Wednesday, October 7, 2026

The Complete Research Paper Folder

How to Organize a Scientific Publication from First Draft to Final Publication

A scientific paper does not begin when you upload a manuscript to a journal, and it does not end when the paper receives a DOI.

Between the first draft and the final publication there may be:

dozens of manuscript versions,
multiple sets of figures,
statistical analyses,
raw and processed datasets,
code,
plagiarism checks,
AI-use checks,
internal reviews,
preprint versions,
several journal submissions,
reviewer reports,
responses to reviewers,
revised manuscripts,
rejected versions,
accepted manuscripts,
proofs,
copyright agreements,
APC waiver applications,
repository deposits,
social-media posts,
press releases,
blog posts,
and many other documents.

If these materials are stored in an ad hoc collection of files such as

paper_final.docx
paper_final2.docx
paper_final_revised.docx
paper_FINAL.docx
paper_FINAL_really_final.docx
paper_FINAL_really_final2.docx

then the problem is not simply untidy file management.

It becomes a research reproducibility and provenance problem.

A better approach is to create a standardized Publication Master Folder for every paper.

The principle is simple:

1. The publication folder should represent the entire life of the paper

A useful top-level structure is:

PAPER_PROJECT/
│
├── 00_PROJECT_ADMIN/
├── 01_RESEARCH_DATA/
├── 02_ANALYSIS/
├── 03_CODE/
├── 04_RESULTS/
├── 05_FIGURES/
├── 06_TABLES/
├── 07_MANUSCRIPT_DEVELOPMENT/
├── 08_INTEGRITY_CHECKS/
├── 09_INTERNAL_REVIEW/
├── 10_PREPRINT/
├── 11_JOURNAL_SUBMISSIONS/
├── 12_PEER_REVIEW/
├── 13_REVISIONS/
├── 14_ACCEPTANCE/
├── 15_PRODUCTION/
├── 16_PUBLICATION/
├── 17_APC/
├── 18_DATA_CODE_REPOSITORIES/
├── 19_PUBLIC_COMMUNICATION/
├── 20_ARCHIVE/
└── README.md

This is deliberately more extensive than what a journal requires.

The journal may receive only a fraction of these files.

The rest form the permanent provenance record of the paper.

2. Start with a README file

The root directory should contain:

README.md

This is the first file someone should read when entering the project.

It should contain:

Paper title:
Short title:
Project ID:
Corresponding author:
First author:
Lab:
Institution:

Current status:
Current manuscript version:
Current journal:
Submission date:
Revision status:
Preprint DOI:
Journal DOI:

Repository:
Code repository:
Data repository:

Last updated:
Maintained by:

It should also contain a short history:

2026-01-10  Project initiated
2026-03-15  First complete manuscript
2026-04-02  Internal lab review completed
2026-04-10  bioRxiv v1 posted
2026-04-20  Submitted to Journal A
2026-06-15  Rejected
2026-06-25  Submitted to Journal B
2026-08-03  Major revision
2026-09-01  Revised manuscript submitted
2026-09-25  Accepted
2026-10-05  Published

This simple file can become extraordinarily valuable years later.

3. 00_PROJECT_ADMIN

This folder contains the administrative information about the paper.

00_PROJECT_ADMIN/
│
├── Project_Overview/
├── Author_List/
├── Author_Contributions/
├── Affiliations/
├── ORCID/
├── Funding/
├── Grant_Information/
├── Conflict_of_Interest/
├── Ethics/
├── Permissions/
├── Institutional_Approvals/
├── Data_Management/
├── Publication_Agreements/
└── Timeline/

For example:

Author_List/
├── Author_List_v01.xlsx
├── Author_List_v02.xlsx
└── Author_Order_Final.pdf

The final author order should be explicitly archived.

This avoids future confusion about whether an author was added, removed or moved during manuscript development.

4. 01_RESEARCH_DATA

This is the scientific foundation of the paper.

01_RESEARCH_DATA/
│
├── 01_RAW_DATA/
├── 02_METADATA/
├── 03_PROCESSED_DATA/
├── 04_FINAL_ANALYSIS_DATA/
├── 05_SOURCE_DATA_FOR_FIGURES/
├── 06_SOURCE_DATA_FOR_TABLES/
├── 07_EXTERNAL_DATA/
├── 08_DATABASE_DOWNLOADS/
├── 09_SEQUENCE_DATA/
├── 10_METADATA_DICTIONARIES/
└── 11_DATA_RELEASE/

The distinction between raw, processed, and analysis-ready data is extremely important.

Never overwrite raw data.

For example:

01_RAW_DATA/
02_PROCESSED_DATA/
03_FINAL_ANALYSIS_DATA/

should represent a direction of processing:

RAW → PROCESSED → ANALYSIS

not three interchangeable copies of the same files.

5. 02_ANALYSIS

This contains the actual scientific analyses.

For computational biology:

02_ANALYSIS/
│
├── 01_Data_Cleaning/
├── 02_Quality_Control/
├── 03_Sequence_Analysis/
├── 04_Alignment/
├── 05_Phylogenetics/
├── 06_Gene_Loss/
├── 07_Statistics/
├── 08_Robustness_Analysis/
├── 09_Sensitivity_Analysis/
├── 10_Alternative_Models/
└── 11_Final_Analysis/

This is where one should preserve analyses that may not ultimately appear in the paper.

That is important.

An analysis that was abandoned can explain why the final analysis looks the way it does.

6. 03_CODE

Code deserves its own dedicated directory.

03_CODE/
│
├── R/
├── Python/
├── Shell/
├── Workflow/
├── Configuration/
├── Environment/
├── Notebooks/
├── Containers/
├── Tests/
└── README.md

For example:

03_CODE/R/
├── 01_import_data.R
├── 02_clean_data.R
├── 03_statistics.R
├── 04_PCA.R
├── 05_phylogeny.R
└── 06_make_figures.R

Also preserve computational environments:

Environment/
├── environment.yml
├── requirements.txt
├── renv.lock
└── software_versions.txt

For serious computational projects, the goal should be:

7. 04_RESULTS

This folder contains the outputs of analyses.

