SPOKE (v5.5) edge/node file validation violations

[jdr0887]

Issues to be fixed for SPOKE to pass minimal checklist

• Node category has invalid delimiter...please use pipes instead of unit separator. For example: biolink:Gene\u001fbiolink:GeneOrGeneProduct\u001fbiolink:GenomicEntity\u001fbiolink:ChemicalEntityOrGeneOrGeneProduct\u001fbiolink:PhysicalEssence\u001fbiolink:OntologyClass\u001fbiolink:BiologicalEntity\u001fbiolink:ThingWithTaxon\u001fbiolink:NamedThing\u001fbiolink:PhysicalEssenceOrOccurrent\u001fbiolink:MacromolecularMachineMixin\u001fbiolink:Protein\u001fbiolink:GeneProductMixin\u001fbiolink:Polypeptide\u001fbiolink:ChemicalEntityOrProteinOrPolypeptide
• Invalid Node id CURIE format. For example ECCODE: 3.5.1.M19 contains a space character which is not allowed.
  Details

  
  ECCODE: 3.5.1.M19
  ECCODE: 4.2.1.M17
  ECCODE: 1.1.2.e
  ECCODE: 3.2.1.M34
  ECCODE: 3.5.1.M20
  ECCODE: 3.2.1.M32

[matentzn]

@matebio do you have an ETA for this request, so we know if we need to copy it to the next sprint?

[matentzn]

Sorry to press @matebio but can you give us a little update for this ticket here? Thank you!

[matebio]

v5.6 has been uploaded to fix node category and ECCODE node ids. Just fyi when we ran matrix-validator on v5.5 it didn't pick up on the issue with the node categories.

[jdr0887]

I ran the matrix validator on the edges & the output was not super informative:
invalid_starts_with_biolink_predicate: interacts_with

However, when I take a closer look ( cat spoke_5.6_edges.tsv | awk '{print $1} | sort | uniq ), there is only one predicate used, "interacts_with" and it does not start with 'biolink:'. I would imagine/expect more than just one predicate in the edges file & those edges need to be prefixed with 'biolink:'.

The nodes validation was fine...meaning no errors reported.

[matentzn]

@matebio this cant be intended, right? That all edges are "interacts with"?

[matebio]

No we seem to have lost the majority of edges in this version. We'll work on it

[piotrkan]

FYI @matebio @matentzn I tried running our ingestion & integration with spoke v5.6, I found out there are duplicates in the file (only two but thought I would flag this, as it leads to an error in our pipeline due to pandera checks within transformers class).

EDIT: also noticed it has no node names, there are only id and category columns available for nodes file, unsure if that's intended.

[piotrkan]

@matebio I also see that nodes & edges are mismatching (i.e many of nodes are not captured within edges subject or object identifiers) - this is related to losing most edges probably, but thought it's worth flagging as we really need to integrated good quality SPOKE for our milestone 5 deadline which is soon.

happy to quickly iterate with you to ensure the quality is good in our pipeline if helpful

[may-lim]

Hi @matebio do you have any updates on this and when a fixed version of SPOKE will be submitted?

[piotrkan]

Hi @matebio thank you for uploading v5.7 of Spoke - I can confirm it looks much better and the normalization rate between SPOKE and ROBOKOP/RTX improved significantly. It's awesome 🚀

That said, I did spot a few issues with it to put on your radar e.g.

• there are some duplicates in the raw nodes file
• predicates seem to be missing 'biolink' suffix - are they all derived from biolink or are some not adherent to the biolink model
• there are no names for nodes

Would be great to have them fixed in the future versions of SPOKE.

[may-lim]

@matentzn @kevinschaper @jdr0887 I think it would be a good exercise to run the validator library on v5.7 of SPOKE, especially as we are about to connect the validator to the main MATRIX pipeline. If you agree, I can set up your permissions to the SPOKE directory in our prod bucket

[jdr0887]

@matentzn @kevinschaper @jdr0887 I think it would be a good exercise to run the validator library on v5.7 of SPOKE, especially as we are about to connect the validator to the main MATRIX pipeline. If you agree, I can set up your permissions to the SPOKE directory in our prod bucket

That would be great. Thanks!

[may-lim]

Have added Nico, Kathleen, Kevin, and Jason to the group to access Embiology and SPOKE per this doc, which also has instructions on how to access the datasets. Let me or @Dashing-Nelson know if you have any issues accessing this!