obophenotype/upheno

'Vulvar hemangioma' logical definition problem

Opened this issue · 11 comments

balhoff commented

This issue is contributing to incoherency in the Ubergraph build.

HPO term

Vulvar hemangioma EquivalentTo has_part some
(increased amount and (characteristic_of some
(Hemangioma and (part_of some external female genitalia))) and (qualifier some abnormal))

Suggested revision of logical definition

The above axiom uses another phenotype (Hemangioma) as an anatomical entity. You could use Hemangioma of the lip as a template:

'has part' some 
    (Hemangioma
     and ('characteristic of' some lip)
     and ('has modifier' some abnormal))

One (much smaller) issue is that style uses a phenotype as a characteristic, but that's a much broader issue about the Sub-Q pattern (@cmungall @matentzn). For either type I think you have to manually classify under Hemangioma, which is unfortunate.

matentzn commented

I would suggest to drop both EQ's for now, they are hard to define in terms of anatomical or physiological characteristics, and the current ones are not right..

pnrobinson commented

@matentzn the suggested revision seems non-crazy -- what is the issue with it?

matentzn commented

The problem is we currently do not have a uPheno patterns for extending the definition of another phenotype. In your example your you are trying to extend the definition of Hemangioma.. Maybe we should work on this more seriously, but right now an EQ needs to be defined in terms of a PAtO quality and some biological entity.. Maybe make a uPheno issue requesting instructions for extending existing EQS. Or comment on the one that is already there..

pnrobinson commented

Not sure if maybe this would be a decent pattern

'has part' some
('increased amount' and
 ('characteristic of' some ('blood vessel') and 
(qualifier some abnormal)))

Hemangiomas are benign neoplasms characterized by excessive amounts of blood vessels.
The definition is not precise but it might be better?
We current do not have any definitions for neoplasms, that would make things a little easier too

matentzn commented

@rays22 can you make a proposal for a pattern for this term?

@pnrobinson I don't think that "increased amount of blood vessel" is a precise enough - Using EQ logical definitions means we need to be very precise. I would suggest, for now, to simply delete the EQ and have @rays22 move the issue to uPheno and discuss it at a call.

rays22 commented

1. Vulvar hemangioma

Hemangioma and
('characteristic of part of' some 'external female genitalia')

where

Explanation:

A Hemangioma HP:0001028 is an Abnormality of the vasculature HP:0002597. An Abnormality of the vasculature HP:0002597 is defined as an abnormal quality:

'has part' some quality and 'characteristic of part of' some 'vasculature' and 'has modifier' some 'abnormal').

No need to add the redundant increased amount and abnormal qualities.

2. Hemangioma EQ

'has_part' some (
'neoplastic, non-malignant' and
('characteristic_of' some 'blood vessel') and
('has_modifier' some 'abnormal')
)

However, I second Nico's concern about finding an EQ that is precise enough for Hemangioma, but distinguishes it from other types of non-malignant neoplastic blood vessel lesions. I am not sure if all benign blood vessel tumours are hemangiomas, but I am not a pathologist.
Are hemangioblastoma, benign carotid body paraganglioma, capillary lymphangioma etc. also hemangiomas or distinct categories?

I am happy to discuss this at an upcoming uPheno editors call. However, I would like to bring a suitable candidate EQ to the call, so that it is easier for the group to find a resolution.

pnrobinson commented

@rays22 @matentzn can we finalize this?

LCCarmody commented

If we are agreed with @rays22 suggestion, I can implement

pnrobinson commented

@LCCarmody which of the two suggestions? Wouldn't it be better to add a term such as hemangiomatous to PATO to avoid the ambiguities described above?

LCCarmody commented

I was arguing for the 1st suggestion (though not preferred to use HP in axiom). We can create a ticket to add the term to PATO. I guess it would be sibling to neoplastic (PATO:0000141)? It would take awhile to implement