Mobilizing Deep Ocean Biodiversity Data

Publishing eDNA to OBIS

Steve Formel

2026-09-09

What is eDNA?

Environmental DNA (eDNA)

“Any DNA collected from an environmental sample without first isolating targeted organisms”

 — Taberlet 2012

• Environments: water, soil, sediment, air
• Free DNA, particle-bound DNA, organelles, cells, tissue
• From viruses to fish, crustaceans, and plants

🌊🌱🍃

Environmental DNA is any DNA collected from an environmental sample without first isolating target organisms (Taberlet 2012). The most common we’ll likely encounter as OBIS nodes is metabarcoding and qPCR. For this workshop we’ll focus more on metabarcoding

The eDNA Metabarcoding Workflow

Figure adapted from NatureMetrics and Gill et al. 2016 · Emilie Boulanger

---

Most common eDNA data types you may encounter:

• eDNA Metabarcoding
• Quantitative PCR (qPCR)

The metabarcoding workflow has six main stages. 1 — Sample collection. 2 — DNA extraction and PCR amplification with specific primers. 3 — High-throughput sequencing generating millions of reads simultaneously. 4 — Bioinformatic treatment: quality filtering and OTU/ASV clustering. 5 — Taxonomic assignment against reference databases. 6 — Ecological analyses producing a presence/absence or abundance matrix across samples.

What is special about eDNA data?

Enormous potential…

• Expands access to biodiversity data at scale
• Captures environmental, specimen, and bulk-sample DNA
• Raw sequences already widely shared in NCBI

Seqs vs ASVs

• Raw sequences cannot be searched by species in time and space
• No link to spatial coordinates or sampling date
• Cannot support distribution maps, trend analyses, or MPA assessments

Special considerations

• Derived information — not a direct observation; extensive lab and bioinformatic processing
• Relies on reference databases — results depend on database completeness
• Large fraction of unknown sequences — no match in any database
• The DwC DNA Extension records how work was done, enabling future re-analysis

Raw sequences deposited in NCBI lack coordinates and dates — they cannot answer where or when a species was detected. OBIS, GBIF, and ALA now accept DNA-derived occurrences linked to space and time. DNA data requires special care: it is derived, depends on reference databases, and often produces many unknown sequences. The Darwin Core DNA extension preserves methodological context for future re-analysis.

Read left to right: what eDNA makes possible, why depositing sequences alone does not deliver it, and what to keep in mind about the data itself.

Why publish eDNA data?

eDNA captures what other methods miss

• Cryptic and rare marine taxa
  
• Non-invasive, scalable sampling
  
• Community-level biodiversity signals

Shared eDNA data is powerful data

• Data contributes to global analyses
• Data has long-term value
• Your sequences become findable via OBIS sequence search tools

Benefit	Why it matters
Open DOI	Citable in publications
Long-term archive	FAIR data
Global reach	OBIS + GBIF
Sequence search	Reuse & reanalysis

Mention the OBIS sequence search prototype at sequence.obis.org — users can search for similar sequences across all OBIS datasets. This is only possible because sequences are published in standardized form.

Resources

Looking up terms

Resource	Link
DNA Extension terms	rs.gbif.org/extension/gbif/1.0/dna_derived_data_2024-07-11.xml
DwC Quick Reference	dwc.tdwg.org/terms/
OBIS Manual (DNA chapter)	manual.obis.org/dna_data.html
GBIF DNA publishing guide	doi.org/10.35035/DOC-VF1A-NR22

Training & community

Resource	Link
These slides	github.com/iobis/obis_edna_slides
OBON 2024 DNA training	github.com/iobis/obon-2024-dna-training
IOOS Bio Mobilization Workshop	ioos.github.io/bio_mobilization_workshop
GBIF-NA DNA Publishing Workshop	sunray1.github.io/2025-05-09-GBIF-NA-DNAPublishing
OBIS helpdesk	helpdesk@obis.org

The community is genuinely welcoming to beginners. The OBIS helpdesk responds quickly. The workshops listed here offer hands-on help with your own data — participants bring actual datasets and work through formatting with instructors.

Left column is what you reach for while mapping a dataset; right column is where to get help and keep learning.

The DNA publishing guide

Publishing DNA-derived data through biodiversity data platforms

Five DNA data categories

Which one fits your data?

