The Minimum Evidence Record is a small set of provenance relationships that should travel with a consequential scientific or regulatory result when it leaves the system that produced it. We’re publishing it as a first open draft, alongside a formatted PDF, ahead of the EHDS implementing acts, the Data Union standardisation work and the Biotech Act.
Comprehensive provenance schemas tend to fail at exactly the point they are most needed. Every domain has a different view of what the complete picture looks like, so a schema rich enough to satisfy genomics will be unusable in pharmacovigilance, and one general enough for both will describe neither well. The result is that when an artefact crosses from one infrastructure to another, the receiving system cannot interpret what arrives and falls back on a PDF.
The alternative is the approach that has actually produced interoperability elsewhere in digital standards: agree a small stable core, and allow richer specialist profiles above it. This record is a proposal for that core. It describes what should travel, not what a system should log internally — an institution’s own audit trail will always be richer than this, and should be.
The record contains nine conceptual elements: persistent identifier, parent artefact(s), actor or process, time, method or workflow, material version information, integrity reference, governance context, and derived artefacts. Each is a relationship to capture, not a document to write. Two are worth calling out specifically.
Material version information is governed by materiality rather than completeness: the test is whether a change to the item could alter the interpretation, not whether the item has a version number. Capturing every package in an environment produces a record nobody reads; capturing the reference genome build, the typing scheme release and the clustering threshold produces one that answers the question actually asked five years later.
Derived artefacts is the field most commonly absent, and the one regulators need most. Without it, when an upstream error is found, there is no way to determine which conclusions are affected.
The full paper and PDF walk through the record at two points in a pathogen genomic surveillance chain: a cluster assignment derived from a consensus assembly and a reference typing scheme, and a public-health interpretation derived from that cluster assignment plus an access-restricted epidemiological line list. A verifier can establish that the resulting control-measure recommendation depends, through two steps, on a specific assembly and typing-scheme release — and that when the typing scheme is next updated, exactly which downstream artefacts are implicated — without ever seeing the restricted line list or any patient-level data. That separation, between verifying lineage and disclosing content, is the point of the design.
This record is intended to be expressed in existing standards, not to replace them. W3C PROV, RO-Crate and research-object approaches, workflow description standards, persistent identifier infrastructures, FAIR Digital Objects and EOSC’s interoperability work already cover most of the ground. We are actively working through which of the nine fields map cleanly onto existing vocabularies today, and where a domain profile will be needed — our own working assumption is that material versioning, governance context and derived artefacts are where that mapping is least settled.
Not a schema — it’s a conceptual model; a schema would need to be developed within an existing standards framework. Not a product specification. Not a requirement: no European instrument mandates this, it’s a proposal offered into work already under way. And not complete: some records will additionally need status, quality assertions, signatures or links to validation evidence, which belong in domain profiles above this core.
The full technical specification — all nine fields in detail, the complete worked example, the standards mapping and licensing — is available as a downloadable PDF from our Resources page. Published under a Creative Commons Attribution 4.0 licence: reuse, adapt and build on it, including commercially, with attribution.
A core has to be small enough that a system in a different domain can implement it without adopting a foreign worldview, and expressive enough that the relationships which determine scientific meaning survive the journey. Nine is the smallest set we found that still supports real verification.
No. This describes what should travel with an artefact when it leaves the system that produced it, not what a system should log internally. An institution’s own audit trail will always be richer than this record, and should be.
The white paper, From Data Governance to Evidence Governance, proposes a European minimum evidence record as its central recommendation. This specification is that record, published in full detail as its own technical document.
No — it’s a conceptual model and a first open draft, version 0.1. A schema would need to be developed within an existing standards framework, which is the point of publishing the core now rather than waiting.
If your organisation generates, manages, analyses or governs biological evidence and would be interested in participating in a UK or European provenance demonstrator, The BioChain would welcome the conversation.
Get in touch
Regulatory developments, technical notes and platform news — sent occasionally, straight to your inbox.