Background
Currently, a Measurement with an abnormal value may imply a PhenotypicFeature, but the schema lacks a built-in mechanism to formally link the two. For instance, a Measurement for LOINC:26515-7 (Platelets [#/volume] in blood), if abnormally low, directly implies a PhenotypicFeature for Thrombocytopenia (HP:0001873). In many clinical integrations, it is better to store both the quantitative measurement and the qualitative ontology term while preserving their relationship.
This is a specific instance of a broader architectural issue: we frequently need to link various building blocks to ensure AI-ready provenance and a clear clinical narrative. Other examples include:
- Linking a
MedicalAction to a PhenotypicFeature or Disease (e.g., administering lithium to treat Mania, or performing an MRI to discover an abnormality of the corpus callosum).
- Linking a
PhenotypicFeature to a Disease to explicitly document which features were key to establishing the diagnosis.
- Modeling the treatment of a recurrent phenotypic feature, which is currently difficult because treatments are linked to particular occurrences rather than to an ontology term.
Proposed solutions for discussion
- Option 1 (Targeted): Add an ontology class called
inferred_phenotypic_feature (OntologyClass) directly to Measurement. (e.g., if blood sugar = 10, inferred feature = Hypoglycemia HP:0001943). Note: We must specify the intended semantics here. A PhenotypicFeature implies a characteristic abnormality, whereas a measurement might simply capture a transient state (e.g., High Blood Sugar 30 minutes after eating).
- Option 2 (Generalized): Introduce a unique ID system to
PhenotypicFeature, Measurement, and MedicalAction elements, allowing them to cross-reference one another within a single Phenopacket payload.
Background
Currently, a
Measurementwith an abnormal value may imply aPhenotypicFeature, but the schema lacks a built-in mechanism to formally link the two. For instance, aMeasurementforLOINC:26515-7(Platelets [#/volume] in blood), if abnormally low, directly implies aPhenotypicFeaturefor Thrombocytopenia (HP:0001873). In many clinical integrations, it is better to store both the quantitative measurement and the qualitative ontology term while preserving their relationship.This is a specific instance of a broader architectural issue: we frequently need to link various building blocks to ensure AI-ready provenance and a clear clinical narrative. Other examples include:
MedicalActionto aPhenotypicFeatureorDisease(e.g., administering lithium to treat Mania, or performing an MRI to discover an abnormality of the corpus callosum).PhenotypicFeatureto aDiseaseto explicitly document which features were key to establishing the diagnosis.Proposed solutions for discussion
inferred_phenotypic_feature(OntologyClass) directly toMeasurement. (e.g., if blood sugar = 10, inferred feature = HypoglycemiaHP:0001943). Note: We must specify the intended semantics here. APhenotypicFeatureimplies a characteristic abnormality, whereas a measurement might simply capture a transient state (e.g., High Blood Sugar 30 minutes after eating).PhenotypicFeature,Measurement, andMedicalActionelements, allowing them to cross-reference one another within a single Phenopacket payload.