Blogs, Genomics

Variant Evidence Management: The Reuse Audit for Clinical Genomics Laboratories

Scientists reviewing variant evidence together

The central problem in variant evidence management is not the total number of records in ClinVar. It is whether the laboratory can recover the reasoning behind a prior classification when that reasoning is needed again.

Many laboratories already store classifications and applied ACMG/AMP criteria in a variant database or interpretation platform. A signed report communicates the approved conclusion, while those systems may retain the detailed assessment. Reuse becomes difficult when the linked records do not preserve a recoverable account of the evidence reviewed, source versions, applicable rules, clinical context, and rationale behind the decision.

Reuse audit record 01
Prior internal classification
Likely pathogenic
Last evaluation
18 months ago
Final report available
Yes
Original evidence packet retrievable
Partial
Applied ACMG/AMP criteria recoverable
Yes
Evidence-source versions recoverable
No
Disease-context scope stated
No
Prior reviewer rationale recoverable
Partial
New external evidence assessed
Unknown
Safe reuse decision
Pending review

The classification is present. The institutional reasoning is incomplete.

Variant evidence management gives a laboratory the ability to preserve, retrieve, evaluate, and reuse its clinical reasoning under controlled review. That capability becomes critical when a previously classified variant appears in a new case, a public assertion changes, a clinician questions a result, or a reanalysis program expands.

Reuse Audit Record 01: Final Classifications Do Not Preserve Full Clinical Reasoning

A signed report is a patient-care artifact. It communicates an approved conclusion for a defined patient, indication, and reporting moment. It is not designed to preserve every source record, rejected line of evidence, data-release version, review discussion, or criterion-level rationale needed for future use.

Where detailed criteria and evidence are already held in a variant database, the internal workflow should link the issued report to the reviewed assessment. The full evidence packet does not need to be duplicated in the patient-facing report; it needs to remain recoverable from the laboratory’s interpretation records.

The ACMG/AMP framework for sequence-variant interpretation uses five classification categories: pathogenic, likely pathogenic, variant of uncertain significance, likely benign, and benign. It weighs evidence from population data, computational evidence, functional studies, segregation, and other relevant observations.

A future reviewer usually needs more than the final category. The reusable record should show the variant representation, gene-disease and inheritance context, source material reviewed, source dates or versions, ACMG/AMP criteria considered, applied strengths, reviewer rationale, approval history, and relationship to reports or amendments.

When those records are dispersed across reports, analyst notes, shared folders, and personal workspaces, the laboratory repeats reasoning it has already performed. It also becomes harder to determine whether a new publication, ClinVar submission, or internal observation is material to an earlier conclusion.

Connect interpretation, clinical review, reporting, and reusable knowledge.

NonStop engineers clinically governed genomics workflows that connect evidence assembly, review, reporting, and the data assets required to make interpretation work durable.

Reuse Audit Record 02: Evidence Without Variant Provenance Cannot Be Reliably Reused

A PubMed link without the reviewed claim, a ClinVar record without its review status, or an internal case note without phenotype context may be stored in the system. That does not make it reusable evidence.

The evidence record needs to preserve both the source and the laboratory’s assessment of the source.

Evidence record
Identity
Canonical identifier, genome build, HGVS, transcript, protein effect
Context
Gene, disease, inheritance model, relevant phenotype or assay scope
Source
Publication, ClinVar record, population dataset, functional study, or internal observation
Source state
Publication date, database release, record version, retrieval date
Assessment
ACMG/AMP criterion, applicable specification, applied strength, rationale
Review
Analyst, approver, review date, decision state
Reuse boundary
Context in which the assessment may be reused

The ClinGen Criteria Specification Registry maintains gene- and disease-specific specifications that refine application of ACMG/AMP evidence criteria. A provenance-aware record must therefore capture more than “PS3 applied” or “PM2 applied.” It needs the evidence item, source state, applicable specification, clinical context, assessment rationale, and reviewer decision.

Reuse Audit Record 03: Internal Precedent Is a Retrieval Signal, Not a Classification Shortcut

“We saw this variant before” should trigger retrieval, not end the review.

