Prior Knowledge: Integrate BioCypher

[vicpaton]

The integration of BioCypher in NC should provide a graph (networkx or CORNETO graph) retrieved from various prior knowledge sources, including OmniPath.

[nilskre]

To keep track of the current state: the issue to create a networkx output adapter for BioCypher. I keep you updated, when this is implemented in BioCypher.

[slobentanzer]

@vicpaton @martingarridorc The NetworkX output adapter is merged: biocypher/biocypher#352

Currently, the workflow is to write a pickle of the NetworkX object to disk. Maybe it's worth considering an addition to do the transfer in-memory as well (as we do with the Pandas object). This could make the workflow of handover to the algorithm part of networkcommons more seamless.

When using the NetworkX mode as described in the docs, BioCypher also writes an ingestion script as an accessibility tool for loading the object back into Python.

[vicpaton]

great! this is very useful thanks @slobentanzer @nilskre

[vicpaton]

@slobentanzer @nilskre do you think it would be possible to directly provide the networkx graph? It would allow better integration with the nc

[nilskre]

By directly you mean, that you have a BioCypher instance (bc) within nc and can directly get the networkx in-memory representation of the KG by calling something like bc.get_networkx() returning the networkx graph, right? (so kind of similar how it is done with Pandas at the moment within BioCypher https://biocypher.org/api.html#in-memory-pandas-knowledge-graph).

[slobentanzer]

Yes, I think that would be most seamless.

[vicpaton]

@nilskre yes, exactly! that would be great

[nilskre]

Alright, got it. Just created an issue in BioCypher to also make this use case directly possible.

Until we have implemented this in BioCypher a fast workaround for your PoC, would be to directly access the networkx graph by the private attributes (not recommended for permanent use!)

    bc = BioCypher(
        biocypher_config_path="biocypher_config.yaml", # with dbms: networkx
        schema_config_path="schema_config.yaml",
    )
    # Run the import
    bc.write_nodes(node_generator())
    bc.write_edges(edge_generator())

    # Write command line call
    bc.write_import_call() # <- important to call this, otherwise the networkx graph is not initialized

    in_memory_graph = bc._writer.G # <- this returns the in memory networkx graph
    print(in_memory_graph)