Reasoning over is_a and part_of

[mikpom]

According to relationship reasoning in the docs if A part_of B is_a C then A part_of C. But this does not seem to hold in the following example. GO:0000902 (cell morphogenesis) part_of GO:0048856 (anatomical structure development). GO:0048856 (anatomical structure development) is_a GO:0008150 (biological_process). BUT GO:0000902 (cell morphogenesis) is_a GO:0008150 (biological_process).

As I understand Biological process category is so broad that is_a assignment to it will be very liberal and those statements about reasoning won't hold for such relationships.

Can anyone give me more inference?

[kltm]

@cmungall You might be able to speak to this?

[cmungall]

On 26 Jun 2018, at 18:55, Mikhail Pomaznoy wrote:

As I understand Biological process category is so broad that  is_a  assignment to it will be very liberal and those statements about reasoning won't hold for such relationships.

Can you give a counter-example where you feel it wouldn't hold? The semantics of the relationship hold regardless of how broad or narrow the class is.

[mikpom]

There are two points here:

1. Main point is that in this particular example (I can provide more of this kind) reasoning if (A is_a B part_of C) then (A part_of C) doesn't hold. So I find that part in the GO documentation misleading. You didn't comment on that yet.

2. Another point is to distinguish in which kind of relationships this reasoning doesn't hold. As I guess here this might happen in case of a category for which is_a relationships assignment is loose. "Biological process" I think is such example and not because it is very general (top level) but because pretty much anything can be assigned to it. Counter-example would be a chain GO:0008135 (translation factor activity, RNA binding) part_of GO:0006412 (translation) is_a GO:0034641 (cellular nitrogen compound metabolic process). Here indeed translation factor activity, RNA binding is_not_a cellular nitrogen compound metabolic proccess but a part_of it and above mentioned reasoning holds.

[cmungall]

Apologies, I read too fast, I see you provided examples. In fact, for all examples 1-3 the inference is true. The first may seem odd but it's how GO models development, morphogenesis is part of development. @ukemi can comment further.

For your second example, every translation factor activity, RNA binding is inferred to be part of some cellular nitrogen compound metabolic process, this is a valid (but trivial) inference

[mikpom]

There is still misunderstanding on my question. I edited the initial comment. Hope it is now more clear what I am asking about.

[ukemi]

Morphogenesis and maturation are both part of the development of an anatomical structure. Morphogenesis, maturation and the development of the anatomical structure are all types of developmental processes. This is valid.

[mikpom]

If I know understood correctly two terms can be simultaneously connected with is_a and part_of relationships. If these relationships are not mutually exclusive then issue is resolved.

[cmungall]

Correct. We typically do not assert X is_a Y, X part_of Y, but inferences of this form are perfectly valid.