Investigate why some diseases match to multiple patterns

[matentzn]

See this comment as an example:

https://github.com/monarch-initiative/mondo/pull/2268#issue-517185148

[nicolevasilevsky]

inflammatory_disease_by_site.yaml is the correct pattern for the example in #2268

specific_infectious_disease_by_location - the variable is MONDO:0000001 disease, it should probably be more specific, like MONDO_0005550 'infectious disease'

specific_inflammatory_disease_by_site.yaml the variable is MONDO:0000001 disease, it should probably be more specific too, MONDO_0021166 'inflammatory disease'

[nicolevasilevsky]

Are there other cases like this, @matentzn?

[matentzn]

I think you are right! You should fix these patterns as you suggest! Other cases of overlapping patterns:

• [ ] allergic_form_of_disease.yaml basis_in_disruption_of_process.yaml
• [ ] allergic_form_of_disease.yaml disrupts_process.yaml
• [ ] basis_in_disruption_of_process.yaml disrupts_process.yaml
• [ ] inborn_metabolic.yaml inborn_metabolic_disrupts.yaml
• [ ] hemangioma.yaml location.yaml
• [ ] leiomyoma.yaml location.yaml
• [ ] leiomyosarcoma.yaml location.yaml
• [ ] lipoma.yaml location.yaml
• [ ] location_top.yaml location.yaml
• [ ] lymphoma.yaml location.yaml
• [ ] melanoma.yaml location.yaml
• [ ] meningioma.yaml location.yaml
• [ ] neuroendocrine_neoplasm.yaml location.yaml
• [ ] rhabdomyosarcoma.yaml location.yaml
• [ ] small_cell_carcinoma.yaml location.yaml
• [ ] specific_disease_by_dysfunctional_structure.yaml disease_by_dysfunctional_structure.yaml
• [ ] specific_disease_by_dysfunctional_structure.yaml disease_series_by_gene.yaml
• [ ] specific_infectious_disease_by_agent.yaml infectious_disease_by_agent.yaml
• [ ] allergy.yaml specific_infectious_disease_by_agent.yaml
• [ ] allergy.yaml environmental_stimulus.yaml
• [ ] specific_infectious_disease_by_location.yaml autoimmune_inflammation.yaml
• [ ] specific_infectious_disease_by_location.yaml inflammatory_disease_by_site.yaml
• [ ] specific_infectious_disease_by_location.yaml specific_inflammatory_disease_by_site.yaml
• [ ] autoimmune_inflammation.yaml specific_inflammatory_disease_by_site.yaml
• [ ] inflammatory_disease_by_site.yaml specific_inflammatory_disease_by_site.yaml

The entire list of all disease with duplicate matches this way can be found here.

[cmungall]

hmm, for the specific_X patterns is there a way to say that filler should be a proper subclass of disease?

[nicolevasilevsky]

I think you just have to say the variable is the more specific subclass, like for specific_infectious_disease_by_location

it should be:

classes: infectious disease: MONDO:0005550 anatomical structure: UBERON:0000061

[matentzn]

@nicolevasilevsky suggestion is the only way - you cant say "proper subclass" as in "any class that is a subclass of the filler, but not the filler itself.

[matentzn]

@cmungall Is there any problem with Nicoles suggestion?

[nicolevasilevsky]

@matentzn suggested that @sabrinatoro read through and try to solve this ticket. This type of problem would be nice to mention in the paper.

[sabrinatoro]

Analysis is here: https://docs.google.com/spreadsheets/d/1stK1FPjNRgNjt0QJiqRKp7nMjvq08yuBfIoIy7uVvaI/edit?usp=sharing

I compared the pattern pairs from the list above. The analysis shows that :

• some patterns are semantically the same. These should probably be merged
• some patterns are parent/child: one pattern is the same as the other but has more specificity/restrictions
• some patterns (2) should be reviewed as the relation used is incorrect (I think)
• only 1 pair of patterns contains different scope. We might want to create a new specific pattern to encompass these 2 concepts when they are present together.