Leveraging Disease Databases (ClinVar, HGMD, OMIM) for Variant Classification: How to Interpret Conflicting and Curated Entries

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-21

Key Takeaways

Variant classification requires systematic evaluation of ClinVar, HGMD, and OMIM, recognizing their distinct purposes: ClinVar archives submitter assertions with review stars, HGMD catalogs published disease-associated variants, and OMIM details gene-phenotype relationships.
ClinVar's star rating system (0-4 stars) indicates submission quality and consistency, guiding evidence review but not replacing it; higher stars generally denote greater confidence, but even expert panels can revise classifications.
Resolving conflicting database entries necessitates reviewing primary literature and applying ACMG/AMP criteria, particularly distinguishing between PS4 (strong evidence of enrichment) and PP5 (reputable source report without sufficient independent evidence).
A tiered evidence adjudication system (Tier 1: direct evidence, Tier 2: curated indirect evidence, Tier 3: minimal/unverifiable evidence) and a two-pass review method are crucial for structured conflict resolution, preventing confirmation bias.
Quantitative functional data, such as residual wild-type transcript levels in splicing assays, should be integrated into conflict resolution, acknowledging limitations of computational predictions and assay variability.
Maintaining a variant conflict log documenting database versions, evidence tiers, review findings, and final classifications ensures reproducibility and facilitates re-evaluation as new evidence emerges.

Variant scientists routinely encounter conflicting classifications when querying public disease databases. A variant labeled pathogenic in one resource may appear as benign or uncertain in another, and the same entry can change meaning across database versions. This article provides a systematic framework for evaluating the reliability of database submissions, understanding ClinVar star ratings, and resolving conflicts by reviewing primary literature and applying ACMG/AMP criteria such as PS4 and PP5 appropriately. The practical outcome is a reproducible workflow for weighing database evidence without overtrusting any single entry.

Scope and Reader Context

This guidance addresses germline variant classification for research and clinical laboratory settings. The intended readers are biology students, researchers, laboratory professionals, and life-science practitioners who need to interpret variant calls from sequencing pipelines. The focus is on three major public resources: ClinVar, the Human Gene Mutation Database (HGMD), and Online Mendelian Inheritance in Man (OMIM). Each database serves a distinct purpose, and understanding those differences is the first step toward resolving apparent conflicts.

The workflow described here applies to constitutional variants identified through germline variant calling. Somatic variant calling follows different logic because tumor heterogeneity, clonal hematopoiesis, and sample purity introduce additional variables. The principles of evidence evaluation remain relevant, but the ACMG/AMP framework discussed in this article was designed for germline interpretation.

At a Glance

Database	Primary Content	Submission Model	Key Limitation for Classification
ClinVar	Variant-level assertions with supporting evidence	Submitter-driven with review status stars	Variable submitter expertise and outdated entries
HGMD	Published disease-associated variants	Literature curation by professional curators	Overrepresentation of rare variants and potential false positives
OMIM	Gene-phenotype relationships and curated summaries	Expert editorial review	Limited variant-level detail and slower update cycles

The table above summarizes the structural differences that explain many apparent conflicts. ClinVar aggregates submissions from multiple laboratories and includes a review status system. HGMD focuses on variants reported in the published literature. OMIM provides gene-level and phenotype-level context instead of comprehensive variant-level classification.

Core Principles of Database Evidence Evaluation

Database Purpose Determines Interpretation Weight

ClinVar, HGMD, and OMIM were built for different purposes. ClinVar functions as a public archive of variant classifications submitted by laboratories, with each submission carrying a review status. HGMD catalogs variants reported in the literature as disease-associated, with a focus on rare alleles. OMIM provides expertly curated summaries of gene-phenotype relationships, inheritance patterns, and clinical descriptions.

The National Center for Biotechnology Information hosts ClinVar and describes it as a resource for relationships between human variations and phenotypes with supporting evidence. The same organization provides access to related sequence resources and analysis services through its broader data platform. Understanding that ClinVar is an archive of assertions instead of a single authoritative judgment is essential for correct interpretation.

HGMD differs in that it applies professional curation to published literature. The database includes variants that have been reported in association with disease, but the threshold for inclusion may be lower than the evidence required for a ClinVar pathogenic classification. A variant present in HGMD as disease-causing may not meet ACMG/AMP criteria for pathogenicity when evaluated systematically.