04_RESULTS/
│
├── 01_Raw_Analysis_Output/
├── 02_Statistics/
├── 03_Model_Output/
├── 04_Phylogenetic_Trees/
├── 05_Alignments/
├── 06_Tables/
├── 07_Figure_Source_Data/
└── 08_Final_Results/

Do not mix raw computational output with polished publication figures.

8. 05_FIGURES

This deserves particularly careful organization.

05_FIGURES/
│
├── 01_Working_Figures/
├── 02_Manuscript_Figures/
├── 03_Supplementary_Figures/
├── 04_Editable_Source/
├── 05_High_Resolution/
├── 06_Web_Resolution/
├── 07_Graphical_Abstract/
└── 08_Figure_Source_Data/

For example:

Figure_01/
├── Figure_01_editable.ai
├── Figure_01_final.pdf
├── Figure_01_300dpi.tif
├── Figure_01_web.png
└── Figure_01_source_data.xlsx

The source data underlying a figure should not disappear after publication.

9. 06_TABLES

06_TABLES/
│
├── Working/
├── Main_Text/
├── Supplementary/
├── Source_Data/
└── Publication_Final/

Tables should preferably remain in editable formats.

10. 07_MANUSCRIPT_DEVELOPMENT

This is where the manuscript evolves before it is submitted to a journal.

07_MANUSCRIPT_DEVELOPMENT/
│
├── 01_Outline/
├── 02_First_Draft/
├── 03_Internal_Drafts/
├── 04_Complete_Drafts/
├── 05_Author_Revisions/
├── 06_Tracked_Changes/
├── 07_Clean_Copies/
├── 08_Final_PreSubmission/
└── 09_Archived_Drafts/

Do not rely solely on filenames such as final.docx.

Use explicit versions:

Manuscript_v01.docx
Manuscript_v02.docx
Manuscript_v03.docx

Better still, associate versions with dates:

2026-05-10_Manuscript_v03.docx

A consistent naming convention makes files easier to retrieve and interpret later.

11. 08_INTEGRITY_CHECKS

This is a particularly important folder.

08_INTEGRITY_CHECKS/
│
├── 01_Plagiarism/
├── 02_AI_Usage/
├── 03_Image_Integrity/
├── 04_Data_Integrity/
├── 05_Reference_Check/
├── 06_Fact_Check/
├── 07_Statistical_Check/
├── 08_Code_Check/
├── 09_Authorship_Check/
└── 10_Final_Integrity_Clearance/

Plagiarism

01_Plagiarism/
├── Originality_Report_v01.pdf
├── Originality_Report_v02.pdf
├── Similarity_Report.pdf
├── Exclusions_Used.txt
└── Author_Review_of_Matches.pdf

Do not merely save the similarity percentage.

Save the actual report.

Also record:

software used
date
manuscript version
similarity score
exclusions applied
interpretation of flagged sections

A 15% similarity score without context is much less informative than the underlying report.

12. AI-usage documentation

Because AI-assisted writing and analysis are increasingly part of research workflows, create:

02_AI_Usage/
│
├── AI_Usage_Log.xlsx
├── AI_Disclosure_for_Journal.txt
├── AI_Check_Report.pdf
├── AI_Generated_Text_Reviewed/
├── AI_Assisted_Code/
├── AI_Assisted_Analysis/
├── AI_Image_Use/
└── Final_AI_Compliance_Check.pdf

The AI usage log might contain:

Date	Tool	Purpose	Material	Human verification
2026-05-02	AI tool	Grammar	Introduction	Yes
2026-05-03	AI tool	Code debugging	R script	Yes
2026-05-04	AI tool	Literature organization	References	Yes

The purpose is not to create unnecessary bureaucracy.

It is to preserve an auditable record of how AI contributed to the research process.

13. Image-integrity checks

For papers containing microscopy, western blots, gels or other image-based data:

03_Image_Integrity/
├── Original_Images/
├── Processed_Images/
├── Image_Processing_Log.xlsx
├── Figure_Comparison/
├── Integrity_Check_Report.pdf
└── Final_Approval.pdf

Never let the only surviving copy of an experimental image be the version embedded in PowerPoint or Word.

14. Statistical and data-integrity checks

Create a final verification folder:

04_Data_Integrity/
├── Raw_vs_Reported_Values.xlsx
├── Figure_Source_Verification.xlsx
├── Table_Source_Verification.xlsx
└── Final_Data_Check.pdf

This can answer:

15. 09_INTERNAL_REVIEW

Before submission, have an independent internal review.

09_INTERNAL_REVIEW/
│
├── 01_Scientific_Review/
├── 02_Methodological_Review/
├── 03_Statistical_Review/
├── 04_Language_Review/
├── 05_Figure_Review/
├── 06_Data_Review/
├── 07_Senior_Author_Review/
├── 08_Final_Checklist/
└── 09_Approval/

Keep comments and responses.

For example:

Internal_Reviewer_01_comments.docx
Internal_Reviewer_01_response.docx

This creates a useful history of scientific decision-making.

16. 10_PREPRINT

If a preprint is appropriate, create a dedicated folder.

10_PREPRINT/
│
├── 00_Preprint_Decision/
├── 01_Preprint_Manuscript/
├── 02_Preprint_Figures/
├── 03_Preprint_Supplement/
├── 04_Submission_Package/
├── 05_Submission_Record/
├── 06_bioRxiv/
├── 07_arXiv/
├── 08_Preprint_Versions/
├── 09_Preprint_DOI/
└── 10_Preprint_to_Journal_Mapping/

The last folder is particularly useful.

