A genetic test can find a difference in DNA within minutes. Deciding what that difference means can take far longer. A laboratory must connect the exact variant with a gene, a biological mechanism, a condition, an inheritance pattern, and a body of evidence that may be incomplete or contradictory.
For germline sequence variants associated with Mendelian disorders, the widely used ACMG and AMP framework has five labels: pathogenic, likely pathogenic, uncertain significance, likely benign, and benign. The label summarizes evidence. It does not replace the diagnosis, predict every person's outcome, or remain fixed when knowledge changes.
First, identify the exact variant#
Classification begins with accurate naming. The report should specify the reference sequence, gene transcript, genomic or coding change, and protein consequence when applicable. The same shorthand can be ambiguous if a transcript or genome build is missing.
The laboratory also asks whether the test detected the change reliably. Read depth, allele balance, mapping quality, pseudogenes, repetitive regions, mosaicism, and the technology used can affect confidence. A classification framework cannot rescue a false variant call.
Next comes the disease context. A DNA change may have different implications in different conditions, inheritance models, or tissues. Germline pathogenicity is not the same as somatic clinical impact in a tumor. A variant can also have reduced penetrance, meaning not everyone who carries it develops the associated condition.
Five labels, not five certainties#
The ACMG and AMP terms are standardized categories for evidence interpretation:
- Pathogenic means the evidence supports a disease-causing role under the stated model.
- Likely pathogenic means the evidence strongly favors a disease-causing role but does not reach the laboratory's pathogenic threshold.
- Uncertain significance means the evidence is insufficient or conflicting.
- Likely benign means the evidence strongly favors no disease-causing role under the stated model.
- Benign means the evidence supports a non-disease-causing interpretation.
“Likely” is not a description of symptom severity. “Pathogenic” does not mean that disease is inevitable, because penetrance, age, other genes, environment, and clinical context can alter outcome. “Benign” is also bounded by the condition and mechanism evaluated.
The framework does not turn every observation into a diagnosis. A clinical team still asks whether the person's features fit the condition, whether the inheritance pattern fits the family, and whether another explanation is more plausible.
Population frequency can rule against rarity#
A variant that is common in a population is unlikely to cause a very rare, highly penetrant dominant disorder. Population databases therefore provide important benign evidence when allele frequency exceeds what the disease prevalence, inheritance, and penetrance could support.
This reasoning needs care. Populations are not interchangeable, and a database may underrepresent some ancestries. A founder variant can be relatively common in one group. Coverage and call quality differ by region. A recessive pathogenic variant can be carried by healthy people. Low frequency alone does not prove pathogenicity.
ClinGen has refined how frequency evidence is applied, including the strong benign criterion and the rarity criterion. Gene and disease specifications can set thresholds that are more defensible than a universal cutoff.
Molecular consequence depends on disease mechanism#
A premature stop, frameshift, or essential splice change may support pathogenicity when loss of function is an established mechanism for the gene and condition. The same kind of change may be uninformative if disease arises through a different mechanism, if the affected transcript is not biologically relevant, or if the change occurs near the gene's end and escapes the expected effect.
Missense variants, which replace one amino acid with another, require a different analysis. The position may lie in a critical functional domain. The same amino-acid change may have been observed before. Other changes at the same residue may help. Structural or conservation information can contribute.
Computational predictions are supporting evidence, not a vote count. Several tools may use overlapping training data and features, so treating ten concordant scores as ten independent experiments exaggerates the evidence. ClinGen guidance emphasizes calibration and criterion-specific use.
Patient and family evidence needs a denominator#
Cases can support pathogenicity when the variant appears in people whose features match a specific condition. The value depends on how cases were selected, how specific the phenotype is, whether alternative causes were examined, and whether the observations are independent.
De novo evidence means a variant arose in the tested person rather than being inherited from either biological parent. Its strength depends on confirmed parentage, the expected disease mechanism, and phenotype specificity. A new variant is not automatically pathogenic.
Segregation asks whether the variant tracks with the condition through a family. Co-segregation can add evidence, but small families provide limited information. Reduced penetrance, age-related onset, phenocopies, and uncertain diagnoses complicate the pattern. Related relatives cannot be counted as though every observation were independent.
For recessive disease, evidence may depend on whether two variants are on different copies of the gene, described as being in trans, and whether the second variant is already established as pathogenic.
Functional evidence is only as good as the assay#
A laboratory experiment can show that a variant alters protein abundance, enzyme activity, splicing, cellular localization, or another function. The key question is whether that assay measures a process relevant to the disease mechanism and whether its controls distinguish known pathogenic from known benign variants.
An abnormal result in an artificial system is not automatically strong pathogenic evidence. Overexpression, nonhuman cells, weak controls, or a readout far from the clinical mechanism can limit interpretation. A well-validated assay with predefined thresholds can carry more weight.
