Python Vibe Coding Course — Part 5 (Final)

Publishing with Your Paper

Upload to Zenodo and GitHub

Illustration of a research paper and a box of code and data, linked by two rings

This is the final part of the series. In Part 4, you created requirements.txt and made your environment reproducible. As a finishing touch, this part explains how to upload your complete analysis package to Zenodo, GitHub, or similar services and publish it alongside your paper. Once it is published, anyone can access your analysis code and environment through the link in your paper and reproduce your figures and tables on the spot. More and more journals require data and code availability, including PLOS ONE, Nature-family journals, and eLife. The same policy is becoming common in agronomy journals such as Field Crops Research.

Dr. Hoo, this is the last part.
It is. Let's publish the analysis folder on Zenodo and GitHub alongside the paper. Then anyone can follow the analysis from the link in the paper.
!

Decide carefully when to make your data public

Should I publish as soon as it's ready?
Be careful there. Publishing too early risks someone else writing the paper first. The safest time is after the paper is formally published.

Before diving into the technical steps, you must carefully consider when to open your data and code to the public. Making them publicly accessible too early risks losing priority or allowing others to publish a similar analysis first.

As a rule of thumb, the safest time to publicly release data and code is after the paper has been formally published as a peer-reviewed article. Posting a preprint first does establish a timestamp, but releasing data and code at the preprint stage still carries risks (e.g., others may publish a similar analysis during peer review, or it may become harder to control how your data is reused). Switching from Private to Public only after acceptance or publication strikes a practical balance between the researcher's interests and open-science principles.

The recommended workflow during submission and peer review is shown below.

  1. Upload the analysis package to GitHub, Zenodo, OSF, or similar services in Private (non-public) mode
  2. In the paper's Data / Code Availability Statement, state that the repository is currently private for data-security reasons and will be made public with a DOI upon publication. Editors and reviewers typically accept this if you explain it clearly
  3. If reviewers need to inspect the code, use a reviewer link (Anonymous Link / Reviewer Link) available in services such as OSF, Figshare, or Zenodo's Closed Access feature
  4. Once the paper is published, switch from Private to Public and formally issue a DOI on Zenodo

If the journal's policy requires data to be public at submission, follow the journal's instructions. Most journals allow a reviewer link during peer review and require Public access only at publication. When in doubt, confirm with the editor in advance.

01

Finalize the folder structure for publication

If you have followed the steps from Part 2 onward, your folder should already look like this. Give it a final check before publishing.

Recommended folder structure for the published package

my_first_analysis/
├── data/
│   └── sample_data.xlsx  # Raw data
├── notebooks/
│   └── analysis.ipynb    # Analysis script
├── results/
│   ├── scatter.png        # Output figure
│   └── statistics.csv    # Output data
├── requirements.txt       # Environment file
├── README.md             # Usage instructions
└── .gitignore             # Files to exclude

.venv/ is excluded by .gitignore, so it is not published.

Separating raw data, analysis scripts, and results into distinct folders makes the structure immediately clear to anyone who opens it. If the layout has drifted, you can ask an LLM to help you tidy it up.

If your raw data contains personally identifiable information (e.g., human subject data), anonymize or process it into a shareable form before publishing. Do not publish data outside the scope of your ethics approval.

02

Choosing where to publish

Several platforms exist for publishing research data and code. This course recommends the Zenodo + GitHub combination. It is also worth knowing the other options so you can respond flexibly to requirements from your institution or journal.

Zenodo

A research data archive operated by CERN (Europe). Every upload receives a DOI. zenodo.org

  • Persistent link guaranteed by a DOI
  • Easy to cite in a paper's reference list
  • Size limit of 50 GB per file and 100 GB in total
  • Suitable for any research field

GitHub

The world's largest code-sharing and version-control platform. Excellent for collaboration and ongoing updates. github.com

  • Full change history is preserved
  • Supports collaborative editing
  • No DOI by default (can be issued via Zenodo integration)
  • Best for collaborative or ongoing projects

Figshare

A research data repository closely integrated with Springer Nature journals. figshare.com

  • DOI issued
  • Supports large files (20 GB free)
  • Private reviewer links available
  • Often recommended by Nature-family journals

OSF (Open Science Framework)

Operated by the Center for Open Science (US). Well-suited for managing entire projects. osf.io

  • DOI issued
  • Supports pre-registration
  • Size limit of 5 GB per file, with no total cap
  • Widely used in psychology and the social sciences

Dataverse

An open-source platform that began at Harvard, often run by individual institutions. dataverse.org

  • DOI issued
  • Ideal when your institution runs its own Dataverse
  • Rich metadata management
  • Suited for large-scale, long-term storage

Field-specific repositories

Some fields have dedicated repositories that are strongly recommended.