Record:

bioRxiv v1 → Journal manuscript v1
bioRxiv v2 → Journal manuscript v3
Journal accepted manuscript → Published article

bioRxiv allows revised versions before formal acceptance, and versions retain the same basic DOI while version-specific URLs identify particular versions.

Therefore, do not treat a preprint as a disposable PDF.

It is part of the publication history.

17. Preprint file-size management

Create:

10_PREPRINT/
└── 04_Submission_Package/
    ├── Full_Resolution/
    ├── Compressed/
    ├── PDF/
    ├── Source/
    └── Upload_Archive/

For example:

Preprint_v01_full.zip
Preprint_v01_submission.zip
Preprint_v01.pdf
Preprint_v01_source.zip

Keep the exact package that was uploaded.

If figures had to be compressed to meet an upload limit, retain both:

Figure_01_original.tif
Figure_01_preprint_compressed.jpg

Do not overwrite the original.

18. 11_JOURNAL_SUBMISSIONS

This is one of the most important folders.

Never assume that there is one manuscript.

There may be five different journal-specific versions.

Use:

11_JOURNAL_SUBMISSIONS/
│
├── Journal_A/
├── Journal_B/
├── Journal_C/
└── Journal_D/

Each journal gets its own complete package.

For example:

Journal_A/
│
├── 00_Journal_Guidelines/
├── 01_Submission/
├── 02_Cover_Letter/
├── 03_Figures/
├── 04_Tables/
├── 05_Supplement/
├── 06_Declarations/
├── 07_Reviewer_Suggestions/
├── 08_Submission_Portal/
├── 09_Submission_Confirmation/
├── 10_Editorial_Correspondence/
└── 11_Submitted_Package/

This prevents a common disaster:

19. Preserve the exact submitted package

Inside every journal folder:

Submitted_Package/
├── Manuscript_SUBMITTED.pdf
├── Manuscript_SUBMITTED.docx
├── Figure_01_SUBMITTED.tif
├── Figure_02_SUBMITTED.tif
├── Supplement_SUBMITTED.pdf
├── Cover_Letter_SUBMITTED.pdf
└── Submission_Metadata.pdf

The word SUBMITTED is important.

This is the immutable historical record.

Never replace these files with later versions.

20. 12_PEER_REVIEW

Once the paper enters peer review:

12_PEER_REVIEW/
│
├── Journal_A/
│   ├── Round_1/
│   │   ├── Reviewer_1.pdf
│   │   ├── Reviewer_2.pdf
│   │   ├── Reviewer_3.pdf
│   │   ├── Editor_Letter.pdf
│   │   └── Decision_Letter.pdf
│   │
│   └── Round_2/
│
└── Journal_B/

Preserve the original reviewer reports.

21. 13_REVISIONS

This folder should capture the entire revision history.

13_REVISIONS/
│
├── Journal_A/
│   ├── Round_1_Major_Revision/
│   ├── Round_2_Minor_Revision/
│   └── Final_Acceptance/
│
└── Journal_B/

Each round should contain:

Round_1/
├── Decision_Letter.pdf
├── Reviewer_Comments/
├── Response_to_Reviewers/
├── Revised_Manuscript_Tracked.docx
├── Revised_Manuscript_Clean.docx
├── Revised_Figures/
├── Revised_Supplement/
└── Submitted_Revision_Package/

22. The response-to-reviewers deserves its own history

For example:

Response_to_Reviewers/
├── Response_v01.docx
├── Response_v02.docx
├── Response_FINAL.docx
└── Response_SUBMITTED.pdf

The final submitted response should never be overwritten.

23. 14_ACCEPTANCE

Once accepted:

14_ACCEPTANCE/
│
├── Acceptance_Letter/
├── Accepted_Manuscript/
├── Final_Figures/
├── Final_Supplement/
├── Final_Data/
├── Final_Code/
├── Copyright/
├── License/
├── Author_Agreement/
└── Publication_Charges/

This is the transition from peer-reviewed manuscript to publication production.

24. 15_PRODUCTION

The production stage often generates a completely new set of files.

15_PRODUCTION/
│
├── Typesetting/
├── Proofs/
├── Proof_Corrections/
├── Author_Corrections/
├── Production_Correspondence/
├── Final_Proof/
├── XML_or_JATS/
├── Published_PDF/
└── Version_of_Record/

Save every proof.

For example:

Proof_v01.pdf
Proof_v02.pdf
Proof_FINAL.pdf

If a production error occurs years later, these records can be extremely useful.

25. 16_PUBLICATION

Once the article is published:

16_PUBLICATION/
│
├── DOI/
├── Published_PDF/
├── HTML/
├── Supplementary/
├── Published_Figures/
├── Published_Tables/
├── Article_Metadata/
├── Citation/
├── Publisher_Page/
└── Final_Bibliographic_Record/

Record:

DOI
publication date
volume
issue
article number/pages
publisher
journal
PMID
PubMed Central ID, if applicable
Crossref information
repository links

26. 17_APC

APC management deserves a separate folder.

17_APC/
│
├── 01_APC_Eligibility/
├── 02_Waiver_Request/
├── 03_Supporting_Documents/
├── 04_Waiver_Correspondence/
├── 05_Waiver_Decision/
├── 06_Discount/
├── 07_Invoice/
├── 08_Payment/
└── 09_Final_APC_Record/

For example:

02_Waiver_Request/
├── APC_Waiver_Letter.docx
├── Institutional_Proof.pdf
├── Funding_Statement.pdf
└── Supporting_Explanation.pdf

Then:

04_Waiver_Correspondence/
├── Publisher_Request.pdf
├── Author_Response.pdf
└── Final_Decision.pdf

The financial history of the publication is therefore preserved separately from the scientific record.

27. 18_DATA_CODE_REPOSITORIES

The paper may have multiple public repositories.

