Which table decided that phenotype?

The notation is consistent, the strand has been checked, and the genotype is finally unambiguous. The next question is the obvious one: what does it mean?

This is the easiest place in the whole process to lose your footing.

The phenotype is not in the data

What the VCF holds is bases. It says G/A. It does not say anything beyond that, anywhere in the file.

The step from G/A to a named type is made by a table that someone else built and published. The phenotype does not emerge from reading the data; it emerges from laying a table over the data.

That ordering may look backwards, but it is what actually happens — and keeping it in view changes what you are able to write down afterwards.

Do not build the backbone yourself

Invent your own cut-offs and you inherit the whole burden of defending them. Why that boundary, why that weighting — with nothing published to point at, every answer has to come from you.

So the backbone of any genotype-to-phenotype call is borrowed from an external published table, and the table is cited alongside the result. No scoring scheme of your own.

That is also where this site stops. Showing what a table defines is one thing; assessing a person’s health is another, and we do not do the second.

The backbone comes in four layers

There is no single all-purpose table. There are layers with different jobs, stacked on top of one another. Mix them up and you can no longer tell how strong your grounds are.

Layer	What it settles	Commonly used resources
Name	What to call this one position	dbSNP, HGVS notation, GA4GH VRS
Frequency	How rare it is in a population	gnomAD, jMorp (ToMMo), TogoVar
Significance	How clinical significance is recorded	ClinVar, ClinGen, the ACMG/AMP framework
Phenotype	Which label a genotype or diplotype maps to	PharmVar allele definitions, CPIC-derived terms

The bottom layer is the backbone proper. It is the only layer that actually defines the mapping from genotype to label; the three above it supply the material that mapping rests on.

Wording like fast or slow enzyme activity carries a fixed meaning rather than a loose impression precisely because that layer has agreed terms behind it — terms that were settled deliberately, as a documented expert consensus (Caudle et al., 2017).

What is moving right now

The backbone is not a stone tablet. The tables themselves have been shifting for several years, and notes written as though they were fixed become unreadable within a few years.

Resources have been consolidated. Since late July 2025, links to PharmGKB redirect to ClinPGx, which now brings CPIC, PharmVar and PharmCAT under one roof. Copying a URL into your notes is no longer enough to record which document, at which version, you actually read.

Allele definitions get revised. Haplotype (star allele) definitions continue to be updated. When a definition changes, the same raw data, run through the same procedure, can come back with a different label. Published guidelines for drug-metabolising enzymes were still being revised through 2026.

Frequency databases differ by version. gnomAD versions differ both in the samples aggregated and in the reference build — the older releases are on GRCh37, the v4 series on GRCh38. “I looked it up in gnomAD” is not enough information.

Computational predictors are evidence, not verdicts. Rather than taking a predictor’s output as the answer, there is now a published method for calibrating tool scores against the strength of evidence they justify (Pejaver et al., 2022). What a tool returns is one input to a judgement.

Japanese population frequencies are a separate lookup. For some variants, frequency shifts by an order of magnitude between populations. ToMMo’s jMorp publishes frequencies derived from tens of thousands of Japanese whole genomes, and TogoVar aggregates Japanese frequencies so they can be read side by side with other populations. Calling something rare on the strength of a global figure alone is premature.

One note of caution. Every framework listed here was built for clinical and research settings. None of it transfers unaltered to a person reading their own data — and anything to do with medicines belongs with a physician or pharmacist, not with whoever can look up a table. What a table returns is the single fact that this table defines that mapping.

When it will not resolve, do not resolve it

Haplotypes, star alleles, differences in copy number. These do not reduce cleanly to a single value. Two variants sitting close together and inherited as a unit are another case: count them separately and you have counted one fact twice.

In situations like these, do not force the result into “high” or “low.” Leaving the fact that it will not resolve, and noting what exactly will not resolve, carries far less chance of being read wrongly.

Which means the categories have to be built with somewhere to put it. A classification with no slot for “will not resolve” and “not readable yet” lies somewhere, always. A table with unfilled cells is more accurate than one where every cell is full.

What to keep when you look something up

It looks laborious. It is five fields.

1. The table (which resource)
2. Its version, or the date you consulted it
3. The input (rsID and genotype, or the diplotype)
4. The output (the label the table returned)
5. Why you chose that table

The first three write themselves. The fifth is the one that separates a record from a note-to-self. If you cannot state the reason, you have not yet decided whether that table was the right one to consult.

It changes how you write

Once you accept that the backbone sits outside your own work, the sentences change shape on their own.

• ✗ “You are the such-and-such type.”
• ✓ “In the such-and-such table, at this version, this genotype maps to such-and-such.”

The second is longer, and it survives a revision of the table. When the mapping changes, you add a line saying so. The first ages together with whatever table you happened to read that day.

Would this still make sense to you in six months? That question turns out to settle most of the others.

If something about your health is on your mind, please take it to a medical professional, separately from any of this.