#	Category	Example
1	DNA-derived occurrences	Metabarcoding (ASVs/OTUs assigned to taxa)
2	Enriched occurrences	Voucher specimen + barcoded
3	Targeted species detection	qPCR / ddPCR assays
4	Name references	Sequence in GenBank only
5	Metadata only	Dataset without sequences

Use the DNA Guidance decision tree to confirm your category

We won’t go into details on this slide - just reiterating that Categories 1 and 3 are by far the most common for eDNA researchers. Category 1 covers metabarcoding; Category 3 covers targeted detection methods like qPCR/ddPCR where no sequences are produced. Be aware that there are other possible types but they are outside the scope of this workshop.

One click boxes Category 1 — that is the one almost everyone in this room has. The next click moves on.

Key Metadata: eDNA Metabarcoding

From field collection to taxonomy — what to capture at each step

⛵ Field Sampling

• Date, Coordinates, Environmental measurements

🧪 Biomarker / PCR

• Primer sequence, primer reference, target gene, PCR conditions, length of product, concentration,

🧬 DNA Extraction

• Concentration, Extraction method, Extracted amount

💻 Sequences & Taxonomy

• Sequencing platform, Library layout, Public repositry ID (NCBI, ENA)
• Taxonomy, annotation confidence and reference, Pipeline, Reference library

Category 1 — DNA-derived occurrences · basisOfRecord = MaterialSample

At field sampling: map to eventID, decimalLatitude/Longitude, samplingProtocol, Env_broad_scale. At DNA extraction: Concentration and ratioOfAbsorbance fields. At PCR stage: target_gene, pcr_primer_forward/reverse, pcr_cond. At sequences/taxonomy: seq_meth, DNA_sequence, scientificName, taxonID, otu_db.

OBIS and eDNA

eDNA data in OBIS

• OBIS currently holds 128 datasets with eDNA-derived occurrences, including 44,5 million records
• About 2 million distinct sequences
• That is 2% of all datasets, but 25% of all occurrence records in OBIS

OBIS is endorsed by UNESCO’s IOC. Data published there can flow to GBIF as well, giving you double visibility with a single submission. Emphasize that OBIS is not just a repository but actively feeds into global biodiversity assessments, climate analyses, and conservation planning.

OBIS DNA Data Management Platform

Through the eDNA Expeditions Project, OBIS is developing a data management platform for DNA data including:

• Bioinformatics
• QC
• Validation
• Publication to OBIS pipeline

To be offered to OBIS nodes to support local eDNA projects! Learn more at https://ednaexpeditions.org/

eDNA data standards

Two major eDNA data standards exist

1. MIxS

Minimum Information about any (x) Sequence — for raw sequences

Used by
INSDC NCBI ENA DDBJ

2. Darwin Core (DwC)

Community vocabulary for biodiversity data — species, location, date, basis of record

Used by
OBIS GBIF ALA

Developing landscape

• Rapidly changing field
• For OBIS future directions include the DWC-DP and FAIReDNA terms

Raw sequences deposited in NCBI lack coordinates and dates — they cannot answer where or when a species was detected. OBIS, GBIF, and ALA now accept DNA-derived occurrences linked to space and time. DNA data requires special care: it is derived, depends on reference databases, and often produces many unknown sequences. The Darwin Core DNA extension preserves methodological context for future re-analysis.

INSDC is shown dashed because it is the umbrella body — NCBI, ENA and DDBJ are its three member archives, mirroring each other’s records, not four separate places to deposit.

One click highlights the Darwin Core card, which is what the rest of the session is about; the click after that moves on.

Structuring your data

Darwin Core + DNA Derived Data Extension

• Darwin Core: shared vocabulary for biodiversity data
  
• DNA Extension: adds MIxS terms for sequences, primers, methods
  
• Developed jointly by GBIF, OBIS, and the genomics community

meta.xml ← links all tables
eml.xml ← metadata (who, what, where, when, how)
occurrence.csv ← the occurrence records
dna_derived.csv ← the DNA extension

MIxS = Minimum Information about any (x) Sequence. The extension incorporates these community standards so the same fields used in INSDC/GenBank submissions can be reused here.

