Overview
What ODDB v1 represents
ODDB v1 is a relational research resource for ocular diseases and their molecular, variant, therapeutic and clinical evidence. The released model keeps biologically different entities separate and preserves the provenance needed to understand how each relationship was established.
Core entities
Current released entity layers
Counts below describe the deployed ODDB v1 enrichment release. Counts are captured from its database when this deployment image is built.
Disease
185 canonical diseases
Canonical ocular disease concepts used as the central disease layer for genetics, variants, therapeutics and clinical evidence.
Gene
1,249 genes
HGNC-aligned gene entities. Non-gene loci and regulatory labels are not treated as genes.
Variant
707,315 variants
Canonical variant records connected to genes and, where explicit evidence exists, to diseases.
Locus
2 structural loci
Non-gene genetic entities retained outside the HGNC gene layer to prevent contamination of gene comparisons.
Drug
893 canonical drugs
Canonical therapeutic entities with explicit drug-form relationships and separate disease and target evidence.
Clinical Trial
1,255 trials
Trial entities linked independently to diseases and drugs, avoiding unsupported disease×drug cross-products.
Clinical Report
1,382 reports
Clinical evidence report entities modeled separately from trials and linked through many-to-many disease/drug associations.
Interactive data model
How the released layers connect
Choose a layer or select a node. Pan and zoom the diagram on desktop or touch devices.
Relationships
Released association layers
Association counts are row counts in the deployed enrichment release unless otherwise stated.
| Relationship | Count | Meaning | Interpretation rule |
|---|---|---|---|
| Disease–Gene Association | 1,662 | Links a disease to a supported gene. | Stable association identity includes DiseaseID + HGNC_ID + inheritance context. |
| Gene–Variant Association | 707,315 | Links a variant to its associated gene. | Does not by itself establish disease relevance. |
| Disease–Variant Association | 106,838 | Explicit disease–variant evidence. | Never inferred solely from disease–gene membership. |
| Disease–Locus Association | 2 | Links disease to a non-gene locus. | Locus entities remain separate from HGNC genes. |
| Drug–Target Association | 1,009 | Molecular pharmacology between drug and gene target. | Target evidence is not equivalent to disease-specific therapeutic evidence. |
| Disease–Drug Association | 424 | Disease-specific evidence, including registered clinical investigation. Efficacy is not implied. | Kept separate from target-overlap discovery. |
| Disease–Drug Evidence | 1,299 | Evidence records supporting disease–drug context. | Evidence records should not be reduced to target overlap alone. |
| Drug Form Relationship | 99 | Connects parent drugs with salts/forms. | Drug forms are not silently collapsed into a parent entity. |
| Disease–Clinical Trial | 1,342 | Many-to-many disease links to trial entities. | A trial association does not prove efficacy or approval. |
| Drug–Clinical Trial | 1,566 | Many-to-many drug links to trial entities. | Do not cross-product trial diseases and drugs into asserted treatment pairs. |
| Disease–Clinical Report | 1,370 | Many-to-many disease links to clinical reports. | Source context is retained at report level. |
| Drug–Clinical Report | 1,575 | Many-to-many drug links to clinical reports. | Source drug IDs and names remain preserved after canonical mapping. |
Taxonomies
Web-facing classification systems
Anatomy / localization
Diseases may belong to more than one curated ocular localization group. Counts by category can therefore exceed the number of distinct diseases.
12 localization categories · 185/185 diseases classified · 187 memberships
Inheritance
Inheritance is evidence-supported and may be multi-valued. Missing support is shown as not classified rather than guessed.
AD · AR · mitochondrial · other · XLD · XLR
Gene category
Gene categories are applied to true gene entities only. Structural/regulatory loci remain outside the gene category layer.
Used in gene landscape and relationship visualizations
Variant consequence
Variant annotations are normalized into the ODDB consequence taxonomy while retaining zero-count categories in the taxonomy itself.
21 consequence categories · 1,400 canonical variants currently annotated
Clinical development stage
Source-reported trial/development stage is displayed conservatively and is not converted into an independent claim of regulatory approval.
Includes explicit unknown/no-stage states
Evidence interpretation
Rules that prevent over-interpretation
01
Target overlap is discovery, not treatment evidence
A disease-associated gene that is also a drug target creates a discovery relationship only. It does not automatically establish that the drug treats the disease.
02
Target action direction matters
Agonism, antagonism, inhibition and other target actions cannot be assumed to be therapeutically compatible merely because the same gene appears in disease and drug-target layers.
03
Variant classification is not gene–disease validity
Variant-level clinical interpretation and gene–disease validity are distinct evidence concepts and are not substituted for one another.
04
Parent disease mapping is not subtype mapping
Disease hierarchy or parent mapping does not automatically justify assigning all evidence to a more specific subtype.
05
Clinical investigation is not approval or efficacy
A clinical trial, report or source-reported maximum development phase indicates investigation context, not proof of therapeutic benefit or independent regulatory approval.
06
Conflicts remain visible
Source identifier conflicts and unresolved canonical mappings are flagged for review rather than silently overwritten.
Identifiers & provenance
Traceable source-aware records
Disease identifiers
Canonical ODDB disease IDs provide stable internal references while source names, aliases and history are preserved.
Gene identifiers
HGNC identifiers anchor canonical gene entities. Non-gene labels from legacy sources are classified separately using genetic entity type.
Variant identifiers
Variant records retain source identifiers and can link outward to resources such as dbSNP, Ensembl and gnomAD where available.
Drug identifiers
Canonical drug identity can be enriched from sources such as ChEMBL while original source drug IDs and names remain preserved.
Publication evidence
PMID/DOI and evidence-source information are retained where available to keep curation traceable to source evidence.
Frozen source record
Normalized MySQL rows retain the complete frozen TSV source row in source_record_json together with source file and source release metadata.
Clinical evidence model
Reports and trials are relational entities
Disease A
Disease B
Clinical trial / report
Drug X
Drug Y
Why ODDB does not create a cross-product
A single trial or report may mention multiple diseases and multiple drugs. ODDB stores disease→report/trial and drug→report/trial links independently.
This avoids asserting four disease–drug treatment relationships merely because two diseases and two drugs co-occur in the same source record.
Release & versioning
Frozen releases remain immutable
1
Source ingestion & curationNormalize entities, preserve evidence and flag unresolved conflicts.
→
2
ValidationRun structural and semantic checks before release freeze.
→
3
Frozen ODDB v1 releaseRelease files are not modified after validation.
→
4
Derived web curationDisplay/taxonomy corrections can evolve without rewriting frozen data.
Using Explore
How to interpret an ODDB search result
Search a canonical entity
Start with a disease, gene, variant or drug name/identifier.
Read direct associations first
Use explicit disease–gene, disease–variant and direct disease–drug evidence before discovery-only relationships.
Inspect provenance
Use source identifiers and evidence context to understand where the relationship came from.
Treat target overlap as hypothesis-generating
Target-overlap paths can suggest opportunities for investigation but are not equivalent to disease-specific treatment evidence.
Scope note
Research resource, not clinical guidance
ODDB is designed for research exploration and evidence organization. Database relationships should not be used as a substitute for clinical diagnosis, treatment recommendations or regulatory information.