Expmeta - GA4GH Experiments Standard group

Group meetings Agenda and Minutes (2026-2027)

👉 Previous meeting notes: Discovery-Metadata_Group_Agendas_ and_Minutes 2022-2025

Overview

The Experiments Metadata Standard Group focuses on defining a minimal set of metadata properties for genomic datasets, to make it usable in a variety of contexts, particularly for discovery and analysis.

• For a quick introduction, please read our Scope Statement
• For more details about the group and its background, see the Context section at the end of this document

Meeting Protocols

• Please note that by participating in meetings, attendees agree to adhere to the GA4GH Standards of Professional Conduct
• Meetings may be recorded for note-taking purposes. Recordings will be deleted within three months of the meeting taking place.

Meeting details:

Meeting schedule: 1st Monday of each month

Time: 2pm UK time / 9am Eastern

Zoom link: https://us02web.zoom.us/j/8713805924

Meeting ID: 871 380 5924

Passcode: GA4GH

GitHub issue tracker: https://github.com/ga4gh/experiments-metadata/issues

Overview        1

Meeting Protocols        1

Meeting details:        1

2026-11-02        2

Action items        6

2026-09-14        7

2026-08-03        11

2026-07-06: revised PDAP        15

2026-06-01        19

2026-04-27        27

2026-04-15: Connect meeting session (Montreal, Canada)        31

2026-03-09: LinkML draft for ExpMeta        31

2026-02-02        33

2026-MM-DD        40

Context        42

Background        42

Problem Statement        42

Goals        42

Use Cases/Challenges        42

Group Structure & Meeting Format        43

External Groups Involved        43

Resources and Links        43

Asynchronous Platforms        43

2026-12-07

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos, Coline Thomas, Charlie Barclay

Link to meeting recording:

Meeting summary:

2026-11-02

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos, Coline Thomas, Charlie Barclay

Link to meeting recording: https://us02web.zoom.us/rec/share/xBfBV3gzSvW5d9BEqzWlJDfIY9O0CupUOhNhSugiUc76l38bjsb_A4g9ZD2Hp0Lt.yuVoHVW9B-ff9vrI 

Meeting summary:

Updates and outreach

• EGA metadata schema: Marcos Casado presented EGA's metadata schema work at the EBI section meeting; he is aiming to keep it as Beacon-compatible as possible.
• dbGaP: dbGaP is re-organising deposited datasets and exposing them through a FHIR API, currently for data users already have access to. Discovery use is a future plan, and they showed interest in Experiments Metadata.
• NCBI SRA: Peter suggested engaging Chris O'Sullivan (SRA lead, NCBI), who supports the group's experiments metadata work.
• Ancient DNA: Peter attended an ancient DNA workshop; its experimental metadata needs are close to environmental DNA / metabarcoding.
• Focus: there are many promising outreach tracks, but the group cannot pursue them all and needs to prioritise.

Reminders

• Connect, Hinxton (April 2027): session proposals are open. Peter and David will attend; Charlie hopes to. Peter wants a focused session to close out the outstanding assays' metadata.
• North American input before Connect: fewer North American members will be in Hinxton, so feedback (NCBI, Canadian groups, FDA) should be gathered beforehand. Charlie suggested using part of a PHA4GE Data Structures Working Group session, which several of them attend.
• GA4GH Structural Review: open for comment until 24 November. It may reshuffle Work Streams, Driver Projects and Communities of Interest; COIs may become strategic partnerships, which matters for the PHA4GE link. Everyone is encouraged to comment.
• Cape Town: save the date only. David noted V2 should ideally be available by then.
• Singapore follow-up: a follow-up document is linked in the notes; outcomes were not discussed in this call.

Methylation as a data type

• Illumina 5-base, PacBio and Nanopore now produce methylation calls alongside WGS, often with little change to sample prep. Kits that detect more methylation types (6-base sequencing) need different preps.
• Peter noted AI gene annotation struggles with regulatory regions, so methylation data would help.
• Likely only one or two new metadata fields are needed, so it could slot in easily (the earlier epigenetics placeholder never had a clear home).
• Whole-genome bisulfite sequencing stays relevant for legacy datasets (e.g. IHEC) still being deposited.
• ENA and EGA cannot easily tell if 5-base data has been deposited; EGA files are encrypted.
• Pan-Canadian Genome Library long-read datasets will include this kind of data.

New GitHub project board

• David walked through the board: one row per GitHub issue, sub-issues with completion rates, assignees and statuses (Backlog, Ready, In progress, Blocked, In review, Done; Done is filtered out).
• Milestones: V1 items should be off the board; V2 items are the current target; out-of-scope items (transcriptomics, arrays, methylation) sit in Backlog until V3 planning.
• Labels: dictionary, ontology, documentation, schema, publication, outreach. Outreach tickets (GSC, INSDC, ancient DNA, etc.) still need creating so threads aren't lost in old minutes.
• Target dates will be set once priorities are agreed, working back from Connect. Beatrice offered a "Connect Hinxton" label.
• Future calls will cover special topics first, then review the board.
• Access: Coline and Charlie could not see the board. David added both to the experiments-metadata repo during the call; visibility still needs fixing at org level.
• Charlie agreed to be assignee on ontology items. Her ontology PR is waiting for IDs from the other group before it can be submitted.
• David is writing a landscape markdown comparing this work to MINSEQE, the IHEC spec and the EGA spec, prompted by Singapore feedback.

Schema: LinkML vs JSON Schema

• Emilio's team, implementing RNAget in Beacon, asked for experiments metadata to support transcriptomic search, referencing a versioned schema as Beacon does with VRS. The group only has a dictionary today, not a schema.
• Ownership: Peter explained INSDC prefers to use others' standards. The pattern that works best (FAIRe, MInAS for ancient DNA) is to finish a model outside GSC, then take it to GSC.
• LinkML can be converted to JSON Schema (and JSON-LD). Peter can reuse GSC's LinkML work, Charlie can reuse PHA4GE's validation and conditional-field approach, and the group has links to Chris Mungall (LinkML author).
• Charlie recommended scripting generation of all outputs from the dictionary so they don't drift out of sync.

Capacity

• Coline has some leftover GA4GH funding and can give time. She will meet Peter at EBI to pick tasks.

Decisions

1. Experiments Metadata will be developed and owned within GA4GH. Once finalised, it goes to GSC as a joint GA4GH/GSC product, keeping GA4GH provenance.
2. The core dictionary (V1) will be formalised as a LinkML schema, including conditional-mandatory field validation, before extending further.
3. A JSON Schema will be generated and published from the LinkML for Beacon/RNAget use (JSON-LD later if needed). Output generation should be automated.
4. The GitHub project board becomes the backbone of meetings: topics first, then a board review.
5. Transcriptomics, arrays and methylation stay in Backlog for now; methylation looks lightweight (one or two fields) and may be revisited.
6. North American feedback will be gathered before Connect, so the Hinxton session can close out outstanding assays.

Action items

2026-09-14

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos, Jessica Chen, Coline Thomas, Charlie Barclay, Mousumi Ghosh,

Expmeta Checklist - v2 working document

Link to meeting recording: https://us02web.zoom.us/rec/share/WaCjIVuJVUeHUjt3h4yDQ6ibDFdVR7IQjfrVFX0Dt3d-TMUfX0Aa5AyBKhcKo253.B-pzTzJnBeGa4bDC 

Actions:

• Peter - continue building ENA experiment-metadata templates/documentation (mandatory + optional fields, example and blank XMLs) for ~6 experiment types; finish within his current sprint (~2.5 weeks left).
• Peter - fix remaining ChatGPT-generated XML errors (e.g., long-read instrument defaults) and produce separate templates for Nanopore vs. PacBio.
• Peter - share the Zenodo AI-readiness resource link and any related material with David; consider mining real submitted data (in addition to the team's own defined concepts) to check for gaps, per Charlie/Coline's suggestion.
• Charlie - follow up on getting ontology ID numbers from Peter to complete the OBO/GenEpiO PR for the instrument ontology mapping.
• David & Peter - reach out to John (HCA) to finalise and validate the single-cell experiment metadata model as the near-term priority ("lowest-hanging fruit"); Peter to produce mock-ups to make this concrete.
• Beatrice - set up a GitHub Projects board for the Experiment Metadata workstream (to-do/in-progress/done style), likely after Plenary; David to identify/contact the GA4GH GitHub org owner to help create it.
• Peter - continue liaising with Chris O'Sullivan (NCBI) to align NCBI/ENA experiment-metadata approaches; invite him to future meetings.
• Peter/Charlie - Pathoplexis and phage-community collaboration to be pursued further given growing external interest.