Two ways to structure the same dataset

Option A — Occurrence core

Option B — Event core

Key DwC fields: Occurrence core

Required

• occurrenceID — unique, stable ID
• scientificName — taxon name to lowest possible taxon rank
• basisOfRecord = “MaterialSample”
• occurrenceStatus = “present”
• eventDate — YYYY-MM-DD
• decimalLatitude, decimalLongitude

Strongly recommended

• scientificNameID (WoRMS AphiaID)
• taxonRank, kingdom, order, genus
• minimumDepthInMeters, maximumDepthInMeters
• locationID, continent
• coordinateUncertaintyInMeters, coordinatePrecision
• year, month
• license, type

Also for eDNA

• organismQuantity (read count of sequence)
• organismQuantityType = “DNA sequence reads”
• sampleSizeValue (total read count in sample)
• sampleSizeUnit = “DNA sequence reads”
• samplingProtocol (field sampling)
• associatedSequences (link to e.g. GenBank or ENA accession)

These are the fields you will use most — the full OBIS list of required and recommended terms is at manual.obis.org/checklist.html

This is the whole field list in one table. Nothing here is wrong, and for a small dataset it is the simplest thing to publish.

Note what it costs: eventDate and the coordinates are identical for every taxon detected in a sample, so they are copied down the table once per row.

The first two columns are what the OBIS checklist marks required and strongly recommended for an Occurrence core. The third holds the terms eDNA publishing adds on top, which the checklist files under the eMoF table but which the community records here.

Key DwC fields: Event core

Event table

Required

• eventID — unique, stable ID
• eventDate — YYYY-MM-DD
• decimalLatitude, decimalLongitude
• parentEventID — required if one exists

Strongly recommended

• eventType
• minimumDepthInMeters, maximumDepthInMeters
• locationID, continent
• coordinateUncertaintyInMeters, coordinatePrecision
• year, month
• license, type

Also for eDNA

• samplingProtocol (field sampling)

Occurrence extension

Required

• occurrenceID — unique, stable ID
• eventID — links the row to its event
• scientificName — taxon name to lowest possible taxon rank
• basisOfRecord = “MaterialSample”
• occurrenceStatus = “present”

Strongly recommended

• scientificNameID (WoRMS AphiaID)
• taxonRank, kingdom, order, genus
• license, type

Also for eDNA

• organismQuantity (read count of sequence)
• organismQuantityType = “DNA sequence reads”
• sampleSizeValue (total read count in sample)
• sampleSizeUnit = “DNA sequence reads”
• associatedSequences (link to e.g. GenBank or ENA accession)

These are the fields you will use most — the full OBIS list of required and recommended terms is at manual.obis.org/checklist.html

Largely the same vocabulary, arranged differently: what describes the sampling goes in the Event table once, what describes a single detection goes in the Occurrence extension, and eventID joins them.

sampleSizeValue and sampleSizeUnit stay on the occurrence side and keep meaning read counts. The volume of water sampled is not recorded here at all — the DNA extension has samp_vol_we_dna_ext for that. samplingProtocol on the Event table is the field sampling protocol, and nothing more; the laboratory protocol is sop in the DNA extension.

Key fields to include: DNA Extension

Required

• occurrenceID — links to the occurrence row

Strongly recommended

• DNA_sequence — ASV/OTU sequence
• target_gene — e.g. “COI”, “18S rRNA”
• target_subfragment — e.g. “V4”
• env_broad_scale — biome
• env_medium — what was sampled

Strongly recommended, cont.

• pcr_primer_name_forward — and reverse
• pcr_primer_name_reverse
• pcr_primer_forward — the sequence itself
• pcr_primer_reverse
• pcr_primer_reference — where the primer pair is published

Recommended

• seq_meth — e.g. “Illumina MiSeq”
• otu_class_appr — e.g. “DADA2 v1.18”
• otu_db — reference database
• sop — link to your protocol
• samp_vol_we_dna_ext — volume sampled

manual.obis.org/dna_data.html

These are the fields you will use most — the full OBIS list of required and recommended terms is at manual.obis.org/checklist.html

WoRMS = World Register of Marine Species. Every marine occurrence should have a WoRMS AphiaID as the taxonID. For freshwater / terrestrial taxa use the appropriate backbone (CoL, etc.). Unidentified sequences should still be included — they can be assigned taxon later.

Everything OBIS marks required or strongly recommended for the DNA table is on this slide. env_broad_scale and env_medium are the two that tend to get missed; they are MIxS core and nothing elsewhere in the archive carries them.

The five primer terms are one idea in five fields: the names of each primer, the sequences of each, and the citation for the pair.

samp_vol_we_dna_ext is where the volume of water sampled goes, and sop is the laboratory protocol. Neither belongs in sampleSize or samplingProtocol on the occurrence side.