ClinGen provides guidance for applying functional evidence and for interpreting splicing. These specifications reduce variation between laboratories, but they do not remove expert judgment.
Combine evidence without double counting#
The original ACMG and AMP framework assigns evidence codes at very strong, strong, moderate, or supporting levels for pathogenicity, with parallel benign evidence. Rules then combine the codes into one of the five categories.
The hard part is dependence. A computational prediction and a functional study may reflect the same molecular assumption. Several published cases may describe overlapping families. A database assertion may be based on evidence already counted directly. Reusing one fact under several codes makes the conclusion look stronger than it is.
ClinGen has developed refinements, including a Bayesian point-based interpretation of the criteria and gene-specific Variant Curation Expert Panels. The practical aim is consistency and a transparent rationale, not automation without review.
The guide to what makes a biomarker useful explores a related principle: analytical detection, biological association, and clinical usefulness are different questions.
Why laboratories can disagree#
Two laboratories can examine the same variant and reach different classifications. One may have newer internal cases. One may use a gene-specific specification. They may interpret functional evidence differently, apply different frequency thresholds, or classify the variant for different conditions.
Dates matter. A 2018 classification may not include a large population database or a later expert-panel review. The underlying evidence can change without the DNA change itself changing.
ClinVar makes these differences visible by archiving submissions from laboratories, researchers, expert panels, and guideline groups. Its gold-star review status indicates the level of review supporting submitted classifications. It is not a simple product rating. NCBI notes that ClinVar does not curate or modify every submitted classification itself.
When submissions conflict, inspect the condition, date, evidence summary, assertion criteria, and review status. An expert-panel classification receives greater weight in ClinVar's aggregate display, but the current clinical laboratory report and the person's context remain essential.
A VUS is a boundary, not a diagnosis#
A variant of uncertain significance means the evidence cannot support a reliable pathogenic or benign classification. It does not mean a fifty-fifty chance of disease. It does not confirm the diagnosis. It also does not prove the variant is harmless.
The original ACMG and AMP guidance states that a VUS should not be used in clinical decision making. Clinical care can still be based on the person's symptoms, examination, family history, imaging, or other validated findings. The VUS itself should not be promoted into a positive result to fill an explanatory gap.
Testing unaffected relatives solely to see who carries a VUS may or may not add useful evidence, depending on the condition and family structure. A genetics professional and the testing laboratory can determine whether targeted family studies are appropriate and whether the laboratory offers them.
Reclassification and follow-up#
Classifications can change when new population data, cases, segregation, functional studies, disease mechanisms, or criteria appear. A VUS can move toward benign or pathogenic, and a prior classification can also be downgraded.
Reanalysis is not identical across laboratories. Policies differ on how often evidence is reviewed, whether the laboratory contacts the ordering clinician, and how updated reports are delivered. Contact details and care teams can change over time.
A practical report review records the laboratory, report date, exact variant, transcript, condition, classification, evidence summary, and follow-up policy. Questions about reanalysis should go to the ordering clinician, genetics service, or laboratory rather than being answered from an undated screenshot or a single database row.
The guide to hereditary cancer risk and BRCA testing shows how variant interpretation fits within a broader process of pretest probability, consent, and clinical management.
Reading the label responsibly#
Begin with the exact report, not the gene name alone. Confirm the condition and inheritance model. Note the classification date and laboratory. Read which evidence was used, which evidence conflicted, and whether a ClinGen expert panel has produced gene-specific guidance.
Then keep the label in its proper role. It is a structured evidence judgment that helps a qualified team interpret a test. It is not a stand-alone instruction for screening, treatment, or family testing.
References#
- ACMG and AMP standards for interpreting sequence variants
- ClinGen Variant Classification Guidance
- ClinGen Variant Curation Expert Panel procedures and resources
- NCBI ClinVar overview
- NCBI ClinVar review-status documentation
- Overview of gene and disease specifications for ACMG and AMP criteria
For your own health, talk with your clinician or a qualified genetics professional.*
Questions and answers
What are the five standard germline variant classes?
For variants in genes associated with Mendelian disorders, the ACMG and AMP framework uses pathogenic, likely pathogenic, uncertain significance, likely benign, and benign.
Does a variant of uncertain significance mean disease risk is fifty-fifty?
No. It means the available evidence does not support a reliable pathogenic or benign classification. It is not a numerical probability or a diagnosis.
Can two laboratories classify the same variant differently?
Yes. They may use evidence from different dates, apply criteria differently, evaluate different conditions or inheritance models, or use different gene-specific specifications.
Can a genetic variant be reclassified?
Yes. Population data, family observations, functional studies, case evidence, or updated criteria can change the evidence balance, so laboratories and expert panels may update classifications.
Should someone change care because a database lists a variant as pathogenic?
Not from a database label alone. The exact variant, condition, evidence, review status, laboratory report, personal and family context, and current professional guidance all matter.