Meeting summary:

Two "discovery sessions" are planned where experiment metadata will feature:

• Consulting-style session with David, Jordi, and Michael (Beacon Group) - three tables where community members can bring their data-discovery use cases; the team is collecting use cases beforehand to structure discussion around common problems.
• "Genomic metadata across the lifecycle" session - covering GA4GH standards for study/dataset provenance, cohort definitions, participant/parental relationships, biological samples, experiments, and results/QC metrics, plus a segment on making datasets "AI-ready." Co-organised by David and Moni (Clin/Pheno Work Stream). Needs to be finalised/published this week.

AI-Readiness Discussion

Peter shared work from his organization (EBI/ENA) on AI readiness, hosted on Zenodo (kept updateable). This has drawn strong feedback from a UK BBSRC-funded project called AirBDS, which is doing detailed AI-readiness metrics (comparable to, but more thorough than, the earlier FAIR metrics work by Mark Wilkinson). Discussion touched on the fact there's no single definition of "AI readiness" - Moni has proposed roughly seven overlapping areas (FAIR, ethics/governance, etc.), and the group discussed mapping existing GA4GH standards against these areas to identify coverage gaps across the metadata lifecycle (dataset → cohort → participant → biological sample → experiment).

ENA Experiment Metadata Template Work (Peter's main update)

Peter is working (per his sprint) on documentation/templates to let users start submitting experiment metadata to ENA now, using the existing SRA.experiment.xsd schema (shared across NCBI, ENA, and DDBJ under INSDC). Key points:

• The schema's flexible "experiment attribute" tag/value fields let domain-specific metadata (e.g., metabarcoding target gene) be captured even though not natively modeled.
• Plan: document core/mandatory vs. optional attributes per use case, provide example and blank XML templates for each experiment type.
• He used ChatGPT to generate draft example XMLs from GA4GH's spreadsheet + the XSD as a starting point, which he is now manually checking/correcting (already fixed some metabarcoding issues; noted ChatGPT incorrectly defaulted to PacBio instead of Nanopore for long-read).
• Discussed whether standard terms should come from GA4GH/GSC rather than being defined unilaterally by repositories (ENA/NCBI), since repositories are archives, not standards bodies - Peter was firm that recommendations should originate from/be endorsed by GA4GH or GSC, not from ENA or HCA alone.
• He's aiming for ~6 experiment-type templates (metabarcoding, exome sequencing, single-cell RNA-seq, bulk RNA-seq, and separate long-read templates for Nanopore and PacBio, since they're too different to merge).
• Charlie suggested combining David's "look at what's already defined" approach with Peter's plan to also mine what's actually being submitted (Coline agreed this could reveal gaps in the existing model - useful as informal domain analysis).
• Instrument model: David raised that XMeta wants to move toward an ontology-based instrument field rather than free text; Peter said ENA can't influence that directly but confirmed change is coming - GA4GH/GSC MixS terms are now hosted in OLS (Ontology Lookup Service), recently updated by Mark Miller. Charlie is keen to get ontology term IDs into GenEpiO/OBO once she has the ID numbers from Peter.
• Standards revision cadence: major schema changes happen roughly every 5–10 years; the last big one was separating biosamples out of ENA, and the current one is separating gene annotation from sequence data (GFF3 submission expected later this year).
• Single-cell metadata is viewed as essentially "done" / aligned with HCA's model (via John from HCA), differing mainly in nomenclature (library naming/strategy), which is solvable via existing mapping tables. The group agreed single-cell is the "lowest-hanging fruit" and should be prioritized/accelerated to finalise soon.
• Positive development: Chris O'Sullivan (head of NCBI's SRA archive) has reached out to Peter and NCBI has independently started similar experiment-metadata work - seen as accelerating alignment across ENA/NCBI. He couldn't attend this meeting but plans to join future ones.
• Coline (EGA) noted EGA's own checklist/metadata work is less advanced and they're tracking this group's progress to align; submissions to EGA currently still route partly through ENA, with a separate CRG-based submission path that EGA is trying to harmonize.

Fall Cleanup / Project Tracking (David's proposal)

David raised losing track of the many in-progress/stalled workstream items over the summer and proposed adopting GitHub Projects (as his Pan-Canadian Genome Library project now does) to track tickets/issues on a board with status columns and target dates, replacing the current spreadsheet-based tracking. Beatrice noted different GA4GH sub-groups already use varied tools (GKS Workstream uses GitHub boards; Pheno Packets just started a to-do/wish-list/done board) and offered to help set up a similar board for experiment metadata (with automation), likely after Plenary. Peter and Charlie agreed a public board is valuable alongside their teams' internal boards, as long as reasonably aligned.

Pathoplexis 

Peter mentioned growing external interest in the experiment metadata work from Pathoplexis, a small Swiss non-profit focused on rapid sequence/metadata sharing for specific pathogens, with tiered public/controlled access before eventually pushing data to ENA. Charlie added that Pathoplexis's data backbone is Bage-standard, partly a response to GISAID bottlenecks, making it easy to map into their existing NCBI/ENA transformation pipelines. A colleague of Charlie's is in contact with someone at Pathoplexis; another (unnamed) North American contact from the phage community also expressed interest.

Closing

The October XMeta meeting is cancelled (Labor Day conflict); the next meeting is scheduled for 2nd November. Peter noted he and David already have separate project-management time booked to follow up offline.

2026-08-03

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos, Ruth Timme, Jessica Chen, Coline Thomas, Charlie Seto, Mousumi

Expmeta Checklist - v2 working document

Link to meeting recording: https://us02web.zoom.us/rec/share/uIRYiXGjUKfm2SY3qDCUBw6-pyM56Spc3648JLVZT_MQPvM-vA5OVk_iTAPAemz0.cfktenPSd6TozI7e 

Meeting summary:

The meeting focused primarily on a proposed standardised metadata specification for long-read sequencing, presented by Jessica Chen and Ruth Timme from the U.S. Genomics for Food and Feed Safety (GenFS) group. The group also discussed opportunities to align this work with existing GA4GH/INSDC metadata standards, gather wider community feedback, and progress related GA4GH activities.

1. Long-read sequencing metadata specification

Jessica and Ruth presented work being developed through GenFS, a U.S. interagency collaboration involving CDC, FDA, USDA and NCBI Pathogen Detection. The group initially focused on whole-genome sequencing but is also working across areas such as metagenomics and potentially AI.

The motivation for the metadata work is the rapid improvement in long-read sequencing accuracy. Public repositories currently contain both older, error-prone long-read data and newer, higher-quality data, making it important to capture metadata that allows users to distinguish between them. FASTA headers can contain useful information about how sequencing data was generated, but this information is generally lost when FASTQ data is retrieved from SRA.

The proposed specification is initially focused on WGS applications, rather than metagenomics, to keep the scope manageable. It is primarily driven by Oxford Nanopore (ONT) use cases, although the team wants to retain compatibility with PacBio where possible.

2. Proposed metadata fields and controlled vocabularies

The working spreadsheet includes proposed SRA experiment attributes, their scope (e.g. ONT, PacBio or both), definitions, guidance, data types, proposed picklists and mappings to relevant GA4GH specifications.

A key discussion was how to manage controlled vocabularies and picklists:

• The team has started developing lists of acceptable values for items such as library preparation kits.
• There was discussion about whether repositories themselves should control these lists.
• Ruth noted that, based on their experience, it is more practical for the submitting community to maintain the lists because they can be updated much more quickly.
• Their approach is to maintain controlled term lists and validation outside the repository, with submitting laboratories checking their metadata before submission.
• There was interest in making these resources available more broadly, potentially through SRA as well.

