search

monarch-initiative / dipper

Data Ingestion Pipeline for Monarch
https://dipper.readthedocs.io/en/latest/
BSD 3-Clause "New" or "Revised" License
57 stars 26 forks source link

Relations to use for G2P associations #254

Open mbrush opened 8 years ago

mbrush commented 8 years ago

We recently reconsidered the use of RO:0002200 ! has_phenotype in G2P associations, deciding to use more precise relations indicating causality or contribution to the condition (disease or phenotype). See #195. The following hierarchy of 'causes or contributes to condition' properties was implemented in RO to support this, as per the RO issue here.

concepts_011

  1. Can we pull the trigger on this and swap out hasphenotype for 'causes condition'_ in our data? What is required for a variant to 'cause' a disease? (deterministically leads to the disorder in any individual bearing it, or has been known to lead to the disorder in some individuals)?
  2. How should we handle assertions that a variant is non-causal (benign for) a condition? Esp in the area of cancer variant classification, variants are asserted with statistical confidence to not be causative for a disease such as breast cancer. We could create a 'non-causal for condition' property to capture such assertions. Or we could use negation (however this ends up working) to say the variant !'causes condition' the disease.
  3. Consider the case of phenotypes getting propagated to punned gene class IRIs. Is it odd to say that a gene could 'cause' a disease (in the same way as a variant)?
kshefchek commented 6 years ago

Addressing this in https://github.com/monarch-initiative/dipper/pull/602

mbrush commented 6 years ago

Great Kent. Will we also switch to these predicates for genotype-phenotype associations?

Also, be aware that the Translator work is pushing us toward defining relations specific for gene-condition associations (since genes in the sense we use the term don't cause a condition, but rather it is variants in these genes that are causative). But the final approach is still being worked out (see here). Once resolved, I suspect these relations will make their way into RO, and we can consider using them in Dipper for gene-condition associations if it suits us.

kshefchek commented 6 years ago

Will we also switch to these predicates for genotype-phenotype associations?

Do we want to use these for entities related to phenotypes or just diseases? My understanding was the latter but we can also add them to phenotypes.

Also, be aware that the Translator work...

We could add place holder predicates, but it probably makes sense to wait until these are in RO. I prefer the type agnostic relations over gene specific ones, but am fine with whichever is decided. I wouldn't be surprised if there is some benefit from a cypher perspective for splitting the relations.