OMIM operates at a different level of granularity. It describes the relationship between genes and phenotypes, including inheritance patterns and clinical features. OMIM entries often mention specific variants as examples, but the database does not provide systematic variant-level classification comparable to ClinVar.

Review Status and Star Ratings in ClinVar

ClinVar assigns review status based on the number of submitters and the consistency of their classifications. The star rating system ranges from zero stars for entries with no assertion criteria to four stars for classifications from a professional guidelines body. One star indicates a single submitter with assertion criteria, two stars indicate multiple submitters with conflicting interpretations, and three stars indicate multiple submitters with consistent classifications.

The star rating provides a quick indicator of confidence, but it does not replace direct evaluation of the underlying evidence. A four-star classification from an expert panel carries substantial weight, yet even expert panels can revise their conclusions when new data emerge. A zero-star entry from a single laboratory may be correct, particularly for recently discovered variants that have not yet accumulated multiple submissions.

The practical implication is that star ratings should guide the order of evidence review, not substitute for it. A variant with conflicting submissions at two stars requires deeper investigation than a variant with consistent three-star classifications. The review status also affects how the variant should be weighted in ACMG/AMP scoring.

The Problem of Outdated and Conflicting Entries

Database entries are not static. Classifications change as new evidence accumulates, and older entries may reflect outdated understanding of variant effects. A study of ClinVar and HGMD archives found that variant misclassification has decreased over time, but both databases still contain errors. The same study reported that ClinVar variants classified as pathogenic or likely pathogenic are reclassified more frequently than HGMD disease-causing variants, which likely contributes to ClinVar's lower false-positive rate.

This finding has direct practical implications. When a variant classification changes, the older classification may persist in downloaded database versions, published papers, and laboratory records. Checking the current database entry is necessary, but even current entries may not reflect the most recent literature. The resolution of conflicts requires reviewing primary evidence instead of relying on database labels alone.

Practical Workflow for Variant Classification

Step 1: Collect All Available Database Entries

Begin by querying ClinVar, HGMD, and OMIM for the variant of interest. Record the following information for each entry: the classification, the submitter or curator, the date of the submission or update, the review status, and any supporting evidence cited. This collection step creates the foundation for conflict resolution.

For ClinVar, note the number of submitters and whether their classifications agree. For HGMD, record whether the variant is listed as disease-causing or disease-causing with unknown significance. For OMIM, identify the gene-phenotype relationship and any variant-specific comments in the entry.

The National Center for Biotechnology Information provides search tools that allow cross-referencing between databases. The same variant may appear under different identifiers in different resources, so confirming that all entries refer to the same genomic position and nucleotide change is essential.

Step 2: Assess Submission Quality and Currency

Evaluate each database entry for quality indicators. In ClinVar, the star rating and the identity of the submitter matter. Submissions from large clinical laboratories with established variant interpretation programs generally carry more weight than submissions from individual researchers. The date of the submission matters because classification standards have evolved.

For HGMD entries, check whether the original publication is accessible and whether the variant was reported in a single family or multiple unrelated individuals. A variant reported in a single family with limited segregation data carries less weight than a variant observed in multiple independent cases.

Step 3: Review Primary Literature

Database entries are secondary sources. The primary literature contains the original observations, experimental data, and clinical descriptions that support or refute pathogenicity. Reviewing the primary literature is mandatory when database entries conflict or when the variant is being considered for clinical reporting.

Search for the variant by genomic position, gene name, and nucleotide change. Look for functional studies, segregation analyses, population frequency data, and case reports. The goal is to assemble an independent evidence base that can be compared against the database classifications.

The Galaxy Training Network provides accessible tutorials on variant analysis workflows that include steps for annotating variants with database information and reviewing supporting evidence. These tutorials emphasize reproducibility and documentation, both of which are essential for clinical variant interpretation.

Step 4: Apply ACMG/AMP Criteria Systematically

The American College of Medical Genetics and Genomics and the Association for Molecular Pathology published a framework for variant classification that uses evidence categories to assign classifications of pathogenic, likely pathogenic, uncertain significance, likely benign, or benign. This framework is the standard for clinical variant interpretation.

Two criteria deserve special attention in the context of database use. PS4 applies when the variant is present in affected individuals at a frequency significantly increased over controls. Database entries can support PS4 if they document multiple unrelated affected individuals, but the strength of the evidence depends on the quality of the case data. PP5 applies when reputable sources report the variant as pathogenic, but the criterion is explicitly designed for situations where the source does not provide sufficient evidence for a stronger criterion.