18_DATA_CODE_REPOSITORIES/
│
├── GitHub/
├── GitLab/
├── Zenodo/
├── OSF/
├── Dryad/
├── Figshare/
├── GenBank/
├── SRA/
├── GEO/
├── ENA/
├── TreeBASE/
└── Other_Databases/

Maintain:

Repository_Register.xlsx

with:

Repository	Material	Accession/DOI	Version	Date
Zenodo	Code	DOI	v1	date
SRA	Raw reads	accession	—	date
GitHub	Code	URL	release 1.0	date

Repository versioning is important because later changes should not silently alter the exact computational object supporting a published result. Persistent versioning systems such as Zenodo explicitly preserve separate versions and identifiers.

28. 19_PUBLIC_COMMUNICATION

Publication should not be the end.

Create:

19_PUBLIC_COMMUNICATION/
│
├── 01_Plain_Language_Summary/
├── 02_Blog_Post/
├── 03_Press_Release/
├── 04_Graphical_Abstract/
├── 05_Social_Media/
├── 06_Lab_Website/
├── 07_Institutional_Website/
├── 08_Presentation_Slides/
├── 09_Video/
├── 10_Podcast/
├── 11_FAQ/
└── 12_Media_Enquiries/

29. Blog post folder

For example:

02_Blog_Post/
├── Blog_Draft_v01.docx
├── Blog_Draft_v02.docx
├── Blog_Final.docx
├── Images/
├── References/
├── Published/
└── URL.txt

The blog post should ideally be prepared before publication, so that it can be released soon after the paper appears.

The blog should link to:

published article
DOI
preprint
data
code
supplementary information

30. Social-media folder

05_Social_Media/
│
├── X/
├── LinkedIn/
├── Instagram/
├── ResearchGate/
├── Bluesky/
└── Institutional/

Within each:

Post_v01.txt
Post_Final.txt
Image.png

Prepare several lengths:

01_One_Line.txt
02_Short.txt
03_Medium.txt
04_Long.txt

31. 20_ARCHIVE

This is the final preservation layer.

After publication, create:

20_ARCHIVE/
│
├── 01_FINAL_MANUSCRIPT/
├── 02_FINAL_DATA/
├── 03_FINAL_CODE/
├── 04_FINAL_FIGURES/
├── 05_FINAL_TABLES/
├── 06_FINAL_SUPPLEMENT/
├── 07_SUBMISSION_HISTORY/
├── 08_REVIEW_HISTORY/
├── 09_PUBLICATION_HISTORY/
├── 10_REPOSITORY_RECORD/
├── 11_CORRESPONDENCE/
├── 12_INTEGRITY_RECORDS/
├── 13_APC_RECORDS/
└── 14_README/

The archive should contain enough information to reconstruct the publication history without requiring someone to search through an email inbox.

32. Use a publication version numbering system

A useful convention is:

v01
v02
v03
...

But version numbers should have meaning.

For example:

M01 = manuscript development version
P01 = preprint version
J1.1 = Journal 1 first submission
J1.R1 = Journal 1 revision 1
J1.R2 = Journal 1 revision 2
J2.1 = Journal 2 first submission
ACC = accepted manuscript
VOR = version of record

Then filenames become:

Manuscript_M03.docx
Manuscript_bioRxiv_v01.pdf
Manuscript_J1.1_SUBMITTED.pdf
Manuscript_J1.R1_SUBMITTED.pdf
Manuscript_J1.R2_SUBMITTED.pdf
Manuscript_J1_ACC.pdf
Manuscript_VOR.pdf

This is much safer than:

final.docx
final2.docx
final_new.docx
final_latest.docx

33. Never overwrite historical versions

This is perhaps the single most important rule.

If:

Manuscript_J1.1_SUBMITTED.pdf

was actually submitted, never modify it.

If the manuscript changes, create:

Manuscript_J1.R1_SUBMITTED.pdf

The same principle applies to:

figures
supplementary files
code
datasets
response letters
cover letters
repository releases

Versioning is not unnecessary duplication: it establishes provenance. Research-data versioning guidance emphasizes the importance of being able to identify exactly which version underlies a published result.

34. Keep a MASTER CHANGE LOG

At the root:

CHANGELOG.md

For example:

# CHANGELOG

## 2026-04-01 — v01
Initial complete manuscript.

## 2026-04-10 — v02
Added phylogenetic analysis.
Revised Figure 3.
Updated Discussion.

## 2026-04-20 — bioRxiv v1
Preprint deposited.

## 2026-05-15 — Journal A submission
Major formatting changes.
Added Supplementary Figure S4.

## 2026-07-01 — Revision 1
Addressed Reviewer 1 comments.
Reanalysed dataset.
Figure 2 replaced.

## 2026-08-15 — Accepted
Final accepted manuscript archived.

## 2026-09-01 — Published
DOI assigned.

This may ultimately be more valuable than a complicated file-management system.

35. Keep a submission register

Create:

SUBMISSION_REGISTER.xlsx

with:

Journal	Version	Submitted	Decision	Reason	Next journal
Journal A	J1.1	10 Apr	Reject	Scope	Journal B
Journal B	J2.1	15 May	Revise	Major revision	—
Journal B	J2.R1	20 Jul	Accept	—	—

This prevents a paper's history from becoming dependent on someone's memory.

36. Keep a manuscript–figure–data mapping

A particularly powerful addition is:

FIGURE_DATA_MAP.xlsx

For every figure:

Figure	Source data	Analysis script	Raw data	Final file
Fig 1	data_01.csv	script_01.R	raw_01	Fig1.tif
Fig 2	data_03.csv	script_05.R	raw_03	Fig2.tif

Do the same for tables.

This creates a direct chain:

RAW DATA
   ↓
ANALYSIS
   ↓
SOURCE DATA
   ↓
FIGURE/TABLE
   ↓
MANUSCRIPT
   ↓
PUBLISHED ARTICLE

That is the essence of reproducible publication.

