QC check for cross-species analog

[katiermullen]

Consider adding a QC check for non-human animal terms that have a cross-species analog term to a human disease with a germline mutation axioms.

Generally, non-human animal diseases should not have cross-species analogs to human diseases with germline mutations:

Example: MONDO:1010291 'achondroplasia, non-human animal' should not have cross-species analog to MONDO:0007037 'achondroplasia' because 'achondroplasia' has material basis in germline mutation in some FGFR3 and 'achondroplasia, non-human animal' does not.

Question: What about cases where the non-human animal disease has a germline mutation in an orthologous gene? Do these relationships belong in the ontology or only in the KG?

Example: Should MONDO:1012791 'osteogenesis imperfecta, COL1A2-related, dog' have cross-species analog to NTR 'COL1A2-related osteogenesis imperfecta' in issue #8757?

@sabrinatoro @nicolevasilevsky

[matentzn]

We use hgnc for human, and ncbigene for animals - we do not know, at least in the ontology as it is right now - whether two genes are orthologous, so we cant build a test around that.

The QC check though is reasonable, @twhetzel will take a stab at it!

[katiermullen]

Thank you @matentzn and @twhetzel!

@twhetzel - chatgpt and I wrote some SPARQL queries to:

1. find non-human animal diseases with OMIA x-refs
2. find non-human animal disease with cross-species analogs
3. find non-human animal diseases with cross-species analogs to human diseases with germline mutations.

Can you take a look at these and see what I did right or wrong as a starting point for the QC check?

[twhetzel]

Sure @katiermullen! Can you point me to the files to review?

[twhetzel]

Hi @katiermullen - I had time to look into this and have worked some things out and also have some questions.

Do you want separate queries for "Find all non-human animal diseases with cross-species analogs to a human disease" and then also "Find all non-human animal diseases with cross-species analogs to a human disease where that human disease has a gene association"?

Also, are these queries intended to be run as reports or as part of the Mondo QC checks that are run with each Mondo PR?

[katiermullen]

Thank you @twhetzel!

I would like to have two separate queries:

1. Find all non-human animal diseases with cross-species analogs to a human disease. This query is just for info and does not need to be run as part of the QC check that is run with each Mondo PR.
2. Find all non-human animal diseases with cross-species analogs to a human disease where that human disease has a gene association. @sabrinatoro can you comment on if you would like to have this query be part of the Mondo QC checks? I think I would like to have this as a QC check. I realize the stakes are low since the cross-species analogs are removed pre-release, but having them wrong could results in wrong classifications of disease (e.g. a non-human animal disease is classified as a non-human animal hereditary disease when it is not IF it mistakenly has a cross species analog to a human disease with a gene association.

[twhetzel]

Thanks Katie. I will treat these as queries for information until we hear back from Sabrina about (2) Find all non-human animal diseases with cross-species analogs to a human disease where that human disease has a gene association.

[twhetzel]

I pushed a minor refactor of nonhuman_animal_cross_species.sparql. Overall, the query looked good to me. I added one alternative way to get the restrictions that is a little more compact, but either format works.

In order to get all results though, I had to use the reasoned version of the ontology which uses the format report-reason-query... and then run as:
sh run.sh make report-reason-query-nonhuman_animal_cross_species

[twhetzel]

For the query to get the cross-species analogs where there is a gene or some value from the list of specific properties, I pushed a separate query for comparison. In the original query, there is an OPTIONAL statement and I am not sure what is intended there. Also, I do not think that all of the RO properties would have a gene as the object, but included them all since they were in the original.

[twhetzel]

@katiermullen let me know if you have any questions about the changes in the PR