LAB PROTOCOL — a practical guide for lab members using Google Colab (Python or R).
Version: Feb 3, 2026

This guide is written for everyone—whether you code in Python, R, or don’t code at all. Please read this carefully to avoid losing your data.

Quick warning: Colab sessions are temporary. Anything not saved to GitHub or downloaded can be lost when the session disconnects.

Part 1 — The Concept (Read This First!)

1. 1. What is a “Jupyter Notebook”?

Colab uses a format called a Jupyter Notebook.

Think of it like a digital lab notebook: you combine text (notes) and code (analysis) in one document.

Why we use it: it keeps our science transparent—anyone can read the logic and see results immediately under the code.

2. Google Colab (The “digital lab bench”)

Colab runs on a Google-hosted virtual machine (VM)—you are not using your laptop’s CPU/RAM.

Ephemeral rule: the VM is temporary. When the session ends, files created/uploaded inside Colab can be deleted.

Inactivity: if you do not run code or interact for ~90 minutes, Colab may disconnect.

Hard limit: even with activity, sessions can end after ~12 hours.

Therefore: save to GitHub or download outputs before ending a session.

Part 2 — Getting started & choosing your language

Step A. Open the lab

Open Chrome (recommended) and go to colab.research.google.com.

Sign in with your Google account.

Start a new notebook: New Notebook.

Open an existing notebook from GitHub: choose the GitHub tab, paste the repository URL, and select the file.

Step B. Python vs R (critical)

Colab defaults to Python.

If you need R: Runtime → Change runtime type → select R → Save.

Do not mix Python and R in the same notebook. Use one language per experiment.

Step C. Get your data into Colab (Files pane)

Open the left sidebar → click the folder icon to open the Files pane.

Drag-and-drop your input files (CSV, Excel, images) into the Files pane.

Wait until upload finishes (the spinner/orange indicator disappears).

You can create temporary folders during a session for organization and download them at the end.

Packages and hardware

Install missing packages with pip (Python):

!pip install -q <package_name>  # -q = quiet install

For heavy computation: Edit → Notebook settings → select GPU or TPU (if available).

Part 3 — Writing your notebook (the “Sandwich” rule)

The goal: no “walls of code”.

Use this structure for each analysis block:

Top bun (text): plain-language explanation.

Meat (code): the actual script.

Bottom bun (result): plots/tables output.

How to add a text cell

Hover between cells → click + Text → describe what you are doing.

Use headers (Markdown) like # Experiment 1: Heat Stress Analysis.

Part 4 — GitHub integration (the safety net)

One-time setup

File → Save a copy in GitHub.

Grant permissions.

Check the option “Access private repositories and organizations” if you work in private repos/orgs.

Daily routine (how to save)

File → Save a copy in GitHub.

Choose the repository.

Write a clear commit message (e.g., “Added pH data”).

Click OK.

Open later

In GitHub, open your .ipynb notebook and click Open in Colab (badge/link).

Part 5 — AI “cheat sheet” (Gemini)

Use built‑in AI to accelerate, not replace, understanding.

Generate code: use the sparkle/Generate icon or Ctrl/Cmd + I and describe the task (e.g., “Read a CSV and filter null controls”).

Fix errors: use Explain error to interpret tracebacks and suggest fixes.

Improve readability: ask it to add docstrings and comments.

Part 6 — Statistical analysis

Reproducible stats over manual Excel

Always check missing values and dtypes before testing.

Common tools:

• Descriptive: df.describe()

• Hypothesis tests: scipy.stats (t‑tests, ANOVA, non‑parametric)

• Modeling: statsmodels (linear/multivariate regression)

Verify assumptions (normality, homogeneity of variance, etc.) before reporting p‑values.

Part 7 — Figures & downloads (critical)

Publication-quality figures

Target: 600 DPI (typical journal requirement).

Python template:

import matplotlib.pyplot as plt

# --- PLOTTING CODE ---
plt.plot(x, y)
# ---------------------

# Save high‑resolution figure
plt.savefig('Final_Fig.pdf', dpi=600, bbox_inches='tight')

# Download to your computer
from google.colab import files
files.download('Final_Fig.pdf')

R template (manual download from the Files pane):

library(ggplot2)

# --- PLOTTING CODE ---

ggsave('Final_Fig.pdf', dpi=600)

Manual download backup

If the file does not pop up automatically:

Files pane → right‑click the file → Download.

Part 8 — End-of-shift checklist

Do not close the tab until all are true:

☐ Saved notebook to GitHub (File → Save a copy in GitHub).

☐ Downloaded outputs (check your computer’s Downloads folder).

☐ Notebook is readable (text cells explain the math/logic).

☐ Figures are high‑res (600 DPI).

End-of-shift checklist

• Saved notebook to GitHub (File → Save a copy in GitHub).
• Downloaded outputs (check your computer’s Downloads folder).
• Notebook is readable (text cells explain the math/logic).
• Figures are high‑res (600 DPI).