The group also discussed existing ontologies. Peter noted that relevant sequencing instrument and manufacturer terms already exist in the Genomic Epidemiology Ontology and are expected to move into OBI. These could provide useful terminology for the new specification.

3. Relationship with SRA and existing standards

The specification is still at an early/draft stage and has not yet been formally implemented in SRA. Jessica explained that the team wanted to seek feedback and identify opportunities for standardisation before approaching SRA.

There was strong interest in aligning the work with existing metadata specifications rather than creating an entirely separate standard. David suggested building on previous work around minimal metadata for single-cell sequencing and potentially incorporating the long-read-specific fields as an extension.

The team agreed that broader alignment with the sequencing community would be valuable.

4. Wider community review

David proposed involving the CGEN group in Canada, particularly the major sequencing centres at BCGSC, SickKids and McGill. The group already meets regularly to discuss sequencing standards, with long-read and particularly nanopore sequencing being an important topic.

Jessica agreed that sharing the draft with CGEN would be useful and welcomed comments.

There was also discussion about involving the infectious disease/pathogen community and the phage community. Ruth noted that she had already discussed the work with Emma Griffiths, who suggested completing the current review first and then keeping her group informed.

The emerging plan is therefore to have several groups review the draft in parallel rather than waiting for one review to finish before engaging others.

5. Agreed next steps on the metadata work

The main proposed actions were:

• Peter and David to review the spreadsheet and add comments.
• David to share the draft with the CGEN/Canadian sequencing community and explore whether the groups could meet jointly.
• Charlie and others to review the metadata, particularly in relation to SRA and existing terminology.
• Explore how the long-read-specific fields could ultimately become an extension to an existing specification, rather than a standalone standard.
• Continue investigating relevant ontologies and controlled vocabularies.
• Ruth to share relevant publications on existing custom SRA attributes/quality-control metadata with Coline.

6. GSC MIXS v7 update

Peter provided a brief update that GSC MIXS v7 has been released and will be formally announced at the upcoming GSC conference.

The new version has undergone substantial cleanup, including:

• Removing ambiguities from definitions.
• Improving examples.
• Making the specification easier to parse.
• Adding a SIP checklist from the NMDC/community working on U.S. microbial/soil microbiome data.
• Adding an ancient DNA checklist.

Peter noted that there are relatively few experiment terms in MIXS, as it remains primarily sample-focused, but the changes should make the specification substantially clearer.

7. GA4GH Product Development Approval Process

David provided an update from the Technical Alignment Subcommittee (TASC). The recent TASC workshop focused heavily on Product Development and Approval Process (PDAP) v2.0.

The revised process is intended to simplify the approval process by reducing duplicated documentation and administrative overhead, while introducing some stronger requirements around technical standards. Areas being considered include:

• Providing starter kits.
• Reference implementations.
• Avoiding overlap with other GA4GH standards.
• Having a GitHub repository.
• Consistent organisation of technical product materials.

Beatrice noted that the group has already gone through v1 of the product approval process and suggested that it would be useful to consider milestones or a roadmap for the next version of the group's work, and how the new approval process fits into it.

David confirmed that TASC is aware the group is working towards a next version and that the new process has multiple gates, including an initial assessment of whether the proposed product is relevant and fits within the GA4GH ecosystem.

8. Singapore Connect

David reported that an Experiments Metadata-specific Connect session was not submitted.

Instead, he and Moni submitted a broader session looking at the metadata workflow across the research lifecycle - from creating a study and cohorts through biosamples and experiments to completion of sequencing and generation of data.

The session would explore which GA4GH products support different stages of this workflow and would include Experiments Metadata as part of that discussion.

A separate Discovery Work Stream session was also submitted. There is discussion about potentially combining or reframing the sessions around Beacon, potentially incorporating experiments metadata into the Beacon discussion.

An important objective is to use the Singapore meeting to obtain feedback from communities in the region that may be less represented at meetings held in European and North American time zones.

(AI generated)

2026-07-06: revised PDAP

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos, Jinny Park, Angela Page, Mousumi Ghosh, Eddie Lulamba

Expmeta Checklist - v2 working document

Link to meeting recording: https://us02web.zoom.us/rec/share/20wYMNwy7VeEVg3gxFsOqhxL6ULItsd258xI3C_OL-d2ewl6ATiC6AcRH92Tqs7t.E3-akI9qVe4obsJn 

Actions:

• David: Review the proposed Metadata session for Connect and provide feedback to Beatrice this week.
• David: Socialise the idea of a metadata session (incorporating Experiment Metadata) with relevant people (e.g., from GKS and Clin/Pheno) at the upcoming TASC meeting and report back to Beatrice.
• David: Submit a comment to the PDAP document regarding the classification of extensions to existing standards (e.g. Experiment Metadata) as minor or major updates.
• David: Discuss the issue of extensions to existing standards with Sasha at the upcoming TASC meeting.
• Beatrice: Write to Mark, Susheel, or Regan (AI workstream) to inquire about opportunities for a session on AI-ready data and metadata.
• Peter: Contact experts who helped during the previous Connect session to validate the experiment type metadata spreadsheet and involve them in firming up the details.
• Peter: Update the team in September regarding the re-establishment of the INSDC working group on experiment metadata.

Meeting summary:

PDAPv2 presentation [slides]:

Jinny Park presented the new PDAP:

The new PDAP introduces a more structured, stage-based approach to product development and approval.

Key changes include:

• earlier checkpoints and gate reviews,
• a risk-tiered review process led by TASC and the foundational workstreams,
• a living product dossier maintained throughout development, and
• a new 14-day "no objection" consent model for Product Steering Committee approval.

The revised process is intended to identify issues earlier, increase stakeholder engagement, reduce late-stage objections, and ensure products demonstrate real-world use before approval.

Discussion centred on how the new process will work in practice.

Questions were raised about:

• the consent model,
• classification of risk tiers,
• treatment of existing products and retirement, and
• how major versus minor updates should be handled.

In particular, the Experiments Metadata team highlighted uncertainty around whether significant but non-breaking extensions should be treated as minor updates or new products, with Angela encouraging formal feedback on this point.

The presenters also clarified that major, backwards-incompatible changes would require re-entering the PDAP from Stage 0, while minor and patch updates would follow a lighter process.

Product teams were encouraged to review the draft PDAP, provide feedback, and begin considering how the new requirements, particularly evidence of use in practice, will affect future product development.

Connect discussion:

Beatrice noted that the Connect submission deadline is next week. She highlighted that many registrants are unfamiliar to the team, so the proposed session should be designed to appeal to a broader audience. David suggested submitting a session focused on Experiments Metadata in the context of discovery and Beacon, rather than proposing another hackathon. He felt that Connect would be a good opportunity to showcase how Experiments Metadata connects with other GA4GH standards and work streams.

The group discussed a broader metadata session that would present the GA4GH metadata landscape, covering the connections between dataset and study descriptors, cohorts, participants, samples, experiments, quality control, and provenance. David explained that the goal would be to “connect the dots” and demonstrate that although standards are developed across different work streams, they contribute to a unified metadata story. Peter agreed that this approach would be valuable, particularly because Connect will bring together people from different communities, but he emphasized the importance of defining a clear outcome for the session rather than having it remain a general discussion.

Peter suggested that the session could also address the growing focus on AI-ready data. He noted that AI readiness is closely linked to FAIR data principles, structured provenance, ontologies, and semantics, and that GA4GH metadata efforts already support many of these goals. The discussion focused on using the session to explore whether existing metadata approaches meet the needs of different communities and whether additional connections or metadata elements are needed to support interpretation and reuse of data.

The group agreed that experiment metadata would likely fit best as part of a wider metadata session rather than as a standalone session. A dedicated experiment metadata discussion may only be needed if there is interest from groups working on additional experiment types. David suggested involving contributors from other areas, particularly Clin/Pheno and Genomic Knowledge Standards (GKS), to ensure broader representation and alignment.

The next steps are for David to review the existing session proposal and determine how experiment metadata can be incorporated. He will also discuss the idea with relevant contributors at the upcoming meeting to identify potential speakers or collaborators.

Beatrice will check with AI workstream leads to understand whether there are related activities that could align with the proposal. The team will then refine the session concept and decide on the best framing before submitting it to Connect.

Peter provided an update on changes he had made to the experiment metadata spreadsheet. He added target enrichment method and protocol as additional fields, following previous discussions about splitting the library selection method into more detailed components. He also added examples and noted that he had not yet extended the changes to long-read sequencing. The group discussed whether the protocol field should be mandatory or recommended, with the general view that the method should be mandatory while the protocol would likely be better as recommended.

Peter explained that some metadata fields are conditional depending on the experiment type. For example, target gene information would apply to metabarcoding experiments but not to exome sequencing, where exome capture is the relevant concept. He also presented a draft comparison of experiment workflows, created to identify common metadata elements across different experiment types. The comparison used colour coding to distinguish metadata that is broadly applicable, commonly used, or specific to certain experiment types. Peter clarified that this was not intended as a final artifact but as a way to check that important metadata requirements had not been missed and to prepare for future expansion to additional experiment categories.

David confirmed that the purpose of the workflow comparison was to ensure that all necessary metadata properties and process steps had been captured. Peter noted that some steps shown in the workflow, such as analysis steps, were included only for context and would not necessarily be part of the experiment metadata scope. The group agreed that the draft would be useful for validation with domain experts, who could confirm whether the workflows and metadata requirements were accurate. Peter suggested contacting experts and asking them to review and refine the shared Google Sheet directly, as an iterative process involving feedback from different communities.

Peter then shared positive news from the INSDC community. The minimal metadata specifications for experiment objects have now been formally published, which is an important step because previously much of this information existed informally in discussions or code rather than as documented requirements. This work was partly driven by INSDC’s aim to make it easier for additional centres, such as those in South Korea, India, and China, to participate. Peter had presented some of the ongoing experiment metadata work during these discussions, and the NCBI team showed interest in continuing collaboration to improve Experiments Metadata practices.

Peter clarified that the published INSDC document reflects the current state of experiment metadata within INSDC rather than the GA4GH product being developed by the group. However, the intention is to move forward and align future work. An INSDC experiment metadata working group is expected to be re-established in September, led by Yushi from DNA Data Japan, providing an opportunity to continue discussions and potentially bring the GA4GH work into broader adoption. Peter also noted that there is growing acceptance of moving toward the use of public ontologies, which represents further progress toward more structured and interoperable metadata.

The meeting concluded with confirmation that Peter would provide further updates once the INSDC working group resumes in September.

Beatrice noted that Ruth Timme will present at the next meeting, likely on FDA-related topics, and thanked everyone for their participation.

2026-06-01

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos, Eddie Lulamba, Coline Thomas, Ruth Timme

Expmeta Checklist - v2 working document

Link to meeting recording: https://us02web.zoom.us/rec/share/HFzLbp6dpiB45eMG-wqSj4Te3G8s5avyR1G4iUrIDTWii-rQmPhRpRPaIUakJ1UQ.fIS_N57QCOSrbw5w 

Pre-meeting notes:

experiment types - update

Targeted_sequencing

1. library_selection

             During the Connect meeting in Montreal, the feedback was that was confusing with some of the examples overlapping with experiment assay.

            On exploration, I could not see any clean implementation in an ontology.

            What about splitting this into two?  target_enrichment_method and target_enrichment_protocol  And try and get ontology terms created if needed for the high level   target_enrichment_method?     (will that keep it cleaner to query and offer flexibility?  It would be less aligned with the current INSDC…, but we are trying to improve discoverability etc. )          

target_enrichment_method

Definition

The general strategy used during library preparation to increase the representation of the target molecules or genomic regions prior to sequencing.

Examples

• amplification
• hybridisation capture
• transcript enrichment
• restriction-based enrichment
• affinity enrichment
• chromatin accessibility enrichment
• none
• other

Comment

This field is intended to support consistent searching, filtering, and comparison across experiments. More specific details of the enrichment procedure should be provided in target_enrichment_protocol.

target_enrichment_protocol

Definition

The specific protocol or laboratory procedure used to perform target enrichment during library preparation.

Examples

• COI PCR amplification using Leray primers
• MiFish multiplex PCR amplification
• 1240K ancient DNA hybridisation capture
• Agilent SureSelect Human All Exon capture
• Poly(A) RNA selection
• Ribo-Zero rRNA depletion
• RADseq using SbfI restriction digest
• ChIP enrichment using anti-H3K27ac antibody

Here is a mockup table of how it could look. (there is still sometime overlap with the experiment_assay)

2) library_capture_details

Def: High-level information on the probes etc.  including the kit used to capture the exome DNA

Would we still need this? Or would  target_enrichment_protocol cover this?

—---------------------------

Meeting summary:

The focus of the ExpMeta group remains on extending the checklist beyond the core V1 spec, released in November 2025, to cover single-cell, targeted sequencing, and long-read sequencing experiments.

At the Connect breakout sessions: 

• The long-read sequencing discussion highlighted the large number of potential metadata properties, making it difficult to review more than a small portion of the existing list.
• The single-cell sequencing category is more concise and builds heavily on work from the Human Cell Atlas, with efforts focused on identifying metadata elements relevant to the ExpMeta dictionary.
• For targeted sequencing, participants agreed that the key metadata properties had been captured, including requirements for environmental and exome sequencing use cases, although some cleanup and refinement work remains.

GA4GH Connect meeting in Singapore:

Beatrice encouraged the group to consider submitting session proposals before the July 15th deadline. David raised concerns about the value of a traditional ExpMeta working session given limited attendance from regular contributors, suggesting that a more introductory and community-focused format might be appropriate.

Drawing on experiences from the Australia Connect meeting, Beatrice noted that sessions aimed at introducing standards and gathering perspectives from new participants had been successful, even if they did not directly advance specification development.

The group explored ideas such as presenting ExpMeta alongside related standards, creating sessions around organising cohort, biosample, experiment, and results data, or grouping several metadata-oriented standards together to tell a broader data-management story.

Additional suggestions included using Connect sessions to work through open GitHub issues in breakout groups and considering the development of a manuscript describing the work, although participants acknowledged the challenge of finding time and resources.

The group will continue considering session formats that both raise awareness of ExpMeta and gather feedback from a wider community while advancing outstanding work on the specification.

Progress on ontology work for sequencing platforms and instruments:

Peter reported that he had compared the latest INSDC configuration files against previous versions and identified a small number of new sequencing-related terms, including one new platform and several new instruments. These terms are being evaluated by Charlie against the Gene Ontology Epidemiology (GenEpiO) ontology, where many already exist and the remaining terms will be added. Charlie is also pursuing incorporation of these terms into OBI, with the longer-term goal of ensuring that sequencing platforms and instruments are represented consistently across relevant ontologies.

Coline welcomed the work and suggested that consideration should also be given to other laboratory equipment such as microarray scanners, protein analysis instruments, and PCR devices. While the current effort is focused specifically on sequencing instruments, participants agreed that a broader ontology structure could eventually accommodate other categories of laboratory devices.

Discussion highlighted the practical challenges of ontology maintenance, noting that progress can be slow because many ontology projects rely on limited volunteer effort and few dedicated resources.

David noted that an upcoming GA4GH TASC meeting would focus on improving internal coordination of standards and recommended ontologies for specific data types. Establishing OBI as the preferred ontology for sequencing instruments could provide a cleaner and more consistent solution across GA4GH products, encouraging other standards and services to adopt the same terminology.

Peter added that having instrument terms established in OBI would also strengthen discussions with INSDC partners and improve interoperability across the broader genomics ecosystem.

Version 2 of the Experiments Metadata Checklist: 

David explained that the current work is based on analysis of the Human Cell Atlas Tier 1 checklist, from which experiment-specific metadata properties were extracted while deliberately excluding biosample-related information that falls outside ExpMeta’s scope.

During Connect discussions, participants observed that many metadata properties identified as important for single-cell experiments are not unique to that domain and can also apply to bulk sequencing, targeted sequencing, and other experiment types.

