Unified Biomedical Knowledge Graph (UBKG)

Source Contexts

The source context for an instance of the UBKG (or UBKG context) describes a collection of sets of assertions, each of which is identified by SAB.

A SAB can correspond to one of the following types of sets of assertions:

1. Ontology data, such as that extracted from OWL files published in sources such as the NCBO BioPortal or the OBO Foundry.
2. Reference data extracted from public sources such as UniProtKB.
3. Result Summarized experimental result data obtained from entities such as NIH-supported laboratories.

Citations

We have sought to identify citations for the sources used in UBKG contexts. The preferred form of citation is a PubMed ID.

The sources for citation references are included. Sources are of the following types:

1. OBO home page
2. Web pages of source stewards
3. GitHub repositories of source stewards

Licensing

We have sought to identify licensing under which sources are distributed. The majority of the sources use Creative Commons licenses.

Adaptations

The following is intended to satisfy requirements for licenses that require descriptions of adaptation (e.g., CC BY-SA).

Adaptations to source data include:

1. Use of subsets of data from the source–i.e., selected codes
2. Modifications of code IDs for the purpose of standardization. See CodeID Standardization below.
3. Use of content from multiple sources to generate new “mappings” sources–e.g., between HGNC and HPO.
4. Transformation of downloaded content into files in UBKG Edge/Node format–e.g., as done with UniProtKB and GENCODE data.

Citation or Licensing Questions

If the citation or licensing information for a source needs to be updated, please contact the UBKG steward.

Jonathan Silverstein, MD

Jonathan Silverstein, MD
Department of Biomedical Informatics
University of Pittsburgh

Contexts

UMLS source context (UMLS-Graph)

The UMLS context includes sets of assertions obtained from over 100 vocabularies organized by the UMLS, including

• SNOMEDCT_US
• ICD10
• LOINC
• HGNC

All of these assertions are considered as being from ontologies.

CodeID standardization

In the UBKG, the colon is a reserved character, used to delimit SAB from code in a CodeID property of a Code node.

A small number of vocabularies from the UMLS have default code formats that either include the name of the vocabulary in the code or contain a colon. These vocabularies and their formats include:

Executing the generation script with the SAB argument of UMLS reformats codes from the UMLS to the format SAB:CODE. HPO codes are standardized to HP as the SAB.

non-UMLS Source Platforms

Sources of data for the UBKG (other than the UMLS base) include:

1. OBO Foundry
2. NCBO BioPortal
3. Sites maintained by stewards, including GitHub repositories, FTP sites, etc.
4. Globus (for sources specific to Data Distillery)
5. Spreadsheets in the custom SimpleKnowledge Editor format, stored in Google Drive
6. Custom ETL scripts

UBKG base context (UMLS plus)

The UBKG base context appends to the UMLS context assertions from a number of other standard biomedical ontologies and data sources. The UBKG base context includes assertions from the following SABs:

RefSeq gene summaries

The base context includes summaries of genes from RefSeq.

• Citation: NCBI Handbook (Reference)
• Licensing: Public Domain

A separate script (refseq.py) adds RefSeq summaries to the UBKG. This import assumes that GenCode has been ingested.

Azimuth-Cell Ontology mappings

The HuBMAP project is building a mapping between tissue codings from Azimuth and Cell Ontology (CL). The AZ-CL mappings will be added to UBKG contexts as needed.

HuBMAP/SenNet context

The HuBMAP/SenNet context appends to the UBKG base set assertions from the following SABs:

Data Distillery context

The Data Distillery context appends to the UBKG base set assertions to support participants in Data Distillery, a project of the National Institute of Health’s Common Fund Data Ecosystem (CFDE). SABs include: