Data management and FAIR

CBIO7501 · Lesson 03

Aziz Khan

Today you will

• recall why “data available upon request” fails
• use FAIR to judge a real dataset record
• use CARE to ask who benefits from data, and who decides
• write a data availability statement
• get the LTEE data the reproducible way: recorded source, checksums, metadata, read-only

“Available upon request”

Gabelica, Bojčić and Puljak (2022) wrote to the authors of 1,792 papers whose data statement promised data on request.

• 93% did not answer or declined
• 123 (6.8%) shared the data

Data (mis)management in practice

Stage	Raw data	Metadata
Acquisition	proprietary format	lab notebook
Analysis	converted; originals lost	hard-coded in scripts
Submission	on a cluster, then a hard drive	spreadsheets by email, then a PDF
Publication	“available upon request”

Ask: which row have you lived through?

The data lifecycle

plan ─► collect ─► process ─► analyze ─► preserve ─► share ─► reuse
  ▲                                                             │
  └─────────────────────── someone else's plan ◄────────────────┘

Decisions at plan (formats, metadata, license, repository) are cheap. The same decisions at share are expensive.

A data management plan

• Check what your funder, institution and journal require (many funders, for example NIH since 2023 and Horizon Europe, ask for a plan)
• Estimate the storage you need, short and long term
• Decide the metadata others will need to understand the data
• Check legal and ethical limits if the data is sensitive
• Choose the repository where the data will end up
• Put the data in its final place at the start of the project, not the week before submission

FAIR (Wilkinson et al. 2016)

	Means	For example
Findable	persistent identifier, rich metadata	DOI, SRA accession
Accessible	standard protocol; metadata stays even if data cannot	HTTPS download; documented access procedure
Interoperable	open formats, shared vocabularies	FASTQ, VCF, TSV; ontologies
Reusable	license, provenance, description	CC0 / CC BY; README

FAIR is not “open”

• Human genomic data can identify people and relatives, even without names
• Shared through controlled access (dbGaP, EGA) under the participants’ consent
• Still FAIR: findable record, clear metadata, documented access procedure
• Usually open anyway: code, environment, metadata schema, safe summary statistics

CARE (GIDA 2019; Carroll et al. 2020)

For data from Indigenous Peoples: their people, lands, resources, knowledge.

	Means
Collective benefit	the communities the data comes from benefit from its use
Authority to control	they take part in deciding how their data is governed
Responsibility	data users are accountable to them; report back
Ethics	their rights and wellbeing come first, now and for future uses

FAIR: can the data be used? CARE: for whom, and who decides?

Be FAIR and CARE

• A dataset can be FAIR and still fail CARE
• Havasupai Tribe, Arizona: DNA given for a type 2 diabetes study, later used in other genetic studies; lawsuit 2004, settled 2010, samples returned (Garrison 2013)
• In biology: human genomes, and samples of plants, animals and microbes from Indigenous lands
• In practice: community agreement before collecting; limits on use recorded in the metadata; benefits and reporting back planned

Individual consent is not community agreement. Ask: whose data in your own project would CARE apply to?

Think, pair, share

5 minutes. Open the Zenodo record SARS-CoV-2 RNA levels in Scotland’s wastewater: https://doi.org/10.5281/zenodo.6339631

For each of F, A, I, R: one thing the record does well, one thing that could be better. Then compare with your neighbor.

Data availability statements

Weak

Data are available from the corresponding author upon reasonable request.

Strong

Raw reads: NCBI SRA, BioProject PRJNA295606 (runs SRR2589044, SRR2584863, SRR2584866). Subset used: figshare, doi:10.6084/m9.figshare.7726454 (CC0). Code and environment: GitHub (MIT), archived on Zenodo (DOI).

From principles to our project

FAIR need	What you add to ltee-project today
where did it come from?	metadata/sources.tsv: URL and provider checksum
is it the same data?	metadata/checksums.sha256
what is each file?	metadata/samples.tsv and a data dictionary
never changed by accident	raw files read-only
can others get it?	download by command, not by clicking

Checksums: a fingerprint for files

$ sha256sum data/raw/reference/ecoli_rel606.fasta.gz
070a03fc2e2813853d5327608ee3ebcb4b0b2fe7faa239169921b3362b24adfa  data/raw/...

$ printf 'x' >> data/raw/reference/ecoli_rel606.fasta.gz     # one byte more
$ sha256sum -c metadata/checksums.sha256
data/raw/reference/ecoli_rel606.fasta.gz: FAILED

Same name, almost the same size, different data. Only the checksum tells.

Quiz

1. A dataset has a DOI, an HTTPS download and no license. Which letter of FAIR suffers most?

   1. F (b) A (c) I (d) R

2. Patient FASTQ files cannot be shared. Can the study still be FAIR? (yes / no)

3. Your download’s md5 differs from the provider’s. What do you do?

   1. continue, it is probably fine (b) download again and check again (c) edit the checksum file

4. Every donor in a community signed a consent form. Is that enough for CARE? (yes / no)

Answers: 1 d (no license, no reuse); 2 yes (findable metadata, documented controlled access); 3 b; 4 no (authority to control: the community, not only each donor, takes part in deciding how the data is used).

Discuss

5 minutes, in pairs.

• Does your group have a data management plan? Who wrote it, and who reads it?
• Which repositories does your field use (SRA and ENA, GEO, PRIDE, Zenodo…)? Have you ever submitted to one?
• Where is the raw data of your last project right now?

Recap

• “On request” rarely works; deposit data with an identifier and a license
• FAIR ≠ open; sensitive data can still be findable and described
• Be FAIR and CARE: ask who benefits and who decides
• Record the source before you download; download by command
• Verify with the provider’s checksum; record your own; make raw data read-only

Now: the hands-on in tutorial 03.

Sources

• Tools for Reproducible Research, NBIS. CC BY 4.0. https://nbisweden.github.io/workshop-reproducible-research/
• FAIR in (biological) practice, The Carpentries Incubator. CC BY 4.0. https://carpentries-incubator.github.io/fair-bio-practice/
• Wilkinson MD et al. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Sci Data 3:160018. https://doi.org/10.1038/sdata.2016.18
• Gabelica M, Bojčić R, Puljak L (2022). Many researchers were not compliant with their published data sharing statement. J Clin Epidemiol 150:33–41. https://doi.org/10.1016/j.jclinepi.2022.05.019
• Scorza L, Baby SV, Zieliński T. SARS-CoV-2 RNA levels in Scotland’s wastewater. Zenodo, CC BY 4.0. https://doi.org/10.5281/zenodo.6339631
• Global Indigenous Data Alliance. CARE Principles for Indigenous Data Governance. https://www.gida-global.org/careprinciples
• Carroll SR et al. (2020). The CARE Principles for Indigenous Data Governance. Data Sci J 19:43. https://doi.org/10.5334/dsj-2020-043
• Garrison NA (2013). Genomic justice for Native Americans: impact of the Havasupai case on genetic research. Sci Technol Human Values 38:201–223. https://doi.org/10.1177/0162243912470009
• FAIR principles figure often shown with this topic (SangyaPundir, Wikimedia Commons, linked not embedded): https://commons.wikimedia.org/wiki/File:FAIR_data_principles.jpg

Text and slides CC BY 4.0 unless marked otherwise. Full list: Sources and credits.