The distinction between PS4 and PP5 is critical. PS4 requires direct evidence of variant enrichment in affected individuals. PP5 is a default criterion that should be used sparingly and only when the database entry cannot be independently verified. Overreliance on PP5 can lead to classification errors, particularly when database entries are outdated or conflicting.

Step 5: Resolve Conflicts Using an Evidence Hierarchy

When database entries conflict, resolve the conflict by applying an evidence hierarchy. Primary experimental data and clinical observations carry the most weight. Well-curated database entries with consistent multi-submitter support carry intermediate weight. Single-submitter entries and literature-only reports carry the least weight.

The evidence hierarchy should be applied consistently across variants. A variant with conflicting ClinVar submissions but strong functional data supporting a deleterious effect may warrant a pathogenic classification despite the database conflict. Conversely, a variant with consistent database classifications but no functional data and conflicting population frequency evidence may warrant downgrading to uncertain significance.

The taxonomic schema of potential pitfalls in clinical variant analysis highlights the range of errors that can occur when databases are used without adequate scrutiny. These pitfalls include outdated entries, misapplied criteria, and failure to recognize the limitations of computational prediction tools. A systematic evidence hierarchy reduces the risk of these errors.

Options and Tradeoffs in Database Selection

ClinVar as the Primary Resource

ClinVar offers several advantages for variant classification. The database is publicly accessible, includes submissions from multiple laboratories, and provides a review status system that indicates confidence. The reclassification rate for ClinVar pathogenic variants suggests that the database is responsive to new evidence.

The tradeoff is that ClinVar contains entries of variable quality. Zero-star entries may lack assertion criteria, and conflicting submissions require resolution. The database also depends on submitter participation, so variants from understudied populations may have limited representation.

HGMD for Literature Coverage

HGMD provides comprehensive coverage of variants reported in the published literature. This coverage is valuable for identifying variants that have been described in case reports or small studies that may not have submitted data to ClinVar.

The tradeoff is that HGMD includes variants based on literature reports that may not meet current classification standards. The study comparing ClinVar and HGMD found that HGMD variants imply a higher disease burden in population samples, suggesting a higher false-positive rate. HGMD should be used as a discovery tool and literature index instead of as a primary classification source.

OMIM for Gene-Phenotype Context

OMIM provides essential context for understanding whether a gene is associated with the phenotype under investigation. The database describes inheritance patterns, clinical features, and allelic heterogeneity that inform variant interpretation.

The tradeoff is that OMIM does not provide systematic variant-level classification. The database is most useful for establishing gene-disease validity and understanding the clinical spectrum associated with a gene. Variant-level decisions should be based on ClinVar, HGMD, and primary literature.

Integration Across Resources

The most effective approach integrates all three databases. Use OMIM to establish gene-disease validity and inheritance patterns. Use ClinVar to assess variant-level classifications and review status. Use HGMD to identify literature reports and discover variants that may not be in ClinVar. Then review the primary literature to resolve conflicts and build an independent evidence base.

The nf-core documentation describes community standards for reproducible bioinformatics pipelines. These standards include version control, containerization, and documentation practices that apply to variant annotation workflows. Integrating database queries into a reproducible pipeline ensures that classifications can be traced to specific database versions and retrieval dates.

Observations and Measurements in Variant Interpretation

Population Frequency as a Key Measurement

Population frequency data provide one of the most powerful measurements for variant classification. Variants that are common in general populations are unlikely to cause highly penetrant Mendelian disorders. The ACMG/AMP framework includes criteria for benign classification based on population frequency.

The study of ClinVar and HGMD accuracy found that removing common variants in accordance with classification guidelines reduced the ancestry-related bias in false-positive rates. This finding underscores the importance of applying population frequency thresholds consistently across ancestry groups.

When evaluating population frequency, consider the ancestry composition of the control database. A variant that is rare in European populations may be common in African or Asian populations. The 1000 Genomes Project data used in the accuracy study included diverse ancestry groups, and the analysis revealed that African ancestry individuals were more likely to be incorrectly indicated as affected when HGMD variants were used.

Segregation Data as Observational Evidence

Segregation analysis provides observational evidence that can support or refute pathogenicity. The ACMG/AMP framework includes criteria for pathogenic classification based on segregation of the variant with disease in multiple affected family members.

Database entries may include segregation data, but the quality of the data varies. A variant that segregates with disease in a large multi-generation family carries more weight than a variant observed in a single affected individual. Review the primary literature to assess the quality of segregation data before applying segregation-based criteria.

