Evaluation for ncatsgardkg

Evaluator: Cortes et al.

Evaluated on: 2025-08-26

This is a manual evaluation intended to identify potential barriers to reuse.

Evaluation Criteria: This evaluation uses the KG-Registry evaluation rubric as described in Cortes et al. (2025) . The rubric assesses knowledge graphs across multiple dimensions including access, provenance, documentation, maintenance, and fitness for purpose.


Access Level and Types

QuestionAnswerComment
Access to data outside of the knowledge graphYThe KG enables extraction of disease mappings, disease profiles, and pathogenesis paths, as shown in the case studies (e.g., 3-hop paths from Wilson disease)
API or online access to the knowledge graphYPublic Neo4j instance available: https://disease.ncats.io
Multiple access options availableN
Source code availabilityYPartially – The in-house framework stitcher is mentioned and linked (https://github.com/ncats/stitcher), but full KG generation scripts for GARD KG are not public
Downloadable knowledge graphYNetworkx pickle

Section Score: 4/5

Provenance of Nodes and Edges

QuestionAnswerComment
Source list providedYIntegrates 34 distinct biomedical resources
Source versions informationYORPHADATA v4.0 Gene Ontology Human Phenotype Ontology GO annotations Pathway Commons (PTC) Pharos v3.8.0
Import dependenciesYPartially – Tools like stitcher and OWL file parsing are mentioned, but not declared formally like a software repo would
Node and edge sourcesYEach node/edge carries provenance via stitch keys (e.g., N_Name, I_CODE)
Edges deduplicationYPartially – Mappings and harmonization discussed, but no explicit deduplication process for edges described
Triples source detailsYObject properties defined per source and integration method (e.g., has_phenotype, I_CODE, N_Name)
Edge type schemaYMeta-ontology schema is described with object properties and their meanings (Table 3)

Section Score: 7/7

Documented standards, schema, construction

QuestionAnswerComment
Biological usable dataYDisease profiles, drug-disease associations, and harmonization rules demonstrate biomedical applicability
Resolvable IDsYThe paper explicitly mentions using "Unique Ingredient Identifier (UNII)," "MONDO ID," and "OMIM ID" for mappings, indicating that external identifiers are a core part of their entity resolution strategy
Construction documentationYDetailed methods section describes data collection, mapping strategies, and harmonization
Transformation documentationYThe paper notes that "data cleanup was performed," mentioning specific examples like restricting prefixes for OMIM IDs and Orphanet IDs. It also discusses the challenges of programmatic annotation, such as preventing the annotation of generic terms like "Disease" or "Syndrome"
Schema usedYThe paper outlines a "meta-ontology" that serves as the schema for the knowledge graph. This includes a clear definition of primary classes and object properties used to structure the data; documented in tables 2–4

Section Score: 5/5

Update frequency and versioning

QuestionAnswerComment
Stable versionsN
Public tracker informationN
Knowledge graph contact informationYCorresponding author: qian.zhu@nih.gov and NCATS/ NIH team affiliation listed
Updated annuallyNSuggests updates over time; indicates continually updated
Prior versions accessN

Section Score: 1/5

Evaluation - Metrics and Fitness for Purpose

QuestionAnswerComment
Use case providedYFour detailed case studies: disease mapping, disease profiling, data harmonization, pathogenesis exploration
Evaluation against other modelsYThe paper compares its approach with other similar efforts, such as the semantic Diseasecard and the Monarch Initiative, placing its own work in the context of the broader field; but no explicit benchmarking
Defined scopeYFocused on rare diseases, especially those curated in GARD, and integrating relevant biomedical data
Multiple evaluation methodsYPartially – Demonstrates utility via case studies but lacks formal quantitative evaluation (mapping stats in tables 6,7,10)
Accuracy metricsYThe paper's semi-automatic mapping process for FDA orphan designations includes a manual curation step where curators labeled mappings as "Done," "Approximate," or "Failed," providing a clear way to measure the confidence or accuracy of the mappings

Section Score: 5/5

License Information

QuestionAnswerComment
LicenseCC BY 4.0