/BioPortal-to-KGX

Assemble a BioPortal Knowledge Graph

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

BioPortal-to-KGX

Assemble a BioPortal Knowledge Graph through the following steps:

  • Transform the BioPortal 4store data dump to KGX graphs, with ROBOT preprocessing
  • Validate the output graphs with KGX to determine alignment to the Biolink Model
  • Obtain additional ontology metadata through the Bioportal API
  • Retrieve mappings for nodes without clear Bioportal analogues through Bioportal

Usage

Prepare a dump of the Bioportal 4store data with the 4s-dump script.

The dump will be in the form of n-triples, with individual sets of records in nested directories and one line of metadata at the top of each file.

Run BioPortal-to-KGX with all validation and metadata retrieval options as:

python run.py --input ../path/to/your/data/ --kgx_validate --robot_validate --pandas_validate --write_curies --get_bioportal_metadata --ncbo_key YOUR_NCBO_API_KEY_HERE

Specify individual ontologies to include or exclude with the --include_only and --exclude options, respectively, each followed by a comma-delimited list of the original hashed file ID from the 4store dump.

For example:

python run.py --input ../path/to/your/data/ --include_only dabd4d902360003975fb25ae56f8,7b95f2cc27c8fb0d5df11fbdb078

Output will be written to the /bioportal_to_kgx directory within /transformed, with subdirectories named for the 4store graph and each subgraph.

Each subgraph will contain:

  • node and edge files ({subgraph_name}_nodes.tsv and {subgraph_name}_edges.tsv, respectively)
  • A JSON version of the ontology ({subgraph_name}_relaxed.json)
  • logs containing any validation messages about the transforms

Troubleshooting

  • The --robot_validate option may fail on larger ontologies like NCBITAXON with java.lang.OutOfMemoryError. Consider omitting this option or running ROBOT on files directly, as needed.