Logical definition of "ovarian granulosa cell tumor" is redundant

[matentzn]

~#8262(closed)~ #8264 caused the following failure as pointed out by @twhetzel:

DISJOINT-SIB-OF ovarian granulosa cell tumor SubClassOf Nothing

• DISJOINT-SIB-OF ovarian granulosa cell tumor SubClassOf granulosa cell tumor
  • granulosa cell tumor EquivalentTo neoplasm and (disease has location some granulosa cell)
    • granulosa cell SubClassOf follicular cell of ovary
    • follicular cell of ovary SubClassOf ovarian surface epithelial cell
      • ovarian surface epithelial cell SubClassOf part of some ovary
• disease has location o part of SubPropertyOf: disease has location
• ovarian granulosa cell tumor EquivalentTo granulosa cell tumor and (disease has location some ovary)
• ovarian granulosa cell tumor DisjointWith DISJOINT-SIB-OF ovarian granulosa cell tumor

  Axiom Impact

  Axioms used 1 times

  • granulosa cell tumor EquivalentTo neoplasm and (disease has location some granulosa cell) [O1]
  • ovarian granulosa cell tumor EquivalentTo granulosa cell tumor and (disease has location some ovary) [O1]
  • granulosa cell SubClassOf follicular cell of ovary [O1]
  • follicular cell of ovary SubClassOf ovarian surface epithelial cell [O1]
  • ovarian surface epithelial cell SubClassOf part of some ovary [O1]
  • DISJOINT-SIB-OF ovarian granulosa cell tumor SubClassOf granulosa cell tumor [O1]
  • ovarian granulosa cell tumor DisjointWith DISJOINT-SIB-OF ovarian granulosa cell tumor [O1]
  • disease has location o part of SubPropertyOf: disease has location [O1]

The problem is a bit opaque as it is caught by an ancient check which has long since been superceded by robot reason --allow-equivalent-classes none (the test injects hypothetical siblings and makes sure they can theoretically exist, thereby ensuring a child is not the same as its parent).

Here is the issue:

granulosa cell tumor is defined as a granulosa cell tumor EquivalentTo neoplasm and (disease has location some granulosa cell) and ovarian granulosa cell tumor is defined as a granulosa cell tumor and (disease has location some ovary). Unfortunately, through a chain of inference in CL, granulosa cell is considered a part of the ovaries. So saying that ovarian granulosa cell tumor is has (disease has location some ovary) is redundant, therefore the two classes are considered the same.

This is a great catch by QC. My assumption is that either the fact that granulosa cell is considered a part of the ovaries is wrong, or ovarian granulosa cell tumor is logically underspecified!

[twhetzel]

Thanks @matentzn - that was my general impression from the qc error log. Did you come across why this happened now as a result of running the refresh imports?

@sabrinatoro or @nicolevasilevsky are either of you able to take the curator role on this to resolve this issue?

[matentzn]

Did you come across why this happened now as a result of running the refresh imports?

It is either due to changes in CL classification (one that made granulosa cell part of the ovaries) or because the role chain disease has location < disease has location o part of was added to RO!

[nicolevasilevsky]

@sabrinatoro or @nicolevasilevsky are either of you able to take the curator role on this to resolve this issue?

no, I can't, sorry

[matentzn]

A quick fix is to remove the logical definition of ovarian granulosa cell tumor, but someone should probably google the following:

Are all granulosa cell tumors are by definition "ovarian"?

[twhetzel]

From a few references I've seen:

• Granulosa cell tumors are the most common type of sex cord stromal tumor accounting for about 2-5% of ovarian neoplasms. REF: https://pmc.ncbi.nlm.nih.gov/articles/PMC3978936/
• However, "Granulosa cell tumors can arise in locations other than the ovary..." REF: https://pmc.ncbi.nlm.nih.gov/articles/PMC8893158/#:~:text=Granulosa%20cell%20tumors%20can%20arise,differentiating%20GCT%20from%20other%20neoplasms.

[cmungall]

here's the issue

id: MONDO:0006036
name: granulosa cell tumor
def: "A slow-growing, malignant tumor, characterize by the presence of granulosa-like cells and Call-Exner bodies, that is almost always found in the ovary. In rare cases, it has also been found in the testicle. There are two types of granulosa cell tumor that can be distinguished under the microscope: the adult and the juvenile. The testicular juvenile granulosa cell tumors are perhaps the most common congenital testicular neoplasms." [NCIT:C3070]
intersection_of: MONDO:0005070 ! neoplasm
intersection_of: disease_has_location CL:0000501 ! granulosa cell

we need better DPs for neoplasm->CL

[sabrinatoro]

I agree with everything said here. In addition:

• "granulosa cell tumor" can occur in ovary and in testis. Therefore, "granulosa cell tumor" is NOT equivalent to "ovarian granulosa cell tumor" (this had to be confirmed).

• After reading the definition for "granulosa cell tumor", I think that the logical definition of "neoplasm located in a granulosa cell" is not correct. "granulosa cell tumor" is "characterized by the presence of granulosa-like cells and Call-Exner bodies". (ie it is NOT "located in"). I therefore removed the logical definition for this term. The resulting classification seems more sane.

• I agree with Chris that the DPs for neoplasm located in cells should be reviewed. A new issue should be created for this (@twhetzel can you please create this issue? thank you)

PR has been updated; waiting for QC checks to run.

[twhetzel]

Sure, I added an issue referencing back to this for more details.