About this site
This site was built by a parent after their child was diagnosed with PRPH2 disease, out of publicly available research, foundation materials, structural data, and materials and talks from the 2026 PRPH2 workshop. It is independent, carries no advertising, and is not affiliated with any foundation, laboratory or company named on it.
A PRPH2 result can arrive with very little explanation, and the research is scattered across papers written for specialists. The hope is that this site gathers it in one place, in plain language, so that you can come to your next appointment with a clearer picture and better questions.
What it is for
It has two aims. The first is to explain the biology of PRPH2 disease clearly enough to be useful in a conversation with an eye doctor or genetic counselor: what the protein does, why one faulty copy can sometimes be worse than none, why the same variant can affect relatives differently, and where treatment research stands today. The second is to let anyone with a PRPH2 variant on a genetic report see what is publicly known about it, all in one place.
What it cannot do
This site is meant to sit alongside your care team, not to replace them. A few limits are worth knowing up front.
- It does not diagnose or classify variants. Deciding whether a variant causes disease is a clinical judgment made by a genetics laboratory, which weighs laboratory studies, family information, population frequency and prediction tools together. This site shows you the public pieces of that picture so the conversation is easier to follow; the decision belongs to your laboratory and specialist.
- It cannot tell you what will happen to any one person. PRPH2 disease varies a great deal, even within a family, and some people who carry a variant never develop symptoms. Statements about outlook on this site describe groups of patients, not individuals.
- The computer prediction scores are research tools. AlphaGenome's terms of use do not allow clinical use, and AlphaMissense is licensed for non-commercial use only. They are here to help explain, not to decide.
- The literature search is automated. It looks for PRPH2 papers that name the variant, then checks each paper's abstract and, where it is freely available, its full text, and shows you the passage or table row where it found the name. It can still miss papers, especially ones that list variants only in supplementary files. The optional AI summary reads the abstracts and those passages rather than whole papers, and it can misread them. Please check anything you plan to act on against the papers themselves and with your care team.
- It does not keep anything about you. Variant lookups run in your own browser. Only the variant name is sent to this site's server, for the literature search and summary, and those results are stored by variant, not by visitor.
Data sources and how each number is worked out
For anyone who wants to check the working, here is where every number comes from.
| Source | What it provides | Retrieved / version |
|---|---|---|
| UniProt P23942 | Protein sequence, topology (transmembrane segments, loops), glycosylation and disulfide annotations | 2026-09-16 |
| Ensembl ENST00000230381 (MANE Select, NM_000322.5) | Exon coordinates on GRCh38, used to translate every possible coding single-nucleotide change and to name variants in c. notation. Verified by reconstructing the coding sequence and translating it back to the UniProt protein. | 2026-09-16 |
| NCBI ClinVar | Every record for PRPH2: classification, review status, conditions, coordinates | 2026-09-16 |
| gnomAD v4 | Allele counts and totals for every PRPH2 variant seen in ~800,000 people, exomes and genomes combined | 2026-09-16 |
| AlphaMissense (Cheng et al., Science 2023) | Pathogenicity score for all 6,574 single amino-acid substitutions. Thresholds 0.34 / 0.564 are the authors' calibrated cutoffs. | Zenodo release, CC BY-NC-SA 4.0 |
| ESM1b (Brandes et al., Nature Genetics 2023) | Log-likelihood ratio for every substitution from a 650M-parameter protein language model. −7.5 is the paper's damaging threshold. | ntranoslab precomputed release |
| AlphaGenome Atlas (Google DeepMind) | Variant Impact (AVI) Phred score for every single-nucleotide change in the three coding exons and ±12 bp of each splice junction: 3,339 variants. Not available for changes that need more than one nucleotide. | AlphaGenome Atlas AVI, GENCODE v46, scanned 2026-09-09 (exons 1-2) and 2026-09-16 (exon 3 + splice flanks) |
| PDB 7ZW1 (El Mazouni & Gros 2022) | 3.7 Å cryo-EM structure of the PRPH2–ROM1 heterodimer. Per-residue relative solvent accessibility (Shrake–Rupley on chain A alone), buried area at the ROM1 interface (difference with and without chain B), nearest-atom distance to ROM1, and distance to the nearest D2-loop cysteine sulfur. Residues 5–321 are resolved. | RCSB download 2026-09-16 |
| AlphaFold DB AF-P23942-F1 v6 | Per-residue pLDDT confidence; the model itself for the second 3D view | 2026-09-16 |
| Europe PMC and NCBI LitVar2 | Live literature search by variant name (protein, cDNA, rsID), limited to papers that mention PRPH2. Each paper's abstract and open-access full text are checked for the variant name, and papers are ranked by where it appears, citations and recency. Cached seven days. | live |
| ClinicalTrials.gov API v2 | Studies mentioning PRPH2 | 2026-09-16 |
The percentile bars
"More damaging than X% of all possible PRPH2 missense changes" ranks the variant's score against every other single amino-acid substitution in the same protein under the same model. It is a within-gene rank, not a probability of disease.
The mechanism class
Assigned by rule from the type of change and its position: nonsense and frameshift changes are called predicted loss-of-function; canonical splice-site changes are flagged as such; missense changes inside residues 124–264 are called D2-loop missense; other missense changes are labeled by domain. The class summarizes what cohort studies report for that category and is explicitly not a per-variant claim.
The curated notes
About thirty well-studied variants carry hand-written notes drawn from the primary literature (cohort papers, functional studies, the 2023 and 2026 PRPH2 workshops). Where a PMID is known it is linked.
Accessibility
The site is designed with readers with low vision in mind first. Base text is 18 px and can be scaled to 156% from the Display menu (independently of browser zoom, which also works). There is a dark theme, a light theme, and a high-contrast mode; a reduce-motion setting; and full keyboard operation of every interactive figure, including the protein map (arrow keys move along the protein) and the heatmap (which also has a text alternative in the result cards). Every chart's numbers are available as text. The typeface is Atkinson Hyperlegible Next, designed by the Braille Institute to keep similar letterforms distinguishable.
Corrections
This is one person's attempt to summarize a field that changes. Corrections or feedback are welcome at admin@prph2.site.
Licensing
Original text and figures on this site are CC BY 4.0. Third-party data keep their own licenses: AlphaMissense is CC BY-NC-SA 4.0; AlphaFold DB is CC BY 4.0; ClinVar and gnomAD are public domain or open; PDB data are free to use; AlphaGenome outputs are subject to Google DeepMind's terms.