37. Keep correspondence

Create:

CORRESPONDENCE/
│
├── Authors/
├── Journal/
├── Editor/
├── Reviewers/
├── Publisher/
├── Repository/
├── Institution/
└── Funding_APC/

Important email correspondence should be exported or otherwise archived where institutional policy permits.

Do not rely on a personal inbox as the only record.

38. The final archive should be self-contained

The ultimate goal is to create something like:

ARTICLE_ARCHIVE/
│
├── README.md
├── CHANGELOG.md
├── PUBLICATION_METADATA.txt
├── SUBMISSION_REGISTER.xlsx
├── FIGURE_DATA_MAP.xlsx
├── MANUSCRIPT/
├── DATA/
├── CODE/
├── FIGURES/
├── TABLES/
├── SUPPLEMENT/
├── INTEGRITY/
├── PREPRINT/
├── JOURNAL_SUBMISSIONS/
├── PEER_REVIEW/
├── REVISIONS/
├── ACCEPTANCE/
├── PRODUCTION/
├── APC/
├── REPOSITORIES/
├── PUBLIC_COMMUNICATION/
└── CORRESPONDENCE/

A compressed archive of the appropriate release can then be preserved as a long-term snapshot.

Repositories such as Zenodo can preserve uploaded collections and provide persistent identifiers; when a collection contains multiple files/folders, a compressed archive can also be used where appropriate.

39. But don't put everything into one ZIP file

There is an important distinction between:

Working storage

and

archival storage.

Your working directory may contain:

temporary/
scratch/
old/
test/
debug/

These should not automatically become part of the public archive.

Instead, create a deliberate release package:

RELEASE/
├── README.md
├── DATA/
├── CODE/
├── FIGURES/
├── TABLES/
├── SUPPLEMENT/
├── LICENSE
└── CITATION.cff

Only materials necessary for understanding, reproducing or reusing the published work should enter the public release.

40. A simple rule for deciding where a file belongs

Every file should answer one of these questions:

A. Is this scientific evidence?

Put it in:

DATA/
ANALYSIS/
RESULTS/

B. Is this how the evidence was generated?

Put it in:

CODE/
WORKFLOW/
ENVIRONMENT/

C. Is this a manuscript version?

Put it in:

MANUSCRIPT/
SUBMISSIONS/
REVISIONS/

D. Is this evidence of research integrity?

Put it in:

INTEGRITY/

E. Is this evidence of communication with a journal?

Put it in:

SUBMISSIONS/
PEER_REVIEW/
CORRESPONDENCE/

F. Is this evidence of publication?

Put it in:

ACCEPTANCE/
PRODUCTION/
PUBLICATION/

G. Is this about communicating the research to the public?

Put it in:

PUBLIC_COMMUNICATION/

41. The golden rules

A lab-wide publication-management system can be reduced to about fifteen rules:

1. Never call a file simply final.

2. Never overwrite a submitted version.

3. Never overwrite raw data.

4. Preserve the exact files actually submitted to every journal.

5. Keep rejected submissions.

A rejected manuscript is part of the publication history.

6. Keep reviewer reports and responses.

7. Keep plagiarism/similarity reports.

8. Keep AI-use documentation and declarations.

9. Keep original figure/source data.

10. Keep code and software environments.

11. Record repository accession numbers and DOI versions.

12. Separate journal-specific submission packages.

13. Keep APC waiver requests and decisions.

14. Prepare communication material before publication rather than after publication.

15. Create a final immutable archive once the paper is published.

42. The ultimate purpose

The purpose of such a folder structure is not bureaucracy.

It is scientific provenance.

A well-organized paper should allow someone to move backwards through the entire history:

Published article
       ↓
Version of record
       ↓
Accepted manuscript
       ↓
Final revision
       ↓
Reviewer comments
       ↓
Original submission
       ↓
Preprint
       ↓
Manuscript development
       ↓
Figures and tables
       ↓
Analysis
       ↓
Code
       ↓
Processed data
       ↓
Raw data

And it should also allow someone to move forward:

Published article
       ↓
DOI
       ↓
Data repository
       ↓
Code repository
       ↓
Preprint
       ↓
Blog post
       ↓
Institutional news
       ↓
Public communication

That means the paper becomes more than a PDF.

It becomes a traceable research object.

43. The ideal lab philosophy

A useful philosophy for a research group is:

Reproducible means another researcher can understand how the results were produced.

Auditable means there is evidence for important scientific and editorial decisions.

Traceable means every important result can be connected to its source data and analysis.

Recoverable means that years later, the lab can reconstruct what was submitted, revised, accepted and published.

This approach also aligns naturally with FAIR-oriented research-data management, which emphasizes findability, accessibility, interoperability and reusability.

44. A final recommended master structure

If I were implementing this as a standard for a computational biology laboratory, I would use:

