Read your report
What a percentile actually means (and what it does not)
A percentile tells you where you rank, not what will happen to you. It is a position in a crowd, not a probability.
The claim. My percentile is my chance of getting the disease: a 90th-percentile score means about a 90 percent risk, and a low percentile means I am in the clear.
Verdict. A percentile is a rank, where you sit in a reference group, not a probability. On its own it does not say whether you will develop a condition, and the rank is built against a European-ancestry reference, so it is materially less accurate if your ancestry differs.
A percentile is a rank. It places a single number, here a polygenic score, on a reference distribution built from many other people, and reports the share of that reference group who scored lower. A result at the 90th percentile sat above roughly nine in ten of the reference group. It does not mean a ninety percent chance of anything. The rank says where a score falls among others, not how likely the condition is, which depends on the rest of the genome, the environment and chance that the score never captures.
A percentile is a rank, not a risk
A percentile orders one result against a reference group of many people and reports how much of that group scored lower. A score at the 60th percentile sat above roughly sixty out of a hundred in the reference set. The number is a position, like a finishing place in a long race, not a measurement of what will happen to the runner.
This is the single point the word most often loses. A high percentile is not a high probability of disease, and a low percentile is not a clean bill of health. The rank tells you where a score falls among others. On its own it is silent on whether the condition ever appears.
Where the rank comes from
The report describes a polygenic score as a sum of many small-effect variants combined into a single statistical tendency. The engine counts how many copies of each scoring variant you carry, weights them, adds them up, and places that total on the score's reference curve. The percentile is simply that placement expressed as a rank.
Because it is a tendency drawn from many tiny contributions, the report states plainly that a polygenic score is not a measurement, not a prediction of the trait itself, and not a diagnosis. The percentile inherits all of those limits. It is a way of reading the score, not a separate, firmer fact.
What the rank is measured against
A percentile only means something relative to the group it is compared with. Every score in the report declares a reference population, and the percentile value is tagged with it, for example as a European reference. The report carries this caveat in its own words: the percentile is computed against a European-ancestry reference distribution, polygenic scores travel poorly across ancestries, and if your genetic ancestry differs the percentile is materially less accurate.
Research on polygenic scores reports the same limit. Scores derived mainly in European-ancestry cohorts lose accuracy when applied to people of other ancestries, which is why the report shows the rank to everyone with the caveat attached rather than hiding it. A rank against the wrong yardstick is still a number, just a less trustworthy one.
Coverage changes how precise the rank is
The percentile also depends on how much of the score your file could read. The report shows a 'Markers used' line, such as a count of how many of the score's scoring variants your file covered, and treats any variant it could not see as the common reference genotype. When coverage is partial the report calls the percentile an estimate, good for a tendency rather than a precise placement, and points to a gVCF or genotyping-chip file for a sharper rank.
When too few of the scoring variants are present, the report does not guess. It marks the score 'Not reportable' and states that this is a coverage gap, not a result. The same person can sit at a slightly different percentile depending only on which file type they uploaded, which is another reason the percentile reads as an estimate rather than a verdict.
A worked example in the sample report
The public sample report at /sample shows how this looks in practice. Its polygenic scores sit behind an opt-in, each one revealed only on a 'Reveal my score' click, and the banner states the page is shown for illustration, is not medical advice, and is not a real genome. Revealing a score shows the percentile chip, the reference tag and the markers-used line together, with the not-a-diagnosis caveat below.
Reading the sample is the clearest way to see the discipline: the percentile appears as a placed rank with its reference and its coverage stated next to it, never as a probability of disease. Aimosti restates where a score falls and names the limits. It does not convert that position into a personal risk it cannot support.
What a percentile is, and is not, telling you
Read plainly, a percentile tells you your relative position on one particular score, against one declared reference group, given the variants your file could read. That is genuinely informative as a tendency, and it is the real ceiling of what the number supports.
It is not telling you your chance of developing the condition, it does not fold in your environment or plain chance, and it is less accurate when your ancestry differs from the reference. A percentile is a rank and a risk factor to weigh, not a verdict to act on alone.
What Aimosti would (and wouldn't) show you
In the report a polygenic score carries an ordinal percentile chip, such as a chip reading the Nth percentile, alongside an evidence line that names the reference, for example 'Percentile NN.N, European reference', and a 'Markers used' line giving how many of the score's variants your file covered. When coverage is partial the report labels the percentile an estimate rather than a precise rank, and when too few variants are present it reads 'Not reportable' and calls it a coverage gap, not a result.
What we won't claim
We won't tell you that a percentile is your probability of developing a condition, that a high rank is a diagnosis, or that a low rank means you are safe. We won't present a rank built against a European-ancestry reference as equally accurate for everyone, and we won't hide that caveat to make the number look cleaner than it is.
Bottom line. A percentile is a rank, where you sit in a reference group, not a probability. On its own it does not say whether you will develop a condition, and the rank is built against a European-ancestry reference, so it is materially less accurate if your ancestry differs.
Related: Polygenic scores (reveal-gated, opt-in). Restated from: PGS Catalog (Polygenic Score Catalog) · NHGRI: Polygenic Risk Scores · Martin et al., Clinical use of current polygenic risk scores may exacerbate health disparities, Nature Genetics 2019.