The Many Levels of Pedro Peres

Peres, P.A. and Bracken-Grissom, H. (2025), Water Volume, Biological and PCR Replicates Influence the Characterization of Deep-Sea Pelagic Fish Communities. Environmental DNA, 7: e70086. https://doi.org/10.1002/edn3.70086

What they did

One CTD cast — 450 m, night

• 3 water volumes: 1 L · 5 L · 10 L
• × 5 biological replicates → 15 filters
• × 1–11 PCR replicates → 149 kept
• PCR replicates never pooled — each is its own observation

The cleanest nesting example I know, because everything came from a single CTD cast — deliberately, so there is no spatial or temporal variation confounding the replicate comparison. One set of coordinates covers all 149 samples.

parentEventID is the only new term here. It is just a pointer from a child event to its parent, the same idea as occurrenceID linking Occurrence to the DNA extension. Three levels: cast, filter, PCR reaction.

Point at the unbalanced design: 1L1 yielded only ONE usable PCR replicate and 1L3 only seven, because the 1 L samples amplified poorly; the authors reallocated to 11 replicates elsewhere. A reuser has to know that. In the spreadsheet it is invisible — you would have to count columns. As explicit events it is self-documenting.

The negative control is an event too, with no occurrences or occurrenceStatus absent. Note it is not even in the published ESV table.

One more: reads were collapsed to presence/absence (1/0), so this is occurrenceStatus, not organismQuantity.

If asked why not Occurrence core: it would still be valid Darwin Core, but you would repeat the same coordinates on every row and lose which filter and which reaction each detection came from — which is the entire point of the study.

Event Modeling

Take 10 minutes, and in groups discuss/draw how Pedro should model his data in his event table + occurrence core.

Example Data Publishing Schematic

Example Scientific Schematic

Hand-drawn data visualization of relationships, with light grey text indicating variable names provided by the data collectors, Image by Kylie Hollis, License CC0

Linking it to the project: DEEPEND. How?

EML

The link lives in the EML or (as of recently), the DwC

EML <project> — the Project section of the IPT metadata editor

EML field	Value
Project title	DEEPEND|RESTORE consortium
Award	NOAA RESTORE NA19NOS4510193
Award	NSF HRD-1547798 · HRD-2111661
Award	ONR #98288/1184488/2
Citation	doi.org/10.1002/edn3.70086

Record in the data too

• Cruise/station DP08-01Aug22-B287N
  → fieldNumber on the cast event
• R/V Point Sur → samplingProtocol
• https://dwc.tdwg.org/terms/#provenance

Why it matters

Someone can find every DEEPEND dataset, then join this eDNA with the trawl and acoustic data from the same cruise.

This is the part people skip. The occurrence records are about fish; nothing in them says DEEPEND. Project affiliation, funding awards and the paper DOI are dataset-level metadata, and the IPT has a dedicated Project section for exactly this — title, personnel, and structured award entries with funder name and award number.

The cruise/station code is the practical join key. DP08-01Aug22-B287N identifies cruise DP08, August 2022, station B287N. Put it in fieldNumber on the cast event and anyone holding DEEPEND trawl or acoustic data from that station can line the two up.

Worth saying out loud: this costs about five minutes in the IPT, and it is the difference between a dataset that is findable as part of a programme and one that is orphaned.

Optimizing your workflow

From raw outputs → long format

The transformation:

One row = one unique sequence in one sample = one occurrence record

The most common stumbling block is this “wide to long” transformation. The OBIS manual has a clear diagram. The Metabarcoding Data Toolkit (MDT) can automate this for common pipeline outputs.

The Metabarcoding Data Toolkit (MDT)

A newer tool to simplify the workflow

• Developed by GBIF specifically for metabarcoding data
  
• Accepts common pipeline outputs (QIIME2, DADA2, etc.)
  
• Handles the wide-to-long transformation
  
• Generates Darwin Core Archive + DNA Extension automatically
  
• Integrates with IPT publishing

🔗 Available at: mdt.gbif.org

Note

Still under active development — check the GBIF documentation for the latest capabilities. Currently not possible to use the event core structure

The MDT is a significant time-saver for complex datasets. It can be used alongside the IPT rather than replacing it — it handles the data formatting step and then hands off to the IPT for publishing.