This raised questions about how best to organise the next version of the checklist. Rather than creating isolated metadata lists for each experiment category, the group recognised the need for a more flexible structure capable of representing properties that span multiple experiment types.

Peter noted that the challenge arises because the spreadsheet-based approach is fundamentally two-dimensional, while the underlying metadata model is inherently more complex. He suggested that the current focus on identifying concepts first is the right approach, with questions of organisation and computable representation to follow later.

Coline proposed a more modular framework based on experimental workflow stages rather than broad experiment categories. In this model, metadata properties would be associated with specific protocol components, such as sample extraction, cell selection, labelling, library preparation, enrichment, or sequencing itself, which could then be reused across many experiment types. For example, properties associated with a sequencing step would naturally apply to any sequencing-based experiment, while properties related to targeted enrichment could be shared between exome sequencing and other targeted assays.

The group saw merit in this modular approach because it could better capture the reuse of experimental procedures across different technologies. Coline suggested that properties could initially be tagged according to the workflow stage they relate to, allowing the team to explore whether such categorisation proves useful without immediately committing to a highly detailed computable model.

At the same time, David expressed caution about increasing complexity. He reflected that one of the strengths of V1 of the Experiments Metadata Checklist was its simplicity, achieved despite years of effort to reach consensus on a concise set of core properties. Introducing detailed protocol subdivisions risks making the model more difficult to understand and adopt, creating a trade-off between precision and usability. The group generally agreed that lightweight tagging of metadata properties by protocol stage could be a useful intermediate step while preserving the overall simplicity of the checklist.

Peter also observed that a more structured, workflow-oriented representation of metadata could have benefits beyond checklist design. In particular, it could support emerging efforts to automatically extract experimental information from scientific publications, as protocol-based organisation would provide useful guidance for identifying and categorising metadata elements during text mining. The discussion concluded with agreement that these ideas warranted further exploration and should continue to inform the development of the next phase of the ExpMeta metadata model.

The discussion continued with consideration of additional identifiers that may be needed in the Experiments Metadata Checklist. David highlighted feedback from the Human Cell Atlas community suggesting the inclusion of a library sequencing run identifier. While the current model captures experiment-level identifiers, it does not explicitly represent individual sequencing runs associated with a library. Participants agreed that such identifiers could be valuable, particularly for downstream analysis and investigation of batch effects.

Peter noted that while this information may not be essential for experiment discovery, it is highly relevant for data analysis and interpretation. The group therefore considered adding sequencing run identifiers as optional rather than mandatory metadata elements. Related attributes such as sequencing dates were also mentioned as potentially useful contextual information.

The group then discussed ontology support for several single-cell metadata fields. David raised the example of “suspension type,” a controlled vocabulary used in Human Cell Atlas metadata, and asked whether a suitable ontology already exists.

Participants suggested exploring existing biological ontologies, including the Gene Ontology and resources already used by the Human Cell Atlas, to determine whether appropriate terms are available before creating new controlled vocabularies. The conversation reinforced the broader principle of reusing established ontologies wherever possible rather than maintaining project-specific term lists.

How experiment-type-specific metadata could be represented in a future computable model:

David revisited the concept of conditional requirements, where selecting a particular experiment type, such as single-cell sequencing, would automatically trigger additional metadata fields relevant to that category. This would allow the core checklist to remain concise while ensuring that specialised metadata is collected when appropriate.

Peter noted that this idea had previously been discussed under the concept of “conditional mandatory” fields and suggested that LinkML should be capable of supporting such logic.

The team agreed that this would likely be an important feature of future schema development.

The group also examined a Human Cell Atlas field describing cell enrichment.

David explained that the current implementation combines terms from the Cell Ontology (CL) with additional notation indicating positive or negative selection, effectively merging ontology terms and qualifiers into a single field.

Participants questioned whether this was the best design. Peter and David considered whether the concept should instead be separated into distinct components, while Coline proposed a cleaner structure with separate fields for cell enrichment and cell depletion. Additional fields could then specify which cell populations were selected or removed.

The group generally preferred this more structured approach, recognising that just because another implementation combines concepts into a single field does not necessarily mean ExpMeta should do the same.

David agreed to discuss the issue further with Human Cell Atlas collaborators to better understand their rationale and assess possible alternatives.

Later in the meeting, Peter returned to the targeted sequencing metadata category and discussed feedback received during the Connect workshop. Participants had noted that the current “library selection” field mixes multiple concepts and could be difficult to search or classify consistently. Peter proposed splitting this into two complementary fields: a high-level, ontology-driven classification describing the general library selection method, and a second field providing more detailed free-text information. This would allow users to search and group experiments according to broad methodological categories while still preserving important technical details that evolve too rapidly for ontology maintenance.

Coline supported the approach, noting that submissions often include both general enrichment strategies and highly specific commercial kits, making a combination of controlled vocabulary and descriptive detail particularly useful. The group also observed that “library capture” and “library selection” appeared to overlap conceptually and may need to be consolidated or better differentiated in future revisions.

The discussion then turned to long-read sequencing metadata. Ruth reported that her team has been developing a specification for long-read sequencing in pathogen surveillance and has been using the ExpMeta draft specification as a starting point. The team has already identified additional attributes needed for their use cases while finding that some fields currently proposed by ExpMeta may not be necessary for their workflows. Ruth expressed interest in presenting their work to the group to ensure alignment and avoid developing incompatible standards. David welcomed the offer and suggested scheduling a future meeting dedicated to reviewing the long-read metadata work. Participants agreed that collaboration would be valuable, particularly since both efforts are progressing on similar timelines.

Looking ahead, the group discussed completion targets for Version 2 of the Experiments Metadata Checklist. While no formal deadline has been set, David and Peter suggested that having the three current focus areas, single-cell, targeted sequencing, and long-read sequencing, substantially completed by the end of the year would be a reasonable goal. This timeline would also support a potential manuscript describing the work. Participants acknowledged that metadata standards are never truly finished, but agreed that clear versioned releases are necessary to enable adoption and publication.

Before closing, Peter noted that the upcoming annual INSDC meeting would include an overview of the Experiments Metadata initiative, providing an opportunity to increase awareness and encourage participation from INSDC partners. He also reiterated a longer-term goal of moving controlled vocabulary content into established ontologies wherever possible. By relying on community-maintained ontologies rather than maintaining project-specific term lists within the schema, ExpMeta could reduce maintenance burden, improve interoperability, and ensure broader reuse of its metadata concepts.

(Aided by Zoom AI)

2026-04-27

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos, Mousumi Ghosh, Charlie Barclay, Heather Creasy, Tony Brookes, Ruth Timme

Link to meeting recording: https://us02web.zoom.us/rec/share/YA-Ua_qr2PZv6QXrD0OM2QWDUAl-nHbKJwaWeOu8AEtISpdpKsQa0BrNwJ6Lokk.Witn3B-BCX_8v3Ze 

Pre-meeting notes:

1. Welcome and Introduction

2. Review of previous action items

• Most of these have been done.
• Highlight: as a result of Charlie’s work the instruments models are now in GenEpiO

https://www.ebi.ac.uk/ols4/ontologies/genepio/classes/http%253A%252F%252Fpurl.obolibrary.org%252Fobo%252FGENEPIO_0001921?lang=en

        Peter - provided the new sequence instruments to Charlie

• Several outstanding.

3. Recap - Connect session in April

• Slides: Experiments Metadata
• Minutes: Experiments Metadata working session
• See Expmeta Checklist - v2 working document particularly “Targeted Sequencing”, “Long Reads”) and “Single Cell”
• Keeping the ball rolling
• Question: is it v2.0 or v1.1 ? as it is non-breaking? - v2.0 agreed
• Beacon API - ideally wants discoverable metadata- that may be different to machine/human readable

AoB:

• Peter has updated the INSDC experiment group. (there may well be feedback in June from the INSDC annual meeting). At the end of 2025 there was a minimal specification published, it is very minimal: https://www.insdc.org/insdc-minimal-specifications/ At ENA aiming to go over and above this. Helps if it goes into GSC.
• GDI adoption? https://gdi.onemilliongenomes.eu/ - Action Tony/David - ask Jordi
• Ruth Timme: I'll run the long-read spec by our US working group looking to implement ONT sequencing for foodborne pathogen surveillance.