Functional Assays as Experimental Evidence

Functional studies provide experimental evidence that can strongly support pathogenicity. The ACMG/AMP framework includes criteria for pathogenic classification based on functional studies showing a deleterious effect.

The study of GT>GC splice-site variants illustrates the complexity of functional assessment. These variants can retain variable levels of residual wild-type transcript, creating a functional gray zone that complicates classification. The study found that approximately 15 to 18 percent of such substitutions retain variable levels of residual wild-type transcript, and that computational prediction and experimental assessment both have limitations in capturing the continuum of variant effects.

This finding has direct implications for variant classification. A functional assay that shows complete loss of function supports pathogenicity, but an assay that shows partial function requires careful interpretation. The intermediate effects may be disease-causing in some contexts and benign in others, depending on the gene, the disease mechanism, and the threshold for clinical significance.

Records and Documentation for Variant Classification

Maintaining a Variant Evidence Log

Document every piece of evidence considered during variant classification. The log should include the database version and retrieval date for each entry, the classification and review status, the submitter or curator, and the evidence cited. This documentation supports reproducibility and provides a basis for revisiting classifications when new evidence emerges.

The Carpentries lessons on data management and reproducibility provide foundational practices for organizing and documenting analysis workflows. These practices apply to variant classification, where the evidence base can change rapidly and classifications must be traceable to specific data sources.

Tracking Database Version Changes

Database entries change over time. A variant classified as uncertain significance in one database version may be reclassified as pathogenic in a later version. Tracking these changes requires recording the database version and retrieval date for every query.

The study of ClinVar and HGMD archives used historical versions of both databases to investigate how misclassification has changed over time. This approach demonstrates the value of maintaining database archives and tracking version changes. Laboratories should implement procedures for periodic re-evaluation of reported variants when database versions are updated.

Documenting Classification Rationale

The classification rationale should document which ACMG/AMP criteria were applied and how each criterion was weighted. The rationale should also explain how database conflicts were resolved and why certain evidence was prioritized over other evidence.

The documentation should be detailed enough that another qualified professional can understand the reasoning and reproduce the classification. This transparency supports quality control and provides a basis for professional consultation when classifications are uncertain.

Common Failure Patterns in Database-Driven Classification

Overreliance on Database Labels

The most common failure pattern is treating database labels as definitive answers. A variant labeled pathogenic in HGMD may not meet current ACMG/AMP criteria. A variant labeled uncertain in ClinVar may have strong evidence for pathogenicity that has not yet been submitted to the database.

The study of secondary findings in a Lebanese cohort illustrates this problem. The researchers found that 16.8 percent of participants had actionable variants based on ACMG/AMP criteria, but only 6 percent had actionable variants based on ClinVar classifications. The discrepancy highlights the limitations of relying on database entries alone, particularly for understudied populations.

Ignoring Population Frequency Data

A second failure pattern is ignoring population frequency data when it conflicts with database classifications. A variant that is common in population databases cannot be pathogenic for a highly penetrant Mendelian disorder, regardless of database labels.

The accuracy study found that removing common variants reduced false-positive rates and eliminated ancestry-related bias. This finding supports the practice of checking population frequency before accepting a pathogenic classification from any database.

Applying PP5 Without Independent Verification

A third failure pattern is applying PP5 based on database entries without independent verification. PP5 is intended for situations where a reputable source reports pathogenicity but does not provide sufficient evidence for a stronger criterion. Applying PP5 when the database entry is outdated, conflicting, or based on limited evidence can lead to classification errors.

The ACMG/AMP framework explicitly cautions against overuse of PP5. The criterion should be applied only when the database source is reputable and the classification is consistent with the available evidence. When database entries conflict, PP5 should not be used to resolve the conflict.

Failing to Recognize Intermediate Effects

A fourth failure pattern is treating variant effects as binary. The study of GT>GC splice-site variants demonstrates that some variants have intermediate effects that are neither fully disruptive nor functionally neutral. These intermediate effects complicate classification because they may not meet the threshold for pathogenic classification based on loss-of-function criteria, yet they may still contribute to disease.

The study authors recommend classification frameworks that incorporate quantitative functional data and better capture the continuum of variant effects. Until such frameworks are widely adopted, variant scientists should be cautious about classifying variants with intermediate functional effects as benign based solely on the absence of complete loss of function.

