd_score = sd(score),
    n = n(),
    se = sd_score / sqrt(n),
    .groups = "drop"
  )

This is a very common scientific plotting pattern. The figure is usually built from an analysis-ready summary table, not always from the raw table directly.

Step 3: Build a Rough First Draft

Start with the simplest plot that answers the question at all.

plot_rough <- ggplot(workflow_table, aes(x = module, y = mean_score, fill = condition)) +
  geom_col(position = "dodge") +
  labs(title = "Rough first draft")

That is enough to test whether the plot type is reasonable.

Step 4: Improve the Plot Deliberately

Then improve it step by step:

• better labels
• explicit colors
• error bars
• faceting by batch
• cleaner theme
• better legend placement

The companion script produces this comparison:

And here is the final version on its own:

Step 5: Highlight the Most Important Result

The series ends with one more refinement: selective labeling.

Instead of labeling everything, the script labels only the highest summary values within each batch. That keeps the figure readable while still drawing the reader’s attention.

Step 6: Adapt This Workflow to Your Own Data

If you want to reuse this pattern for your own project, replace these pieces first:

• the raw input table
• the grouping variables
• the summary statistics
• the color scale
• the faceting variable

You do not need to rewrite the whole plot from zero every time. You only need to swap the parts that define your own scientific question.

```

How to Confirm It Worked

• The script creates all three figure files and both CSV files.
• You can explain why the rough plot is weaker than the final plot.
• You can point to at least three design decisions that improved the final figure.
• You can identify which columns you would replace in your own dataset.

Common Questions

Why use synthetic data here instead of a downloaded real dataset?

Because the goal is a stable, self-contained tutorial that still feels close to a real scientific workflow without adding fragile external download steps.

Why make a rough version first?

Because rough drafts help you test the question and the plot type quickly. They are part of a good workflow, not a sign of failure.

What is the main lesson of the whole series?

A good ggplot figure is built by a chain of small decisions: data, mapping, geometry, structure, style, and export.

Review Score

Score: 95/100 Verdict: This draft is ready for human review and works well as the series finale because it turns the earlier isolated lessons into one complete workflow.

```

Personnel

• ✍ Creator: Chenglin Cai
• 🤖 AI Collaboration: ChatGPT
• 🧪 Data Provider: synthetic bioinformatics-style example data
• 💻 Code Contributor: ChatGPT