Meeting summary:

Connect recap:

At Connect, the group reviewed progress on the Experiments Metadata core checklist released in Nov last year, which defines properties for general sequencing experiments. The discussion focused on the next phase of work: extending this framework to specific experiment categories. Three priority areas were selected based on available expertise:

• long-read sequencing,
• targeted sequencing, and
• single-cell sequencing.

During the session, participants split into groups to work on each category using pre-prepared datasets. Data for long-read and targeted sequencing were extracted from ENA, while SC sequencing properties were drawn from the Human Cell Atlas. The aim was to compare and refine existing property sets to further develop the metadata standard.

Tony raised questions about when Experiments Metadata is expected to be created and used, noting uncertainty over whether it should be required for all experiments or mainly for data discovery purposes. He also asked whether the three focus areas, long-read, targeted, and single-cell sequencing, are mutually exclusive, and suggested they likely overlap.

In response, David explained that the team is still exploring how to formalise the metadata model, potentially using LinkML or similar schema approaches, to support overlaps between categories. The intention is to produce a structured “dictionary” of properties across experiment types, including core fields shared by all categories and additional fields specific to each domain, with a need to represent these relationships computationally.

Tony further questioned how required vs optional fields should be handled, warning that overly strict requirements could reduce adoption, while too much optionality would harm data discovery. Peter responded that the model would likely use conditional requirements, where core fields are always mandatory and additional fields become required depending on experiment type (e.g. metabarcoding or long-read sequencing). He added that formal tooling would ideally be integrated into laboratory information management systems (LIMS) or submission workflows (such as ENA), so metadata is captured during experiment execution rather than retrospectively.

David and Peter also emphasised that modern systems increasingly automate metadata capture, but flexibility is still needed for less-resourced users. They noted that the current checklist approach has already helped define relevant fields and highlight those most important for discovery. Peter noted strong engagement in the workshop sessions, where participants made tangible progress refining and validating the metadata structure across the three groups.

How to maintain momentum after the Connect session:

David said that some groups, particularly long-read sequencing, were unable to review all proposed fields within the limited breakout time and lacked ongoing representation in regular calls. He raised the challenge of driving these category-specific efforts to a more complete state. Tony responded that such standards are inherently never “finished” and should be expected to evolve, emphasising that the core checklist is the most critical component and should be stable, widely adopted, and potentially enforced by major stakeholders like ENA, while extensions will likely remain flexible and variably adopted across communities.

Ruth, rejoining the group, highlighted parallel efforts in the US to develop long-read sequencing specifications for pathogen surveillance and stressed the importance of aligning with existing work rather than duplicating it. She suggested engaging broader communities, such as the Phage Data Structures Working Group, to validate and promote adoption. Charlie supported this, proposing stronger coordination and knowledge-sharing between groups. Peter noted ongoing communication with NCBI and other international database collaborators, mentioning that they are monitoring progress and may engage further following discussions at upcoming INSDC meetings.

The conversation also explored technical challenges in modelling long-read sequencing metadata across different technologies such as Nanopore and PacBio. David raised the issue of whether to generalise fields across platforms or create technology-specific extensions, with Ruth advocating for as much generalization as possible to avoid excessive complexity, while acknowledging some platform-specific requirements. The group agreed on the value of broader collaboration and informal coordination channels (e.g. mailing lists or GitHub) to refine these approaches.

Updates from the targeted sequencing group indicated solid progress, with contributions from domain experts helping validate that existing metadata fields are largely sufficient, particularly for use cases like exome sequencing. Prior work, including environmental DNA checklists, was leveraged to accelerate development. The group also highlighted the importance of ensuring interoperability with related frameworks such as the Ph4ge data structures, reinforcing a broader goal of alignment across standards initiatives.

SC sequencing category: David explained that the current list is a reduced subset derived from the Human Cell Atlas tier-one checklist, with non-experimental and core fields removed. While some participants noted that many of these properties are broadly applicable across experiment types, the key distinction is that, for single-cell experiments, these fields are considered mandatory to properly interpret the data, whereas they may remain optional elsewhere.

Versioning of the metadata standard: Peter suggested a move to version 2.0 given the scale of additions, while Charlie outlined a structured versioning approach used in related work, distinguishing between major, minor, and patch-level changes. There was general agreement to follow a similar logic, while also aligning with broader GA4GH guidance where possible.

Attention then shifted to how the metadata could integrate with discovery platforms such as Beacon. Tony supported incorporating the core checklist into Beacon models but highlighted that metadata designed for discovery requires stricter structure, such as controlled vocabularies and limited value ranges, compared to metadata intended for human or machine readability more broadly. David noted that the core checklist already distinguishes discovery-relevant fields, though not all fields (e.g. free text descriptions) are suitable for search. This sparked a broader point about the growing number of metadata initiatives within GA4GH, with Tony suggesting a dedicated workshop to align efforts and identify synergies.

Action items: Peter noted that most previously assigned tasks had been completed. Charlie reported progress on aligning targeted sequencing metadata with existing frameworks such as Phage, though ontology work remains ongoing.

Additional updates included plans to publish the checklist, improvements to ontology-linked instrument metadata, and ongoing maintenance of controlled vocabularies. Peter also described coordination with international databases through INSDC, where minimum experiment metadata specifications are being formalised to support broader global participation. He emphasised that this foundational work will enable further standardisation and adoption, with the expectation that alignment across organisations such as GA4GH, GSC, INSDC, and others will strengthen interoperability.

2026-04-15: Connect meeting session (Montreal, Canada)

Experiments Metadata working session

2026-03-09: LinkML draft for ExpMeta

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos, Coline Thomas (EGA@EBI-EMBL), Mousumi Ghosh, Karen Cranston, Xiaojian Shao

Regrets: Charlie Barclay

Link to meeting recording: https://us02web.zoom.us/rec/share/SRUrm8ci-9krIqEd4zeTNmg9r7hd_nLpgiYYbasi52V70WjumiZ4THC33v4VMbfj.GiXwQyipRbeYGVIg 

Meeting notes:

LinkML:

Linkml, GSC and Experiments Metadata

First iteration of YAML: https://drive.google.com/file/d/1TmhoJa5reXGpGxNreXIOLeQ8Fl04L6aK/view?usp=sharing 

Meeting summary:

The team currently has a core checklist for general sequencing experiments and plans to expand it into specialised categories (like long-read, single-cell, targeted sequencing). To support this, they intend to transition the current GitHub Markdown property list into a more computable schema format, likely using LinkML for now, even as broader GA4GH discussions on formal specification technologies continue within the TASC group.

Connect agenda:

The Experiments metadata Connect session is scheduled on Weds 15th from 2:00–3:30, which may be late for European participants but still manageable (around 7 p.m.).

David explained that the session will be structured as a working meeting rather than a beginner-oriented introduction, though it will still include a brief overview of the group’s work. The main goal is to organise breakout discussions focused on specific experiment metadata categories, such as long-read sequencing, targeted sequencing, and single-cell sequencing, with facilitators guiding each subgroup.

Karen offered to help facilitate discussions despite not having deep expertise, and Mousumi noted her background is primarily in proteomics but was willing to contribute where possible. Beatrice highlighted the challenge of planning breakout groups in advance, as attendee lists and expertise are often unclear, especially for virtual participants, and suggested preparing ahead to improve productivity. The group considered circulating a Google Sheet to gauge interest and expertise among participants and possibly identify additional topics such as proteomics, while limiting the number of breakout groups to around four for manageability.

👋 Xiaojian introduced himself as a research officer at the National Research Council of Canada and explained that attending conferences typically requires presenting related work; he expressed interest in understanding how his work on data analytics and single-cell genomics might connect to GA4GH activities.

LinkML and its use for formalising the experiment metadata dictionary:

David summarised his initial exploration of LinkML, explaining that it allows metadata properties (“slots”) and their types to be defined in YAML documents, which aligns with the group’s current approach of maintaining a dictionary of properties rather than a strict schema with classes.