A previous internal assertion may have been made for another disease, phenotype, inheritance model, transcript policy, assay reportable range, evidence set, ACMG/AMP specification, or reporting policy. The molecular change may match, while the clinical decision does not.

A controlled knowledge record separates two tasks:

Precedent
  • What did the laboratory conclude previously?
  • Which evidence, context, policy version, and approval supported it?
  • Has the assertion been superseded, amended, or flagged for review?
Reuse review
  • Does the current case fit the prior assertion’s defined scope?
  • Has external or internal evidence changed?
  • Do current assay, transcript, indication, or policy conditions alter applicability?

This is the operating point of clinical variant knowledge management. Historical decisions remain valuable, but they remain linked to their original scope and evidence. They are not promoted to automatic answers.

ClinVar aggregates submitted relationships between human variation and health, but its variant-level pages retain records from different submitters, conditions, and evaluations. The ClinVar VCV record model is useful for seeing why one matching public classification cannot replace a laboratory’s contextual assessment.

Reuse Audit Record 04: ClinVar Evidence Tracking Must Create a Reanalysis Work Item

ClinVar’s review-status model differentiates submissions with no stated criteria from assertions with criteria provided by a single submitter, multiple submitters without conflict, expert-panel review, and practice guidelines.

A “criteria provided, single submitter” assertion requires the submitter to provide a comprehensive evidence review consistent with current practice, state the criteria used, and provide supporting evidence or rationale, or identify a public contact who can supply it. Those expectations are defined in the ClinVar submission review guidelines.

These signals should create a controlled reanalysis workflow, not disappear into email alerts.

Reanalysis queue item
Change event
New ClinVar expert-panel assertion
Affected internal assertion
Internal assertion ID and canonical variant ID
Current internal classification
Likely pathogenic
External change
Condition-specific assertion differs from current internal scope
Affected reported cases
Identified by approved policy
Required action
Interpretation lead review
Outcome
Retained, revised, escalated, or no action
Report amendment
Determined by qualified clinical and laboratory leadership

The queue should also receive relevant literature, population-data updates, gene-disease validity changes, internal segregation or phenotype data, transcript-policy changes, and laboratory-policy updates. Qualified laboratory and clinical leaders remain responsible for reclassification, report amendments, and provider or patient communications. The workflow system preserves the evidence trail, prioritization, ownership, and audit record.

Reuse Audit Record 05: Automation Can Retrieve Evidence Without Replacing Clinical Judgment

Automation can normalize variant representations, retrieve sources, compare external records, detect changes, identify prior internal assertions, assemble evidence packets, and prioritize review queues.

Those capabilities do not remove the need for qualified review of whether evidence meets a criterion, whether a prior assertion applies to a new case, whether a classification should change, or whether a report amendment is appropriate.

Where generative AI is used to draft an evidence summary, label the output as AI-generated and retain links to the underlying sources. The workflow should preserve the relevant model/version information, draft date, and reviewer edits or approval. The qualified reviewer checks the summary against the original evidence; an AI-generated summary is not an independent source of clinical evidence.

Automation may supportQualified clinical review determines
Variant normalization and identity matchingWhether the match is clinically applicable in the current context
Evidence retrieval and comparisonWhether evidence meets the applicable ACMG/AMP criterion and strength
ClinVar and literature monitoringWhether a change materially affects an internal assertion
Prior-precedent retrievalWhether prior context matches the current disease, inheritance, assay, and patient context
AI-generated draft evidence summaries, where usedWhether the AI-generated summaries accurately reflect the cited sources and are complete and clinically appropriate
Queue prioritizationWhether to classify, reclassify, amend, report, or communicate

The ACMG/AMP guideline states that clinical interpretation should include evidence supporting the classification and consider whether the finding may explain the patient’s indication for testing. That requirement places patient context and professional judgment at the center of the process, even when automated systems improve evidence retrieval and review preparation. Read the ACMG/AMP guideline.

A governed workflow directs automation toward retrieval, normalization, comparison, and prioritization. Qualified professionals retain responsibility for classification, reportability, patient-specific relevance, approval, and downstream action.

Build structured interpretation workflows around evidence and review.

Varion supports clinical variant-analysis workflows, evidence assembly, review queues, and controlled handoffs from tertiary analysis into clinical reporting.