PAPER_ID_SHORT_TITLE/
│
├── 00_ADMIN/
│   ├── Authors/
│   ├── Funding/
│   ├── Ethics/
│   ├── Permissions/
│   ├── Timeline/
│   └── README.md
│
├── 01_DATA/
│   ├── Raw/
│   ├── Metadata/
│   ├── Processed/
│   ├── Analysis_Ready/
│   └── Source_Data/
│
├── 02_ANALYSIS/
│   ├── QC/
│   ├── Primary/
│   ├── Secondary/
│   ├── Robustness/
│   └── Sensitivity/
│
├── 03_CODE/
│   ├── R/
│   ├── Python/
│   ├── Shell/
│   ├── Workflow/
│   ├── Environment/
│   └── README.md
│
├── 04_RESULTS/
│
├── 05_FIGURES/
│   ├── Working/
│   ├── Main/
│   ├── Supplementary/
│   ├── Editable/
│   └── Source_Data/
│
├── 06_TABLES/
│
├── 07_MANUSCRIPT/
│   ├── Drafts/
│   ├── Tracked_Changes/
│   ├── Clean/
│   └── PreSubmission/
│
├── 08_INTEGRITY/
│   ├── Plagiarism/
│   ├── AI_Usage/
│   ├── Image_Integrity/
│   ├── Data_Integrity/
│   ├── Statistics/
│   └── References/
│
├── 09_INTERNAL_REVIEW/
│
├── 10_PREPRINT/
│   ├── bioRxiv/
│   ├── arXiv/
│   ├── Versions/
│   ├── Submission/
│   └── DOI/
│
├── 11_JOURNAL_SUBMISSIONS/
│   ├── Journal_A/
│   ├── Journal_B/
│   ├── Journal_C/
│   └── Journal_D/
│
├── 12_PEER_REVIEW/
│
├── 13_REVISIONS/
│   ├── Round_1/
│   ├── Round_2/
│   └── Final/
│
├── 14_ACCEPTANCE/
│
├── 15_PRODUCTION/
│   ├── Proofs/
│   ├── Corrections/
│   └── Final/
│
├── 16_PUBLICATION/
│
├── 17_APC/
│   ├── Eligibility/
│   ├── Waiver/
│   ├── Supporting_Documents/
│   ├── Correspondence/
│   └── Payment/
│
├── 18_REPOSITORIES/
│   ├── Code/
│   ├── Data/
│   ├── Sequences/
│   ├── Preprint/
│   └── Metadata/
│
├── 19_PUBLIC_COMMUNICATION/
│   ├── Blog/
│   ├── Press_Release/
│   ├── Social_Media/
│   ├── Graphical_Abstract/
│   ├── Website/
│   └── Presentations/
│
├── 20_CORRESPONDENCE/
│
├── 21_FINAL_ARCHIVE/
│
├── README.md
├── CHANGELOG.md
├── SUBMISSION_REGISTER.xlsx
└── FIGURE_DATA_MAP.xlsx

The important conceptual change

The biggest change I would make compared with the way most researchers organize papers is this:

Don't create a folder called "Paper."

Create a folder called the paper's entire lifecycle.

The manuscript is only one component of that lifecycle.

The same folder should tell the story of the research from:

raw data → analysis → manuscript → integrity checks → preprint → journal submission → peer review → revision → acceptance → production → publication → repository → public communication → long-term archive.

That is what turns ordinary file storage into a research information management system.

Potatoes Do Not Domesticate Like Wheat: The Forgotten World of Tuber Crops

Most classic domestication research has focused on cereals.

That creates a hidden bias.

Wheat, barley, rice and maize reproduce mainly through seeds and preserve relatively well archaeologically.

But enormous human populations have historically relied on crops that reproduce very differently.

Potato.

Cassava.

Sweet potato.

Taro.

Yams.

These plants force us to rethink what domestication even means.

Vegetative propagation changes evolution

Many underground crops are propagated using pieces of the plant itself.

A tuber, corm, rhizome, sucker or cutting can generate a new individual.

That means successful plants can effectively be copied.

This drastically changes evolutionary dynamics.

Instead of waiting for sexual reproduction to recombine genes, cultivators can preserve a useful phenotype immediately.

Find a particularly tasty yam?

Replant part of it.

Find a cassava plant with useful characteristics?

Clone it.

Plasticity comes first

Many vegetatively propagated crops display strong phenotypic plasticity.

The same genotype can produce different forms depending on soil, shade, moisture and cultivation.

The article proposes that early cultivators may initially have manipulated these phenotypes through environmental management.

Over long periods, characteristics produced plastically could become increasingly genetically assimilated.

Domestication therefore may have begun not by selecting mutations, but by repeatedly creating environments that induced useful forms.

Domestication can reduce sexual reproduction

The evolutionary trajectory can even run opposite to cereals.

Seed crops often become tightly adapted to sexual reproduction through human-controlled sowing.

Vegetative crops may become increasingly dependent on clonal reproduction.

Some cultivated forms eventually become sterile.

Triploid cultivars provide famous examples.

Once such plants arise, humans become essential to their propagation.

Sweet potato shows that clonality is not absolute

Even clonally propagated crops can retain sexual reproduction.

Sweet potato is normally propagated vegetatively.

Yet sexually produced seedlings can occasionally appear within fields or feral populations.

Farmers may recognise useful new plants and incorporate them back into the clonal stock.

Crop evolution therefore becomes a fascinating mixture of cloning, mutation, sexual reproduction and human selection.

Why archaeology struggles with tubers

The archaeological record strongly favours hard seeds.

Tubers are soft.

They rot.

Their remains are often fragmentary.

Researchers therefore depend on approaches such as:

phytolith analysis,

starch-grain analysis,

microscopy,

microCT,

charred parenchyma analysis,

and potentially ancient DNA or molecular residue techniques.

Page 6 and page 7 of the paper emphasise how much more difficult this evidence is to interpret.

No vegetatively propagated crop yet has an archaeological domestication sequence comparable to the beautifully documented non-shattering trajectories of some cereals.

That gap is not merely inconvenient.

It means a major part of humanity's agricultural history remains partly invisible.

################################################################################

################################################################################

Tuesday, October 6, 2026

Darwin's “Unconscious Selection” Explains a Huge Part of Domestication

Charles Darwin introduced an extremely useful distinction when discussing domesticated organisms.

Selection does not need to be either completely natural or carefully planned.

Between these lies unconscious selection.

Humans repeatedly alter environments and reproductive opportunities. Organisms then evolve in response, even when humans are not deliberately selecting particular genes.

Plant domestication is full of such cases.

Different farming activities create different selective pressures

The article distinguishes several operations that matter:

tillage,

sowing,

harvesting,

and propagation.

Each changes plant fitness differently.

Harvesting may favour plants whose seeds remain attached.