Laura’s Elegant Solution

Laura’s Data

Two workbooks — one per primer

Workbook	OTUs	Samples
vert_… Vert01	460	15
teleo_… Tele01	113	13

• One station sampled over 5 dates
• 2-3 locations sampled within the station on each date
• Each sample == 120L

Each sheet is one table

In the file	Table	Becomes
otu_table	OTU_table	reads → organismQuantity
taxon	Taxonomy	scientificName, DNA_sequence
sample	Samples	sampleID → eventID
defaultValues	Study	terms applied to every record

Laura’s Elegant Solution: MDT to IPT

Two ways into the same place. Both end at the IPT, and both produce a Darwin Core Archive; the difference is who does the mapping.

Laura’s route runs the standard outputs of an eDNA pipeline through the MDT, which reads the OTU table, the taxonomy and the sample sheet and writes Darwin Core for her. The other route is the one most people picture: build the tables yourself, in the right terms, and hand them to the IPT directly. That is drawn dashed because nothing helps you along the way.

Neither is wrong. The point of the next few slides is what the MDT expects in return for doing that work.

This is the same idea as the Study table doing what defaultValues does: say a thing once instead of on every row. Nine terms here — env_medium, target_gene, both primers and their names, sop, seq_meth, otu_db — cover all 460 x 15 records.

The two-workbook split is not an accident or an oversight. The guide is explicit: different primer sets on the same samples are separate datasets, because an OTU from Vert01 and an OTU from Tele01 are not comparable units. Note the OTU numbering runs continuously across both files, so they clearly came out of one pipeline and were split afterwards.

Sample IDs become eventID. That is the hinge between this spreadsheet and everything we said about Event core.

MDT Demo

Special cases

Unknown sequences

• Many sequences in any given study will remain uncharacterized
• Populate the scientificName field with Biota incertae sedis
• Or the lowest taxonomic information if available
  • e.g. only assigned to Class level, scientificName = Teleostei, scientificNameID = urn:lsid:marinespecies.org:taxname:293496

• For Biota incertea sedis, scientificNameID should be populated with urn:lsid:marinespecies.org:taxname:12.
• Recommend that you populate verbatimIdentification with the name that was originally documented (e.g. phototrophic eukaryote).

Sequences without a WoRMS ID (AphiaID)

• WoRMS is an extensive database of linnean names of marine species, confirmed by experts in each taxonomic group
• Within eDNA datasets there are usually many taxonomic assignments (protists, prokaryotes) that cannot be found in the WoRMS database currently (i.e. no AphiaID available)
• The name might be available and mapped to GBIF, making sharing datasets between GBIF and OBIS complicated
• The submitter of the data may in these cases record the known species name, in scientificName, the ID of the taxon in another database in taxonConceptID and leave scientificNameID empty

Control samples

• Control samples are important for quality control of resulting data
• Currently, there is no feature for this implemented in the IPT
• In the future, OBIS and GBIF are working on a solution to submit them as separate files within the same dataset
• They will then not be mapped, to avoid that they end up in the GBIF or OBIS occurrence indexes, but will be available for anyone looking for that information

Metadata with no DwC terms

• A lot of information, especially detailed parameters e.g. of bioinformatic workflow, do not have controlled vocabulary fields yet
• The FAIRe project, “making eDNA data FAIR”, is in the process of developing new metadata fields to record all the relevant information of eDNA projects
• The checklist is already available online: https://fair-edna.github.io/download.html
• A lot of existing fields, a lot of new fields
• Currently, the new fields are dropped from the IPT. But in the process of adding them to DwC and MIxS.

Don’t forget…

The DNA publishing guide

Publishing DNA-derived data through biodiversity data platforms

Resources

Looking up terms

Resource	Link
DNA Extension terms	rs.gbif.org/extension/gbif/1.0/dna_derived_data_2024-07-11.xml
DwC Quick Reference	dwc.tdwg.org/terms/
OBIS Manual (DNA chapter)	manual.obis.org/dna_data.html
GBIF DNA publishing guide	doi.org/10.35035/DOC-VF1A-NR22

Training & community

Resource	Link
These slides	github.com/iobis/obis_edna_slides
OBON 2024 DNA training	github.com/iobis/obon-2024-dna-training
IOOS Bio Mobilization Workshop	ioos.github.io/bio_mobilization_workshop
GBIF-NA DNA Publishing Workshop	sunray1.github.io/2025-05-09-GBIF-NA-DNAPublishing
OBIS helpdesk	helpdesk@obis.org

The community is genuinely welcoming to beginners. The OBIS helpdesk responds quickly. The workshops listed here offer hands-on help with your own data — participants bring actual datasets and work through formatting with instructors.

Left column is what you reach for while mapping a dataset; right column is where to get help and keep learning.