The Variant Evidence Reuse Decision Has Four Outcomes

A prior internal classification does not require a binary choice between reuse and full reinterpretation. A controlled workflow can record four outcomes.

OutcomeMeaningRequired record
Reuse within defined scopeIdentity, approved evidence, context, and policy align with the defined reuse boundaryScope match, retrieval date, and reviewer confirmation where required
Reuse after targeted confirmationPrior reasoning remains relevant, but a source refresh or limited review is neededFocused review, updated sources, and approval
Full reassessmentNew evidence, context, policy, or uncertainty could alter the assertionNew assessment version linked to prior evidence and assertion history
Do not reuseIdentity, provenance, context, rationale, or policy alignment is insufficientNew interpretation with the prior assertion retained as historical context

This approach prevents unnecessary reinvention when a provenance-rich assertion remains applicable. It also prevents teams from treating incomplete precedent as a shortcut.

A Reusable Clinical Variant Assertion Is a Defined Knowledge Object

A reusable assertion needs identity, evidence, assessment, and assertion states.

Identity

Coordinates, reference sequence, HGVS, gene, transcript, molecular consequence, VRS or other canonical identifier

Evidence

Source, source version or release, source and retrieval dates, evidence type, context, limitations

Assessment

ACMG/AMP criterion, applicable specification or policy version, applied strength, rationale, reviewer, approver, review state

Assertion

Classification, condition and inheritance scope, reportability, effective date, supersession history, linked reports or amendments, monitoring rule

The GA4GH Variation Representation Specification provides a computational framework for representing genomic variation consistently and assigning globally unique computed identifiers. VRS v2.0 added standardized descriptions intended to support integration of variant evidence across sequence collections and resources.

A canonical identifier does not replace HGVS notation, transcript-aware interpretation, disease context, or clinical judgment. It gives a laboratory a stable technical handle for linking evidence, assertions, reports, and external updates that might otherwise be stored under incompatible representations.

What NonStop Engineers Around Reusable Variant Knowledge

Durable clinical variant knowledge management connects the laboratory’s existing variant database, evidence records, review decisions, and reporting history.

It is a connected operating system for evidence, review, approval, reuse, reanalysis, and reporting.

NonStop engineers that operating system through two connected capabilities.

  • Varion supports clinical variant-analysis workflows, review queues, evidence assembly, and controlled movement from tertiary analysis into clinical reporting. NonStop describes Varion as building on a shared GA4GH VRS store with GENVAR and feeding approved results to reporting.
  • GENVAR turns a laboratory’s classifications into a versioned, queryable knowledge base on a GA4GH VRS store and includes ClinVar diff monitoring.
  • Clinical reporting engineering connects an approved assertion to report content, sign-out, structured result fields, amendments, and downstream delivery.
  • Data engineering and platform engineering establish the canonical data model, evidence-ingestion pathways, change monitoring, access controls, audit records, and interfaces required for a client-owned knowledge asset.
  • Forward-deployed engineering works with interpretation, laboratory, and technology teams in the real workflow: mapping evidence sources, defining assertion scope, designing review queues, connecting reporting, and transferring durable operational control to the laboratory.

The objective is not to automate medical judgment.

It is to make the laboratory’s validated reasoning accessible, reviewable, and reusable under the governance that clinical work requires.

Build an owned, versioned clinical variant knowledge base.

GENVAR provides the knowledge layer for reusable classifications, evidence provenance, internal precedent, ClinVar change monitoring, and controlled reanalysis workflows.

Assess Whether Your Variant Classifications Are Reusable Knowledge

NonStop’s Variant Knowledge Architecture Review maps how variant identity, ACMG evidence management, ClinVar tracking, internal precedent, clinical review, reanalysis, reporting, and audit records currently connect.

The review identifies:

  • Classification rationale trapped in reports, spreadsheets, PDFs, or individual analysts’ workspaces
  • Evidence records that lack source, date, version, context, or reviewer provenance
  • ClinVar and literature changes that do not create controlled reassessment queues
  • Internal precedent that cannot be reused safely because its original context is unclear
  • Gaps between approved classifications, report templates, structured results, and amendment workflows
  • A phased architecture for a client-owned clinical variant knowledge base