Choose based on your target journal's requirements and your institution's policies. This course recommends the "Zenodo + GitHub" combination. Keep your work updated on GitHub, then link your GitHub repository to Zenodo at the time of publication to issue a DOI (described below).

03

Upload to GitHub

Uploading to GitHub uses GitHub Desktop and follows almost the same steps as in another article on this site, "Building a Research Portfolio Website with No Coding". Here is a brief overview.

  1. Create a free account at github.com
  2. Click "+" (top right) → "New repository" and create a repository (e.g., my-first-analysis, set to Public)
  3. Install GitHub Desktop from desktop.github.com
  4. In GitHub Desktop, choose "File → Add Local Repository" and select your analysis folder
  5. Click "Publish repository" to push it to GitHub

This creates a public URL at https://github.com/your-username/my-first-analysis. Anyone can visit this URL and download the folder via "Code → Download ZIP".

04

Issue a DOI on Zenodo

Isn't a GitHub link enough?
For a paper, a Zenodo DOI is better. It lasts, and it is easy to cite. Linked with GitHub, it can be issued almost automatically.

For the URL you include in your paper's Appendix, we recommend a Zenodo DOI because it is permanent and easy to cite. Linking Zenodo to GitHub lets you issue a DOI in a semi-automated way.

  1. Visit zenodo.org and sign in with your GitHub account
  2. Open the user menu (top right) and select "GitHub" to view your repository list
  3. Toggle the switch next to the target repository (my-first-analysis) to ON
  4. Return to GitHub, go to your repository page, and click "Releases" (right sidebar) → "Create a new release"
  5. Enter a tag (e.g., v1.0) and a title, then click "Publish release"
  6. Within a few minutes, Zenodo automatically creates a snapshot and issues a DOI

The issued DOI takes the form 10.5281/zenodo.XXXXXXXX and is permanently linked to this snapshot of your code and environment. As long as the paper is not retracted, the snapshot stays accessible from anywhere in the world.

05

Include the link in your paper's Appendix or Methods

In the Methods section, or in a Data Availability Statement or Code Availability Statement, include the link with wording like the example below.

Example wording for your paper

The Python code and processed data used to generate all figures and tables in this study are openly available at Zenodo (DOI: 10.5281/zenodo.XXXXXXXX) and GitHub (https://github.com/yourname/my-first-analysis). The analysis was performed using Python 3.12 with the libraries listed in requirements.txt.

Check your journal's author guidelines, as the exact format may vary. Most journals provide a dedicated section titled "Data Availability" or "Code Availability."

06

Ongoing updates after publication

It is common to refine the analysis or add data after a paper is published. The workflow for continuing to use GitHub and Zenodo together looks like this.

  1. Edit the code locally with LLM assistance in VSCode, then check that it runs
  2. Commit and push via GitHub Desktop (the latest version on GitHub is updated)
  3. When needed, publish a new release (e.g., v1.1) → Zenodo issues a new DOI for the new version

If an older DOI is already cited in your paper, that's fine. Zenodo issues both a "Concept DOI" (always pointing to the latest version) and a "Version DOI" (pointing to a specific release), so you can use whichever is appropriate.


Series Complete

Over the five parts, you have learned everything from setting up a Python environment to vibe coding with an LLM, ensuring reproducibility, and publishing your analysis alongside your paper.

From here, the only thing left is practice. Run through the process with your own research data, with ChatGPT at your side. Whenever you hit an error, the LLM will always help you find a solution. Welcome to the community of researchers who share code and build reproducible science.

→

Next steps

  • Go through the workflow in Parts 2 to 4 again with your own research data
  • If you want higher-quality code generation, try a paid plan such as ChatGPT Plus or Claude Pro
  • To learn more advanced version control, read an introductory article on Git
  • Share this series with your co-authors and make it the standard workflow for your whole lab

Questions, feedback, and suggestions are welcome through Contact.

Thank you for all five parts.
Well done. Next, try Parts 2 to 4 again with your own research data.