Peter provided background on the decision to use LinkML, noting that the Genome Standards Consortium (GSC) has adopted it for its MIxS standards, and that LinkML enables conversion into multiple formats such as JSON, JSON-LD, RDF, and tabular outputs. He demonstrated an early prototype generated from the group’s spreadsheet using a Python script that converts entries into LinkML slot definitions, though additional work is needed to handle ontologies, controlled vocabularies, and enumerations more precisely.

The group discussed how far to go in structuring the model, whether to remain at a simple property dictionary level or introduce classes representing experiment types, which would allow required fields but could make the standard harder to maintain.

Beatrice asked about governance and editing, and Peter explained that the canonical source will eventually be a YAML file in GitHub, with community edits managed through pull requests.

Karen noted that although GA4GH’s TASC data model discussions have not mandated a single schema format, using LinkML should not conflict with GA4GH practices, especially since it can be converted to JSON Schema for discoverability in the schema registry.

The conversation then briefly shifted to ontology choices for instrument and experiment terms, comparing OBI with the Experimental Factor Ontology (EFO); Peter explained that EFO aggregates terms from multiple ontologies and may be more complete but less cleanly structured, and agreed to investigate its suitability further.

The plan is  to continue refining the LinkML model before the Connect meeting, share updates via the mailing list, and potentially use Connect discussions to resolve open modelling decisions.

2026-02-02

Chairs: David Bujold, Peter Woollard

Attendees: Sonja Mathias, Charlie Barclay, Coline Thomas (EGA@EBI-EMBL), Ramona Rego,

Regrets: Beatrice Amos

Action Items (Zoom AI generated):

• Peter: Loop Charlie and Nadeem in on the ELIXIR pathogen meeting topic and coordinate further discussion
• David: Reach out to local long-read sequencing experts to review and comment on additional properties relevant to long-read sequencing for the checklist
• David and Peter: By next meeting, review LinkML and come up with a proposal for using it as a source format for the experiment metadata checklist, including considerations for validation and conversion to other schema formats
• Charlie: Get the instrument terms into GenEpiO before the next meeting and coordinate with Damien and the OBI team regarding inclusion in OBI
• Charlie: Provide an update at the next meeting (or the one after) on the targeted sequencing and overlap in the ontology community, after consulting with Damien, Ko, and Chris
• David (with Beatrice and Peter): Look into scheduling a morning session for the Connect meeting in Montreal to maximise European participation
• Peter: Add James Fellow Yates' name to the notes regarding the ancient DNA/ancient environmental DNA checklist work
• Charlie: Check with Damien about suitable journals (e.g. ontology-focused) for potential manuscript submission
• David and Peter: Prepare the experiment metadata checklist for GSC submission, including mapping fields to LinkML format and providing example values, aiming for inclusion in GSC MixS 7.1 if possible - [summer 2026]
• Peter: Coordinate with INSDC people to point to the new GenEPO/OBI terms once available
• Charlie: Update the group on the status of ontology community targets at the next meeting(s) to identify possible overlaps or completed work that could benefit the group
• David: Follow up with Beatrice and Peter on the results of the Connect session proposal in the coming weeks

Link to meeting recording: https://us02web.zoom.us/rec/share/y0w0LUddQokahtfVj9Liox78K-Oq_h5okFhVZpaR2ReOIhe8jzTITHg38ToAX454.MyTNHlierS-Brhag 

Meeting notes:

• Action Peter: WRT Pathogen-ELIXIR community discussion about experiments metadata - contact Nadim and Charlie to give with the extra metadata needs for pathogens particularly around the pathogen workflows.
• Targeted sequences - Charlie (Canadian equivalent  CFI- and ECC) analysis to the UK DEFRA and BGE plus for metadata.
• For EGA Coline says that mainly genomic DNA, but some transcriptomics and even proteomics
• Talking about the “schema” for storing our metadata template in. Consensus was to try linkml. Need to think what the users want. Schema.org. Best for validating. Can start with linkml and convert to schema.org and JSON-LD etc.    Advantages that linkml community is embedded with OBO. Action: Peter and David, explore over the next month and come back with a proposal.
• 4) EGA - would like the ENA & ArrayExpress experiment metadata to be integrated. Also the metronome.
• From Pathogens: MLST - genotyping for multi genotypes.
• Sonja - at SciLife - want to know what experiment metadata
• For publications

• GSC - prepare linkml
• Thinking of journals: NAR, bioinformatics (no), Pensoft (no - too biodiversity)

• Ontologies

• instruments/platforms - Charlie hoping to go into GenPIO by March

• Damion asking the OBI people,.

• GSC checklists - overlapping ours:

• FAIRe - MIxS by Miwa Takahashi -   A Metadata Checklist and Data Formatting Guidelines to Make eDNA FAIR (Findable, Accessible, Interoperable, and Reusable)

• Environmental DNA 2025-05-01 | Journal article DOI: 10.1002/edn3.70100
• CONTRIBUTORS: Miwa Takahashi; Tobias Guldberg Frøslev; Joana Paupério; Bettina Thalinger; Katy Klymus; Caren C. Helbing; Cecilia Villacorta‐Rath; Katherine Silliman; Luke R. Thompson; Sean P. Jungbluth et al.

• Ancient DNA - by James Fellows Yates(MInAS) - aiming to have that in the

Meeting summary:

Since the approval of the GA4GH Experiments Metadata Checklist v1, David, Peter, and Beatrice have met twice to discuss next steps for the group, focusing on prioritising future work given limited time and many possible directions.

Phage Infectious Disease Community of Interest / ELIXIR Pathogen community meeting:

Peter provided an update from a recent ELIXIR Pathogen community meeting, where he presented the checklist and received very positive feedback. The pathogen community expressed strong interest in using the checklist but also advocated for expanding its scope, particularly beyond sequencing to encompass broader experimental metadata and pathogen-specific guidance.

Discussion highlighted interest in capturing more detailed experimental context, including pipelines and analysis-related information. Charlie noted that Phage currently integrates experiment and biosample metadata within a single standard and exports relevant subsets downstream, and mentioned the upcoming GA4GH/Phage Infectious Disease Community of Interest as a potential venue for collaboration.

A key theme was the challenge of defining boundaries between experimental and analysis metadata, particularly for pathogen and wastewater use cases, and how such information maps to existing repositories. While some of this may fall outside the current scope of the sequencing-focused checklist, participants agreed there is clear community demand and potential alignment opportunities, with Peter planning to connect Charlie and Nadeem to continue the discussion.

What’s next for ExpMeta:

David outlined the proposed plan for the year - there’ll be a shift from the core checklist toward developing extensions for specific experiment types.

1. A key priority identified was single-cell experiments, building on prior engagement with the Human Cell Atlas team, who have already developed a single-cell–focused checklist. The idea would be to review their work, identify reusable components, and integrate relevant blocks into the GA4GH framework, similar to the group’s earlier spreadsheet-based harmonisation process, while seeking broader community feedback. This was seen as a relatively low-effort, high-impact opportunity.

2. Targeted sequencing was another major focus area, with Peter informally identified as a lead given his biodiversity work and community connections. He shared that a new project grant involving Sanger and Biodiversity Genomics Europe creates momentum to refine targeted sequencing metadata across biodiversity, pathogen, and exome contexts. Additional related efforts include collaborations with DEFRA in the UK on biomonitoring, while Charlie noted parallel work in Canada with CFIAR and Environment and Climate Change Canada, highlighting potential for international alignment and interoperability across environmental and microbiome use cases.

3. The group also discussed feedback requesting more attention to long-read sequencing strategies, such as nanopore and PacBio. Although in-house expertise is limited, David has begun reaching out to colleagues at his sequencing center to review the checklist and suggest additional long-read–specific properties, making this an active work in progress.
4. Transcriptomics was revisited but deprioritised for now due to a lack of strong, clearly expressed demand from driver projects, despite some familiarity within the group and confirmation from Coline that both DNA and transcriptomic datasets are handled in her context. The consensus was to remain open to expanding into transcriptomics if someone is willing to lead it, but to stay focused to avoid overextending the group’s capacity and diluting progress across too many areas at once.