Sowing may favour rapid germination.

Tillage may alter competition among seedlings.

Vegetative propagation may preserve particular clones.

These pressures accumulate.

Environmental selection

Non-shattering provides a good example.

Wild cereal plants release their seeds naturally.

A harvesting system disproportionately collects grains still attached to the plant.

Those grains become next year's seed stock.

The harvesting environment therefore creates a reproductive threshold.

Plants whose seeds remain attached gain an enormous advantage inside the agricultural system.

This is environmental selection.

Competitive selection

Seed size follows a different dynamic.

Larger seedlings may compete more successfully for soil resources and sunlight.

As the population changes, the competitive environment changes too.

Selection can therefore accelerate or proceed in episodes.

The paper suggests that archaeological patterns of grain enlargement in several species fit this type of process.

Weak selection can be powerful

Another important idea is that domestication may often have involved relatively weak selection.

At first this sounds paradoxical.

How can weak selection transform wild plants into crops?

Time.

Thousands of generations allow small fitness differences to accumulate.

Weak selection may actually have been advantageous to early human communities because extremely strong selection could reduce the available food population.

Domesticated populations were not laboratory experiments. They were dinner.

People could not discard 90% of their food supply simply because the plants lacked a desired allele.

Food security itself may therefore have constrained the speed of selection.

This helps explain why early domestication looks so different from modern crop breeding.

Modern breeders can deliberately isolate particular traits.

Early cultivators were maintaining entire food-producing populations while evolution unfolded within them.

################################################################################

################################################################################

Monday, October 5, 2026

Why Domesticated Seeds Wake Up When We Tell Them To

Wild plants live in uncertain environments.

A seed that germinates immediately after a single rainfall may encounter disaster if drought follows.

Many wild species therefore possess seed dormancy.

Seeds wait.

They may respond to temperature, day length, fire, abrasion or prolonged moisture before germinating.

Agriculture changes the calculation.

Farmers want synchronous germination

A cultivated crop works best when a large fraction of seeds germinate after sowing.

If half remain dormant until next year, the farmer loses much of the expected harvest.

Crop populations therefore often evolve reduced dormancy.

The evolutionary puzzle is how this happened.

Was reduced dormancy deliberately selected?

Some hypotheses proposed that early cultivators discovered rare non-dormant plants and deliberately propagated them.

If that were true, non-dormant forms should appear very early in archaeological cultivation.

But the evidence discussed in the paper does not support such a simple scenario.

Horsegram provides unusual evidence

One of the clearest archaeological cases comes from horsegram in southern India.

Researchers used X-ray tomography to examine ancient seed coats.

Dormancy in many legumes is associated with thick seed coats.

Through time, horsegram seed coats became progressively thinner.

Importantly, the change was gradual and stepped.

This pattern is much more compatible with prolonged evolutionary selection than with farmers finding a single miraculous non-dormant mutant and immediately propagating it.

Cultivation changes the reproductive lottery

Under repeated sowing, seeds that germinate promptly contribute disproportionately to the current harvest.

Those plants therefore have more descendants among the seeds that farmers collect and resow.

Dormant seeds may survive in the soil, but they contribute less frequently to the human-managed reproductive cycle.

Generation after generation, the balance shifts.

The result is another domestication trait that can emerge without intentional plant breeding.

The farmer changes the schedule.

Evolution gradually synchronises the plant to it.

################################################################################

################################################################################

Sunday, October 4, 2026

How Farming Accidentally Created Bigger Seeds

Why are crop seeds often larger than those of their wild relatives?

The obvious answer seems to be human preference.

People preferred bigger seeds because bigger seeds provided more food, so farmers consciously selected them.

The article argues that this explanation is often too simple.

Seed enlargement may initially have been an unintended consequence of cultivation.

A seedling arms race

Imagine a cultivated plot.

People disturb the soil and sow seeds into comparatively favourable conditions.

Some seeds germinate slightly earlier.

Some seedlings grow slightly faster.

Some seeds contain larger nutrient reserves.

Those seedlings gain early access to sunlight, water and nutrients.

Once competition begins, small differences matter.

Larger seeds often produce stronger seedlings.

Those seedlings can suppress smaller neighbours.

If farmers harvest the successful plants and sow their descendants, the cultivated population gradually shifts toward larger seeds.

Nobody needs to deliberately choose “large-seed genes.”

Agricultural ecology generates the selection pressure.

Competitive selection

The paper describes this as competitive selection.

The important feature is that fitness depends partly on what neighbouring plants are doing.

Suppose the average seed size in a population increases.

A seed size that was previously adequate may now be disadvantageous because surrounding seedlings have become stronger competitors.

The evolutionary race can therefore accelerate.

This differs from a simple constant selection pressure.

Archaeology can see the change

Seed size is particularly useful archaeologically because charred grains often preserve their dimensions.

Researchers can measure ancient seeds and compare populations through time.

The lower portion of Figure 1 in the article charts increasing grain size for numerous crops, including wheat, barley, lentils, peas, rice, millet, soybean, chickpea and horsegram.

Many show prolonged trajectories rather than instantaneous jumps.

Bigger food was not necessarily the target

This leads to a delicious evolutionary irony.

Humans ultimately benefited from larger edible seeds.

But the original selection pressure may have been less “farmers choosing bigger dinner” and more “seedlings competing inside environments created by farmers.”

Cultivation changed the competitive arena.

Evolution responded.

Humans later inherited the culinary consequences.

###################################################################################

###################################################################################

Saturday, October 3, 2026

The 3,000-Year Domestication Experiment

For much of the twentieth century, domestication was imagined as relatively rapid.

Perhaps early farmers noticed useful mutations, selected them, replanted their seeds and transformed wild cereals into crops within a few centuries.

Some models suggested even shorter timescales.

Archaeobotany has dramatically changed that picture.