Limitations of Database-Driven Classification

Database Coverage Gaps

All three databases have coverage gaps. ClinVar depends on submitter participation, and variants from understudied populations or rare diseases may have limited submissions. HGMD covers published literature but may miss variants that have not been reported in publications. OMIM provides gene-level context but limited variant-level detail.

The Lebanese cohort study found significant discrepancies between ACMG/AMP and ClinVar classifications, with many ACMG-classified pathogenic variants absent from ClinVar. This finding underscores the need for harmonization of variant databases and updates to ClinVar entries, particularly for understudied populations.

Computational Prediction Limitations

Computational prediction tools are commonly used to support variant classification, but these tools have inherent limitations. The GT>GC splice-site study found that computational prediction and experimental assessment both have limitations in capturing the continuum of variant effects. SpliceAI scores, while useful, cannot fully predict the functional consequences of variants with intermediate effects.

The taxonomic schema of variant analysis pitfalls emphasizes that computational tools should be used as supporting evidence instead of as definitive classifiers. The limitations of these tools should be documented in the classification rationale.

Evolving Classification Standards

Classification standards have evolved over time, and older database entries may reflect outdated standards. The ACMG/AMP framework was published in 2015, and subsequent refinements have addressed specific evidence categories and gene-specific considerations. Variants classified before the framework was adopted may not meet current standards.

The accuracy study found that both ClinVar and HGMD have improved over time, but the improvement has been uneven. ClinVar variants are reclassified more frequently than HGMD variants, which likely reflects the different submission and curation models. Variant scientists should check the date of database entries and consider whether the classification reflects current standards.

Safety and Regulatory Context

Clinical Reporting Requirements

Variant classifications used for clinical reporting must meet professional standards. The ACMG/AMP framework provides the standard for germline variant classification, and laboratories should document their application of the framework. Classifications should be reviewed by qualified professionals and supported by adequate evidence.

The secondary findings study raises important ethical questions about which criteria to prioritize for patient disclosure. The researchers found that relying on ACMG/AMP criteria identified more actionable variants than relying on ClinVar alone, but the absence of ACMG-classified variants in ClinVar complicated reporting due to a lack of evidence linking them to disease in other individuals.

Professional Escalation Criteria

Variant scientists should escalate uncertain classifications to qualified professionals for review. Escalation is appropriate when database entries conflict, when the variant has not been reported in the literature, when functional studies show intermediate effects, or when the classification would change clinical management.

Ethical Considerations in Database Use

Database use raises ethical considerations related to data quality, population representation, and clinical impact. The Lebanese cohort study found that understudied populations may have limited representation in variant databases, which can lead to misclassification and inequitable clinical outcomes.

Variant scientists should be aware of these ethical considerations and should advocate for increased diversity in variant databases. The harmonization of variant databases and updates to ClinVar entries are essential for improving the accuracy and equity of variant classification.

Building a Structured Conflict Resolution Protocol for Database Discrepancies

Establishing a Tiered Evidence Adjudication System

A systematic conflict resolution protocol begins with assigning each database entry to a defined evidence tier before any ACMG/AMP scoring occurs. This tiering step prevents the common error of weighting all database entries equally during classification. The protocol described here uses three tiers that correspond to the structural differences between ClinVar, HGMD, and OMIM described earlier in this article.

Tier 1 contains entries with direct experimental or clinical evidence that can be independently verified. This includes ClinVar submissions with three or four stars, peer-reviewed functional studies, and segregation data from multiple unrelated families. Tier 2 contains entries with curated but indirect evidence, including HGMD disease-causing entries with accessible primary publications and ClinVar submissions with two stars where the conflict is documented. Tier 3 contains entries with minimal or unverifiable evidence, including zero-star ClinVar submissions, HGMD entries without accessible publications, and OMIM variant mentions that lack supporting detail.

The tier assignment must be recorded before any conflict resolution begins. This sequencing matters because it prevents the adjudicator from being influenced by the eventual classification outcome. A variant that ultimately receives a pathogenic classification should not retroactively receive a higher evidence tier than its database entries warrant. The tiering system creates a fixed reference point that anchors the entire classification process.

Implementing a Two-Pass Review Method

The two-pass review method addresses the problem of confirmation bias during conflict resolution. In the first pass, the adjudicator evaluates each database entry independently without considering the other entries. This pass produces a preliminary classification for each entry based solely on that entry's evidence quality and content. The first pass should be documented in a standardized format that records the entry identifier, the evidence tier, the preliminary classification, and the specific evidence that supports that classification.

