Retinopathies Curation Project

Overview

Curate retinopathy and retinal dystrophy branches for the dismech knowledge base with explicit granularity rules. The goal is to avoid duplicated pathographs while still separating entities whose proximal pathobiology, diagnosis, or management genuinely diverge.

This project treats "retinopathy" as a mixed-use term. In some contexts it is a phenotype label within a broader systemic disease. In others it denotes a retinal disease family with a shared mechanistic trunk and clinically meaningful subtypes. The project therefore starts from mechanism-first curation rather than from ontology depth alone.

Current Tension

Retinal diseases create a recurring lump-vs-split problem:

1. Broad umbrella terms such as retinopathy or inherited retinal dystrophy are often too heterogeneous for a single disease-level pathograph.
2. Gene-based MONDO descendants can be clinically useful, but creating one file per descendant often duplicates most of the mechanistic graph.
3. Some gene-centered families are coherent enough for a shared parent entry, but only if the child terms diverge late from a common proximal module.

Working Granularity Principles

1. Do not create one dismech file per MONDO term by default. MONDO is multi-level; dismech curation should be guided by mechanism and modeling utility, not by ontology descent alone.

2. Do not treat generic retinopathy as a default top-level disease entry. In many contexts it is better modeled as a phenotype within a broader disorder.

3. Use a disease-family umbrella when child entities share a strong proximal mechanism. A family entry is justified when most descendants can reuse the same early pathograph trunk and differ mainly in retinal cell-type bias, topography, inheritance mode, or downstream clinical pattern.

4. Split into separate top-level entries when the shared pathograph would be too thin. Separate entries are favored when proximal pathobiology, diagnostic framing, prognosis, or treatment strategy materially differ.

5. Subtype labels are not the same as severity labels. Different retinal phenotypes may reflect rod-predominant, cone/macular- predominant, early severe, or syndromic branches of the same module rather than mere differences in degree.

6. Digenic or inheritance-specific branches can stay under a shared parent. A second-gene interaction or dominant/recessive split can be modeled as a meaningful subtype when the affected molecular complex or pathway is still the same.

7. Treat MONDO placement as informative, not decisive. Some ontology descendants may need special review if their canonical literature points to a different proximal mechanism than the parent family would imply.

Initial Findings

Broad project-level conclusions

• Retinopathy is already used in the current KB mostly as a phenotype label inside broader disorders rather than as a standalone disease.
• Inherited retinal dystrophy is too broad to serve as a default disease-level dismech entry.
• Gene-centered retinal families are often more useful than phenotype-numbered legacy labels when they capture a coherent proximal module.

PRPH2-related retinopathy

PRPH2-related retinopathy is an initial positive test case for a shared umbrella entry.

Why it looks coherent:

• PRPH2/peripherin-2 is a photoreceptor outer-segment rim/disc protein.
• PRPH2 also functions in a structural complex with ROM1.
• Many descendant diseases appear to share a common proximal trunk involving altered PRPH2 dosage, folding/trafficking, oligomerization, or PRPH2-ROM1 complex assembly.
• Descendant terms such as retinitis pigmentosa 7, vitelliform macular dystrophy 3, patterned macular dystrophy 1, central areolar choroidal dystrophy 2, and Leber congenital amaurosis 18 look more like downstream rod-biased, cone/macular-biased, or early severe branches than entirely separate molecular pathways.

Implication for modeling:

• The shared trunk should live in the parent entry.
• Child terms should branch later in the graph where retinal cell type, topography, onset, severity, or inheritance context diverge.
• retinitis pigmentosa 7, digenic should likely be retained as a subtype or explicit genetic-context branch rather than as a fully separate root entry, because the additional ROM1 contribution still acts within the same structural module.

Caution:

• Terms such as fundus albipunctatus and retinitis punctata albescens require extra review before being folded into a PRPH2 family pathograph, because their canonical literature often points to visual-cycle mechanisms in other gene families.

Candidate Curation Strategy

Family umbrellas worth evaluating

Likely standalone retinal disease entries

Concepts to keep as parents / project labels, not default disease files

• retinopathy
• inherited retinal dystrophy
• hereditary macular dystrophy

Existing Related Entries in KB

STATUS

Principles and Scoping

• [x] Established a working granularity framework for retinal disease families
• [x] Distinguished phenotype-level retinopathy use from disease-family use
• [x] Identified PRPH2-related retinopathy as a candidate shared-mechanism umbrella
• [x] Flagged fundus albipunctatus / retinitis punctata albescens as caution terms for blind inheritance from ontology structure

Initial Curation Targets

• [ ] Curate PRPH2-Related_Retinopathy
• [ ] Review RPGR-related retinopathy branch for family-entry viability
• [ ] Review RPE65-related recessive retinopathy for standalone-entry viability
• [ ] Review RDH5 and RLBP1 branches with attention to visual-cycle mechanisms

NOTES

2026-03-18

• Added a retinopathy-specific project note to capture general lump-vs-split findings before individual disorder curation.
• Recorded the working rule that shared-mechanism retinal families may justify a parent entry, but ontology descendants should not automatically become top-level dismech files.
• Documented PRPH2-related retinopathy as the first detailed test case for a shared-trunk retinal family entry.