Formalising the published checklist into a machine-readable schema:

While earlier discussions had suggested aligning with recommendations from TASC (formerly DaMaSC), there is still no clear top-down directive on a specific schema technology.

David and Peter agreed it would be better to move forward with a concrete implementation rather than wait indefinitely, with the understanding that adjustments could be made later if needed.

JSON was widely supported in the discussion, with several people already using it in their projects.

JSON-LD and schema.org were also considered, particularly because of their ability to structure linked data and, in some cases, validate identifiers against external services such as identifiers.org.

However, questions arose about differences in validation capabilities, with Sonja emphasising the practical advantage of JSON Schema for validating required fields and checking whether completed templates conform to expectations.

Coline noted that at EGA they use JSON and support transformations between formats, reinforcing the idea that interoperability is feasible.

LinkML emerged as a strong candidate for a “source of truth” modeling layer, particularly since it allows export into multiple formats such as JSON Schema and JSON-LD.

Peter referenced its use within the Genome Standards Consortium, and Charlie highlighted its growing integration with ontologies through leadership connected to the OBO Foundry, including features like dynamic pick lists derived from ontology branches. Although LinkML is still maturing, participants viewed its active development as a strength rather than a liability.

What are user needs: different communities vary in technical capacity, and many current users still rely on spreadsheets and manual integration into their own schemas. More technically advanced communities, such as parts of the marine biodiversity space, are already working with JSON-LD.

David noted that in many real-world use cases, the checklist serves primarily as guidance for mapping and alignment rather than as a standalone validated document, with projects often implementing their own validators. The group tentatively agreed that exploring LinkML as a central modeling approach, with exports into other formats, made sense, and Peter and David proposed reviewing it further and returning with a concrete recommendation at the next meeting to avoid prolonged indecision.

Open discussion about priorities for the year:

The group then shifted to an open discussion about priorities for the year.

Peter noted that the checklist will be adopted by ENA by the end of the year regardless, but emphasised the importance of aligning with broader community interests.

Coline shared that from the EGA perspective, expansion into additional experiment types would be highly relevant, particularly long-read and transcriptomics.

She highlighted a current challenge: sequencing submissions follow the ENA schema, but transcriptomics, especially array-based data from resources like ArrayExpress, are handled through a separate and less structured schema. Harmonising sequencing and transcriptomics under a more unified framework would be very beneficial.

She also mentioned epigenetic data, such as methylation assays, where metadata capture is currently quite loose and often relies on free text (e.g. inconsistent naming of arrays).

Future plans at EGA include requesting more structured information based on declared assay type, similar to how sequencing submissions trigger specific required fields under the ENA schema, but this work is still in progress.

Additional experiment types were raised, including multilocus sequence typing (MLST) in the pathogen space, which Charlie noted is of interest to Phage for microbial strain typing. This further reinforced the broader theme of going beyond standard sequencing into more specialized assay types.

David explored whether EGA’s evolving model might incorporate more assay-specific properties in future iterations, particularly for methylation arrays, where technical details such as array design and probe content could be captured more consistently.

Sonja shared her perspective from SciLifeLab, where the focus is on supporting data submission workflows, especially to ENA, by defining practical experiment templates that ensure researchers collect the necessary metadata upfront.

She described herself as still getting oriented in the complexity of standards and schemas but emphasised that the checklist discussions are directly relevant to enabling smoother downstream submission.

Longer-term strategy:

David mentioned a potential manuscript describing the Experiments Metadata work beyond its status as a GA4GH product. David and Peter agreed that a key step toward a strong publication would be adoption by the Genome Standards Consortium (GSC).

Peter explained that there is an upcoming release of MIxS version 7 in April, but the group is unlikely to be ready in time for that release; however, inclusion in a subsequent minor update (e.g. 7.1) is realistic. To be incorporated into GSC, the terms would need to be modeled in LinkML format, with defined value patterns, controlled vocabularies where appropriate, and example values. While this requires effort, Peter emphasised that it is manageable if the group prepares the materials themselves, as GSC operates largely on volunteer capacity. Aligning LinkML work with GSC submission would effectively merge two objectives, schema formalisation and broader adoption, and strengthen the case for publication.

On the ENA side, Peter noted that while the checklist can be recommended in the short term (e.g. as structured XML attributes), full integration with validation and submission portal support would require development resources and likely funding. Still, incremental “baby steps,” including example submissions that flow through ENA to INSDC, could both demonstrate value and provide strong material for a future paper.

Ontology integration:

Charlie confirmed that most of the remaining work involves formally adding mapped instrument terms into GenEpiO (GenoPO) and eventually OBI. Progress had been briefly paused due to restructuring of the upper-level ontology hierarchy, but that work is now complete, and the terms should be added before the next meeting.

Peter expressed interest in encouraging INSDC to reference these ontology terms once available. Broader ontology developments may also reduce the group’s workload: Peter highlighted related efforts such as the FAIR environmental DNA (eDNA) checklist led by Miya Takadishi and work from the ancient DNA community, both of which are contributing structured terms into MIxS that overlap with targeted sequencing and pathogen use cases. These parallel initiatives may help cover portions of the metadata space the group is aiming to address.

GA4GH Connect meeting:

Finally, David noted that a session proposal has been submitted for the upcoming GA4GH Connect meeting in Montreal, with the aim of hosting focused discussions on experiment-type extensions, ontology alignment, and community engagement. While Peter may not attend in person due to funding constraints, the group discussed scheduling considerations to maximise remote participation, especially from Europe.

The meeting concluded with agreement to continue advancing schema formalisation, GSC alignment, ontology integration, and experiment-type extensions over the coming month, and to reconvene with updates at the next meeting.

Template Agenda

2026-MM-DD

Chairs: David Bujold, Peter Woollard

Attendees: Beatrice Amos,

Regrets:

Link to meeting recording:

Meeting notes:

Meeting summary:

Context

Background

The Minimal Metadata Study Group was formed after a GA4GH COVID Genomics Interoperability Network project workshop which took place in January of 2022. This study group will act as a venue for interested individuals/groups to determine the problem space and identify areas for collaboration.

Problem Statement

Metadata on (epi)genomic experiments are crucial to understanding output data. Many initiatives have their own standards to describe experiments for the data that they are sharing; however, the level of detail, methods of organization, and use of controlled vocabularies differ between these groups. Thus, there is a need to standardize a set of minimal metadata properties in order to increase shareability of (epi)genomic datasets.

Goals

Technical

• Understand the landscape of experiments metadata specs, which includes:

• Control vocabularies for experiment types and recommended ontologies
• Experiment types
• Experimental metadata schema linking between sample and sequencing experiment

Non-technical

• Identify active v.s. passive contributors
• Ensure representation of all relevant initiatives and external groups
• Agree on need for this group
• Determine focus areas for the group

Use Cases/Challenges

• See here

Group Structure & Meeting Format

Initial meetings of the study group meeting will focus on engaging relevant stakeholders in presentations containing the following information:

• An overview of their method for organisation (e.g. Data portal and relevant annotations)

• Minimal info on sample but not going too into detail re:linking

• Best practices and tools used
• Scope of their project
• Challenges and barriers/ pain points

After this information has been collected from all stakeholders, the study group will begin to identify areas of overlap and specific problem areas where creation of a standard may be beneficial. The group will develop this idea and make a standards proposal to the GA4GH Steering Committee once ready.

External Groups Involved

• IHEC experiment schema
• ENCODE: https://academic.oup.com/database/article/doi/10.1093/database/baw001/2630135
• GSC
• INSDC
• PHA4GE
• MINSEQE https://github.com/FAIRsharing/mircat/tree/master/minseqe 
• https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1008260#sec004 

Resources and Links

Experiments Metadata Working Document - 231204

Experimental Metadata - Participating Projects Description (2022)

Experimental Metadata Study Group Kick Off (2022-05-31)

10th_Plenary-Infectious_Disease_Community_Workshop-220819

Experiments Metadata Standard - Product Approval Questionnaire

Asynchronous Platforms

Slack Channel: #ga4gh-metadata-study-group

Mailing list: metadata-study-group@ga4gh.org

Please email beatrice.amos[at]ga4gh.org to request an invitation to these groups