The second pass compares the preliminary classifications across all entries and identifies the points of agreement and disagreement. This pass focuses on the evidence itself instead of the database labels. For example, if ClinVar contains a two-star conflicting submission and HGMD lists the variant as disease-causing, the second pass examines whether the underlying publications cited by each entry actually support the respective classifications. The second pass should also check whether the entries reference the same genomic position and nucleotide change, because different transcripts or genome builds can create apparent conflicts that do not actually exist.

The two-pass method is particularly valuable for variants with intermediate functional effects. The study of GT>GC splice-site variants demonstrated that approximately 15 to 18 percent of such substitutions retain variable levels of residual wild-type transcript, creating a functional gray zone that complicates classification. A single-pass review might accept a database classification without recognizing that the functional evidence supports an intermediate effect. The two-pass method forces the adjudicator to evaluate the functional data independently before integrating it with database classifications.

Creating a Conflict Resolution Decision Matrix

A decision matrix provides a structured way to resolve conflicts when database entries disagree. The matrix uses two axes: the evidence tier of the conflicting entries and the type of evidence that supports each classification. The matrix output is a recommended action that ranges from accepting the higher-tier classification to escalating the variant for professional review.

The matrix operates as follows. When Tier 1 entries conflict with Tier 2 or Tier 3 entries, the Tier 1 evidence takes precedence unless the primary literature reveals a specific flaw in the Tier 1 entry. When Tier 2 entries conflict with each other, the adjudicator must review the primary literature cited by each entry and determine which entry more accurately reflects the current evidence. When Tier 3 entries conflict with any other tier, the Tier 3 entries should be treated as hypothesis-generating instead of classification-determining.

The matrix also addresses the situation where entries within the same tier conflict. This situation requires a deeper review of the primary literature and may require consultation with a qualified professional. The Lebanese cohort study found that relying on ACMG/AMP criteria identified 16.8 percent of participants with actionable variants, while relying on ClinVar alone identified only 6 percent. This discrepancy demonstrates that same-tier conflicts can have substantial clinical implications, particularly for understudied populations where database representation is limited.

Applying Quantitative Functional Data to Resolve Conflicts

The GT>GC splice-site study provides a model for incorporating quantitative functional data into conflict resolution. The study used residual wild-type transcript as a quantitative functional readout and revisited disease-associated variants previously shown to retain substantial wild-type transcript, including SPINK1 c.194+2T>C, HBB c.315+2T>C, and BRCA2 c.8331+2T>C. These examples illustrate how intermediate splicing effects complicate clinical interpretation across distinct genes and disease contexts.

When database entries conflict for a variant with available quantitative functional data, the adjudicator should record the specific functional measurement and compare it against the thresholds used in the ACMG/AMP framework. The study found that computational prediction and experimental assessment both have limitations in capturing the continuum of variant effects. SpliceAI delta donor-loss scores, while useful, cannot fully predict the functional consequences of variants with intermediate effects. The adjudicator should therefore treat quantitative functional data as complementary to computational predictions instead of as a replacement for them.

The study also performed a locus-wide assessment of all 26 theoretically possible GT>GC substitutions in CFTR, integrating SpliceAI scores with classifications from expert-curated databases. Minigene splicing analyses of four selected CFTR variants, together with full-length and minigene analyses of a BAP1 GT>GC variant with conflicting clinical interpretations, revealed heterogeneous and context-dependent splicing outcomes. These findings underscore both inter-assay variability and the inherent limitations of commonly used splicing assay systems. The adjudicator should document which assay system produced the functional data and whether the results were consistent across multiple assay types.

Establishing a Variant Conflict Log

A variant conflict log provides the documentation structure needed for reproducible conflict resolution. The log should record the variant identifier, the date of the conflict assessment, the database versions queried, the evidence tier assigned to each entry, the preliminary classification from the first pass, the conflict resolution outcome from the second pass, and the final classification recommendation. The log should also record any functional data reviewed and the specific measurements that informed the classification.

The conflict log serves multiple purposes. It provides a basis for revisiting classifications when new evidence emerges. It supports quality control by allowing supervisors to review the adjudication process. It creates a training resource for new variant scientists who need to understand how conflicts are resolved in practice. The Carpentries lessons on data management and reproducibility provide foundational practices for organizing and documenting analysis workflows that apply directly to maintaining a variant conflict log.

The log should be maintained in a version-controlled format that allows tracking of changes over time. The nf-core documentation describes community standards for reproducible bioinformatics pipelines, including version control, containerization, and documentation practices. These standards apply to variant annotation workflows and ensure that classifications can be traced to specific database versions and retrieval dates.

