RO compatibility issue

[matentzn]

On our path to upgrading GENO to ODK, the following two issues came up:

1. GO_0003674 (molecular function) is hardcoded in GENO as function while RO asserts it to be a process
2. GENO makes a lot of use of the 'has role' relationship which formally applies to independent continuants. However, most if not all of the entities in GENO are declared to be "generically dependent continuants"; this causes about 40 classes to be unsatisfiable.

How do you want to proceed here?

1. I assume we just remove the GENO assertion; in the end, we need to get rid of all non-native axioms from GENO in any cases (non-native: axioms that belong to another ontology).
2. The quick fix is to create a new object property or, even better, find a more suitable one in RO that reflects the geno use case. Example:

genomic background = 'genomic genotype' and ('has role' some reference)

More:

@dosumis @cmungall Any suggestions for a better relation in this case?

[dosumis]

GO_0003674 (molecular function) is hardcoded in GENO as function while RO asserts it to be a process

GENO is wrong and RO is right (although this should, of course, be coming directly from GO. @cmungall - Do we have a GO base, derived from go-edit.obo, that includes this? I guess this also points to the need to ODK-ise RO...)

GENO makes a lot of use of the 'has role' relationship which formally applies to independent continuants. However, most if not all of the entities in GENO are declared to be "generically dependent continuants"; this causes about 40 classes to be unsatisfiable.

I understand and support GENOs use of 'has role' here. We need an RO ticket on this, but perhaps in the meantime need a more generic relation in GENO?

[cmungall]

go-edit doesn't include this at the moment, we need to improve our go-base, Eric is working on this

instead of has-role use inverse of RO_0000052 (will be characteristicof)

On Mon, Sep 9, 2019 at 8:08 AM David Osumi-Sutherland < [email protected]> wrote:

GO_0003674 (molecular function) is hardcoded in GENO as function while RO asserts it to be a process

GENO is wrong and RO is right (although this should, of course, be coming directly from GO. @cmungall https://github.com/cmungall - Do we have a GO base, derived from go-edit.obo, that includes this? I guess this also points to the need to ODK-ise RO...)

GENO makes a lot of use of the 'has role' relationship which formally applies to independent continuants. However, most if not all of the entities in GENO are declared to be "generically dependent continuants"; this causes about 40 classes to be unsatisfiable.

I understand and support GENOs use of 'has role' here. We need an RO ticket on this, but perhaps in the meantime need a more generic relation in GENO?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/monarch-initiative/GENO-ontology/issues/47?email_source=notifications&email_token=AAAMMOJRXMIA4UQLEKPGKULQIZRIFA5CNFSM4IU4IAS2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD6H57YI#issuecomment-529522657, or mute the thread https://github.com/notifications/unsubscribe-auth/AAAMMONHZLQJXBVYMUNP2D3QIZRIFANCNFSM4IU4IASQ .

[mbrush]

I removed the molecular function import from GENO, as it was there for exploratory purposes and is not needed.

To deal with the 'has role' issue, for now I created a geno-specific 'has role' object property that is defined to allow for generically dependent continuations as bearers. I didn't use the suggested inverse(inheres in) approach because I want to be clear that being a 'reference' is a role assigned to a sequence feature or allele, not an intrinsic quality/characteristic of it.

object property: GENO:0000968

• label: 'sequence role'
• definition: A role assigned to a sequence feature, collection, or genotype, e.g. serving as a 'reference' against with other sequences are compared.
• comment: The RO:0000087 (has role) property cannot be used here because its domain is explicitly constrained to independent continuants, and sequence features in GENO are generically dependent continuants.

If RO or some other ontology ends up providing a 'has role' property that can be applied here, I will use this and obsolete the GENO sequence role term. If this all sounds reasonable, we can close this issue.

[cmungall]

RO_0000052 would be fine

It gets confusing if we reuse labels with different IDs. What about something more specific like 'has sequence role'?

this will presumably be coordinated with SO's feature qualities?

This should do for now. However, I would like to revisit some of these classes and what the use case is for them.

Also, text defs should align with logical defs (see my blog post). Compare:

"An allele whose sequence matches what is consdiered[sic] to be the reference sequence at that location in the genome." <==> (allele and has_role some reference)