Add some mp_efo_impc.sssom.tsv from IMPC

[matentzn]

No description provided.

[matentzn]

@sbello these are just a handful of MP EFO mappings I got from Violeta MF from IMPC a few years ago that I didn't want to get lost.. Can you take a look?

[matentzn]

Awessssome @rays22 THANK YOU! I will trust your judgment on this, and let @sbello review the rest! Thaaaaanks!

[rays22]

👍

[sbello]

I have some questions about the validity of using broad match to map efo to mp. My understanding is that broad/narrow imply that term A could be a parent/child of the mapped term B in a shared ontology but that is never the case when mapping efo to mp. For example the efo term low density lipoprotein cholesterol measurement is mapped as a broad match to abnormal circulating LDL cholesterol level but you would not want to say that abnormal circulating LDL cholesterol level is a type of low density lipoprotein cholesterol measurement. The phenotype maybe identified by measuring LDL cholesterol but it isn't itself a subtype of LDL cholesterol measurements.
Looking at the SKOS website I see
The properties skos:broadMatch and skos:narrowMatch are used to state a hierarchical mapping link between two concepts.
The property skos:relatedMatch is used to state an associative mapping link between two concepts.
So I question why efo and mp would be mapped with anything other than relatedMatch.

[sbello]

Is the purpose of the mappings to indicate which measurements may be used to identify a phenotype? Assuming that is the case then this mapping (first on the list) is incorrect:
MP:0000013 abnormal adipose tissue distribution skos:relatedMatch EFO:0007800 body fat percentage
abnormal adipose tissue distribution is about where on the body the adipose tissue if located body fat percentage is measuring the total amount of adipose tissue relative to body weight. A better MP term for this would be decreased total body fat amount or increased total body fat amount.

[matentzn]

I have some questions about the validity of using broad match to map efo to mp. My understanding is that broad/narrow imply that term A could be a parent/child of the mapped term B in a shared ontology but that is never the case when mapping efo to mp. For example the efo term low density lipoprotein cholesterol measurement is mapped as a broad match to abnormal circulating LDL cholesterol level but you would not want to say that abnormal circulating LDL cholesterol level is a type of low density lipoprotein cholesterol measurement. The phenotype maybe identified by measuring LDL cholesterol but it isn't itself a subtype of LDL cholesterol measurements.

This is a great observation, and of course conceptually correct! However, we (see our mapping guide) decided to conflate measurement and trait terms - even though they are distinct, for this project, it is safe to do. So you can read low density lipoprotein cholesterol measurement as low density lipoprotein cholesterol phenotype for the purpose of this mapping.

Given that, and @rays22 suggested fix of the body fat percentage mapping, are you ok with the rest of this?

[sbello]

I'm not sure why this is called close, cyanosis is the skin discoloration that results from low blood oxygen. If the point is to get the best match for the EFO term hypoxemia then why not use the MP term hypoxemia (MP:0012549), which would be an exact match.
Another question, why is there a phenotype term in EFO? Wouldn't the experimental factor be low oxygen.

[rays22]

The targeted set of MP terms are those that have been requested by IMPC to be mapped to EFO so that their data can related to GWAS Catalog genetic associations. I suspect that these are the terms with existing associated IMPC data. It seems they have no hypoxemia associated annotations.

"why is there a phenotype term in EFO?"

There are all kinds of terms in EFO. I guess the phenotype terms have been added in response to requests from phenotype curators associated with EBI.

[sbello]

Please add some more details to the description:
what is the use case for the mappings
how are you using the skos matches since these are not using the standard meaning

[rays22]

I have added more details to the description. The use case is to help link mouse phenotype data from IMPC to GWAS traits (annotated with EFO terms).
I thought we are applying the standard skos definitions:
"The properties skos:broadMatch and skos:narrowMatch are used to state a hierarchical mapping link between two concepts."

[sbello]

Copying this over from Slack
I'm really struggling with reviewing this, not knowing what the purpose is. For example why map erythrocyte count as a broad match to hematopoietic system phenotype? Why map edema as a broad match to hydrops fetalis when above that you have the exact match of edema to edema? Why map energy intake measurement to abnormal metabolism and not to abnormal food intake?
Ignoring the use of broad I don't see others that seem wrong just not as precise as I would have chosen if trying to map from EFO terms to MP terms. The ones to the system headers seem particularly odd to me. There are definitely better terms to map these to.

[rays22]

The updated mapping table looks OK to me.

[matentzn]

thanks @rays22!