Conducting a Structured Literature Re-Review

When database entries conflict and the primary literature is ambiguous, a structured literature re-review is required. This re-review differs from an initial literature search because it focuses specifically on resolving the identified conflict instead of discovering all available evidence. The re-review should follow a defined protocol that includes searching for the variant by genomic position, gene name, and nucleotide change across multiple databases and publication repositories.

The re-review should prioritize recent publications because classification standards have evolved over time. The accuracy study found that both ClinVar and HGMD have improved over time, but the improvement has been uneven. ClinVar variants are reclassified more frequently than HGMD variants, which likely reflects the different submission and curation models. A publication from 2010 may not reflect current understanding of variant effects, particularly for genes where functional assays have advanced substantially.

The re-review should also assess whether the variant has been reported in understudied populations. The Lebanese cohort study found that many ACMG-classified pathogenic variants were absent from ClinVar, complicating reporting due to a lack of evidence linking them to disease in other individuals. The study emphasized the urgent need to harmonize variant databases and update ClinVar entries, particularly for understudied populations. When a variant is identified in an understudied population, the adjudicator should consider whether the absence of database entries reflects a true lack of evidence or a gap in database representation.

Defining Escalation Criteria for Unresolved Conflicts

Some conflicts cannot be resolved through the tiering system, the two-pass method, or the decision matrix. These unresolved conflicts require escalation to a qualified professional. The escalation criteria should be defined in advance so that the adjudicator knows when to stop the resolution process and seek consultation.

Escalation is appropriate when the conflict involves a variant that would change clinical management. Escalation is also appropriate when the functional data show intermediate effects that do not clearly support either pathogenic or benign classification. The GT>GC splice-site study found that variants capable of generating appreciable residual wild-type transcript exemplify a broader class of intermediate-effect alleles that expose the limitations of both computational prediction and experimental assessment. These variants require professional judgment that goes beyond the structured protocol.

Escalation is also appropriate when the conflict involves a variant in a gene with limited published evidence. The Noonan syndrome study identified variants in PTPN11, LZTR1, SOS1, and RAF1, with PTPN11 being the most frequently affected gene. Variants in less frequently affected genes may have limited database representation and require professional review to determine whether the available evidence supports classification.

The escalation process should include documentation of the evidence reviewed, the criteria applied, and the rationale for the classification. The process should also include a mechanism for revisiting classifications when new evidence emerges. The taxonomic schema of potential pitfalls in clinical variant analysis highlights the range of errors that can occur when databases are used without adequate scrutiny. A structured escalation process reduces the risk of these errors by ensuring that unresolved conflicts receive appropriate professional attention.

Measuring Protocol Effectiveness Through Reclassification Tracking

The effectiveness of a conflict resolution protocol should be measured through reclassification tracking. This measurement involves recording the initial classification, the date of the classification, and any subsequent reclassifications that occur when new evidence emerges. The accuracy study found that ClinVar variants classified as pathogenic or likely pathogenic are reclassified sixfold more often than HGMD disease-causing variants, which has likely resulted in ClinVar's lower false-positive rate.

Reclassification tracking provides a quantitative measure of protocol performance. A protocol that produces classifications that are frequently reclassified may be overrelying on database labels without adequate primary literature review. A protocol that produces stable classifications may be appropriately weighting the available evidence. The tracking data should be reviewed periodically to identify patterns that suggest protocol adjustments.

The tracking data should also be used to assess whether the protocol is addressing the specific failure patterns described earlier in this article. These patterns include overreliance on database labels, ignoring population frequency data, applying PP5 without independent verification, and failing to recognize intermediate effects. The reclassification tracking data can reveal whether these failure patterns are being corrected by the protocol.

Integrating the Protocol with Existing Laboratory Procedures

The conflict resolution protocol should be integrated with existing laboratory procedures instead of treated as a separate process. The integration should occur at the point where database entries are collected and assessed. The tiering system should be applied during the initial data collection step, and the two-pass review should be applied before ACMG/AMP scoring begins.

The integration should also address the documentation requirements described earlier in this article. The variant evidence log should include the tier assignments and the conflict resolution outcomes. The classification rationale should explain how the conflict was resolved and why certain evidence was prioritized. This integration ensures that the protocol is not an add-on but a core component of the variant classification workflow.