For several major crops, domestication appears to have unfolded across thousands of years.

Archaeological seeds are evolutionary time capsules

Plant remains survive surprisingly well at archaeological sites.

Seeds can become charred during cooking or fires. Pieces of cereal chaff may survive. Plant impressions can remain embedded inside ancient pottery.

These fragments allow archaeobotanists to reconstruct plant populations from different periods.

The paper's Figure 1, on page 4, is especially important. It plots changes in traits such as non-shattering and seed enlargement through time across crops including wheat, barley, rice, pearl millet and sorghum.

Instead of an abrupt leap from wild to domesticated forms, the graphs show gradual evolutionary trajectories.

Barley tells the story clearly

Wild barley disperses its grain by breaking apart when mature.

Domesticated barley retains its grain.

Archaeological barley from early Holocene sites in Syria, Jordan, Israel and western Iran shows changing proportions of these forms.

Before roughly 9000 BC, domesticated-type non-shattering forms were essentially absent.

Over subsequent millennia their proportion increased.

After roughly 7000 BC, some populations contained predominantly or almost entirely domesticated-type forms.

That transition took around two thousand years or more.

Similar patterns occur in wheat.

Rice followed its own long trajectory

Rice spikelet bases from archaeological sites in the Lower Yangtze provide another sequence.

Non-shattering forms gradually increased.

The process had begun before around 6000 BC and appears to have been largely completed before 4000 BC.

Again, we are dealing with millennia rather than a few generations.

Why slow domestication matters

The timescale radically changes explanations for why agriculture began.

If wheat domestication took several thousand years, it cannot easily be explained as a direct response to one short climatic event.

Neither can it be reduced to one cultural innovation or a sudden burst of population pressure.

During three thousand years, climates changed repeatedly.

Human settlements changed.

Harvesting tools changed.

Population density changed.

Cultivation practices changed.

Trade networks changed.

Domestication therefore unfolded while both human societies and environments were moving targets.

Thousands of plant generations

For an annual cereal, three thousand years can mean roughly three thousand generations.

For people it may correspond to around 100 to 150 generations.

Imagine an experiment started by your ancestors more than a hundred human generations ago, modified continuously by changing climates and technologies, and completed by people who had no memory of why it started.

That resembles early domestication more closely than a deliberate breeding programme.

Domestication without a domestication plan

This creates one of the most fascinating ideas in the paper.

The people whose activities generated domesticated crops did not necessarily have “domestication” as their objective.

They harvested.

They sowed.

They cleared.

They moved plants.

They altered soils.

Plants possessing characteristics better suited to those environments gradually left more descendants.

Domestication could therefore emerge from repeated economic behaviour without anyone intending to redesign the species.

Evolution was doing the bookkeeping.

###################################################################################

###################################################################################

Friday, October 2, 2026

Cultivation, Agriculture and Domestication Are Not the Same Thing

Three words are often treated as synonyms in discussions of early farming:

cultivation, agriculture, and domestication.

They are not.

Understanding the difference completely changes how we interpret the origins of farming.

Cultivation is a human behaviour

Cultivation refers primarily to what people do.

Humans may clear vegetation, loosen soil, sow seeds, transplant plants, water them or protect useful species.

These behaviours are culturally transmitted.

A child can learn when to sow seeds, how deeply to plant them or which patch of soil produces the best crop without any genetic change occurring in the plant.

Crucially, people can cultivate completely wild plants.

That means cultivation can exist long before domestication.

Agriculture is an economic system

Agriculture involves cultivation becoming sufficiently regular and important that it becomes central to community subsistence.

Instead of occasionally managing plants, societies begin organising substantial amounts of labour, land and seasonal activity around food production.

Agriculture can also involve livestock, although animal husbandry is not necessary for the definition.

Agriculture is therefore principally an ecological and socioeconomic system.

Domestication is biological evolution

Domestication concerns the plant.

A domesticated plant population has evolved characteristics that adapt it to environments created or maintained by humans.

Some crops ultimately became so dependent on people that they could no longer reproduce efficiently without human intervention.

Consider wheat.

Wild wheat naturally disperses its seeds.

Its seed heads break apart when mature, allowing seeds to fall to the ground.

For a wild plant, this is excellent evolutionary engineering.

For a farmer, it is terrible.

Seeds that fall before harvest disappear into the soil.

Domesticated wheats therefore evolved seed heads that remain attached until humans harvest them.

From the plant's perspective, this sounds disastrous. It has lost part of its natural dispersal mechanism.

But inside an agricultural ecosystem it is extremely successful because humans harvest, store, transport and sow those seeds.

The plant effectively outsourced dispersal to people.

Domestication syndromes

Domesticated plants often evolve suites of correlated features known as domestication syndromes.

For seed crops, these can include:

larger seeds,

reduced dormancy,

non-shattering seed heads,

changes in plant architecture,

more synchronous maturation,

reduced physical or chemical defences,

and adaptations to human harvesting and storage.

But different kinds of crops evolve different syndromes.

The comparison in Table 1 of the paper is especially revealing.

Cereals tend toward sexual reproduction through seeds and annual growth cycles.

Vegetatively propagated crops such as potatoes, cassava, taro and yams may move in almost the opposite evolutionary direction. Human cultivation can favour perennial growth and reproduction through tubers, shoots, corms or other vegetative organs.

There is therefore no universal domestication blueprint.

The distinction solves an archaeological puzzle

Once cultivation and domestication are separated, something important becomes visible.

Humans may have cultivated plants for hundreds or thousands of years before recognisably domesticated forms became dominant.

Archaeologists therefore should not expect the first evidence of cultivation and the first evidence of domesticated morphology to occur simultaneously.

The emergence of agriculture was a process, not a switch.

And that observation leads directly to one of the biggest discoveries in modern domestication research: domestication was surprisingly slow.

###################################################################################

###################################################################################