Population reference datasets help researchers ask whether a variant is compatible with the frequency expected for a disease and whether a gene shows evidence of selective constraint. These are powerful filters, but they are not direct diagnostic tests. A rare variant can be harmless, and a gene-level metric does not establish the effect of every variant within that gene.
Read constraint at its intended level
The 2020 gnomAD study quantified patterns of variation across a large human dataset and developed measures of constraint, including depletion of predicted loss-of-function variation. These metrics describe aggregate evidence about genes and are affected by statistical power. A low tolerance for one class of variation does not mean that any missense change in the gene causes the same disease. Read the gnomAD constraint study.
Distinguish a gene’s biological importance from a proven clinical association. Constraint can prioritize a research hypothesis, but inheritance, disease mechanism and variant-level evidence still need evaluation. Conversely, limited evidence of constraint may reflect insufficient information or a mechanism not captured by the metric.
Interpret frequency with the denominator
gnomAD’s explanation of genetic ancestry distinguishes ancestry estimates from race and ethnicity and describes the value of examining group-specific as well as overall frequencies. A variant diluted in a global average may be more common in a particular ancestry group. These groupings are analytic tools with limitations, not discrete biological categories that perfectly describe individuals. Read gnomAD’s ancestry explanation.
When a variant is absent, check whether the site was adequately covered and how many relevant alleles were assessed. Underrepresentation can make an otherwise ordinary variant appear unusually rare. Database version, filtering and sample composition should be recorded so a later reviewer can reproduce the observation.
Turn database observations into bounded evidence
- Specify the variant representation and reference assembly.
- Check quality flags and the number of alleles observed at the site.
- Examine relevant frequency summaries without substituting racial categories for genetic analysis.
- Keep gene-level prioritization separate from variant classification.
The defensible conclusion is often narrower than the tempting one: a frequency may be inconsistent with a proposed highly penetrant disorder, or absence may supply limited supportive information. State that inference explicitly. Reference data are strongest when they constrain a hypothesis rather than being treated as an automatic verdict.