The European Bioinformatics Institute provides training materials on using biological data resources effectively. These materials emphasize the importance of understanding data provenance and quality indicators when interpreting database entries. The Galaxy Training Network provides accessible tutorials on variant analysis workflows that include steps for annotating variants with database information and reviewing supporting evidence. These training resources support the implementation of a structured conflict resolution protocol by building the skills needed to execute each step effectively.

Frequently Asked Questions

How do I determine which database entry to trust when ClinVar and HGMD conflict?

Resolve conflicts by reviewing the primary literature and applying ACMG/AMP criteria systematically. ClinVar entries with multiple consistent submitters and high review status carry more weight than HGMD entries based on single literature reports. Check the date of each entry and consider whether the classification reflects current standards. Population frequency data and functional studies should be prioritized over database labels.

What does the ClinVar star rating actually tell me about variant classification confidence?

The star rating indicates the review status of the entry. Zero stars means no assertion criteria were provided, one star means a single submitter with assertion criteria, two stars means multiple submitters with conflicting interpretations, three stars means multiple submitters with consistent classifications, and four stars means classification from a professional guidelines body. The star rating guides the order of evidence review but does not replace direct evaluation of the underlying evidence.

When should I apply PP5 versus PS4 in ACMG/AMP scoring?

PS4 applies when the variant is present in affected individuals at a frequency significantly increased over controls, based on direct case-control evidence. PP5 applies when a reputable source reports the variant as pathogenic but does not provide sufficient evidence for a stronger criterion. PP5 should be used sparingly and only when the database entry cannot be independently verified. When database entries conflict, PP5 should not be used to resolve the conflict.

How do I handle variants with intermediate functional effects that are neither fully disruptive nor neutral?

Variants with intermediate effects require careful interpretation. The GT>GC splice-site study found that approximately 15 to 18 percent of such substitutions retain variable levels of residual wild-type transcript, and that computational prediction and experimental assessment both have limitations. Document the functional data, consider the disease mechanism and threshold for clinical significance, and escalate uncertain classifications to qualified professionals.

Why do ClinVar and HGMD sometimes disagree on the same variant?

The databases use different submission and curation models. ClinVar aggregates submissions from multiple laboratories with review status indicators, while HGMD applies professional curation to published literature. The accuracy study found that ClinVar variants are reclassified more frequently than HGMD variants, which likely contributes to ClinVar's lower false-positive rate. The databases also have different inclusion thresholds and update cycles.

How should I account for population frequency data when interpreting database entries?

Population frequency data should be checked before accepting a pathogenic classification from any database. Variants that are common in general populations are unlikely to cause highly penetrant Mendelian disorders. Consider the ancestry composition of the control database, because a variant that is rare in one population may be common in another. The accuracy study found that removing common variants reduced false-positive rates and eliminated ancestry-related bias.

What documentation should I maintain for variant classification?

Maintain a variant evidence log that records the database version and retrieval date for each entry, the classification and review status, the submitter or curator, and the evidence cited. Document which ACMG/AMP criteria were applied and how each criterion was weighted. Explain how database conflicts were resolved and why certain evidence was prioritized. This documentation supports reproducibility and provides a basis for revisiting classifications when new evidence emerges.

When should I escalate a variant classification to a qualified professional?

Escalate when database entries conflict, when the variant has not been reported in the literature, when functional studies show intermediate effects, when the classification would change clinical management, or when the evidence base is insufficient to support a confident classification. The escalation process should include documentation of the evidence reviewed, the criteria applied, and the rationale for the classification.

Related Bioinformatics Guides

Related Clinical & Scientific Guides

References and Further Reading

NCBI Data Resources. National Center for Biotechnology Information.
EMBL-EBI Training. European Bioinformatics Institute.
Bioconductor. Bioconductor Project.
Galaxy Training Network. Galaxy Project.
nf-core Documentation. nf-core.
The Carpentries Lessons. The Carpentries.
When splicing is not all or none: GT&gt,GC 5' splice-site variants as a model for intermediate effects and challenges in variant classification.. 2026.
Certain vs. uncertain actionable secondary findings in a cohort of 500 Lebanese participants: What to report to the patient?. 2025.
Molecular Diagnosis and Phenotypic Variability of Noonan Syndrome: Experience from a Romanian Multicenter Study.. 2026.
A taxonomic schema of potential pitfalls in clinical variant analysis based on real-world evidence.. 2023.
ClinVar and HGMD genomic variant classification accuracy has improved over time, as measured by implied disease burden.. 2023.

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.