model-organism-phenotype

1. When to use, and when not

Use this skill when you need model organism phenotype data or human phenotype ontology associations for a gene. Entry points include:

• Checking what mouse knockout or knockin phenotypes are reported for a gene -- the observable consequences of gene disruption in a model organism.
• Looking up HPO (Human Phenotype Ontology) terms associated with a gene through disease-gene mappings.
• Building the phenotypic evidence layer of a target validation case: does disrupting this gene produce phenotypes relevant to the indication?
• Identifying safety-relevant phenotypes from mouse knockouts as part of a preclinical safety review -- with the standing caveat that mouse phenotypes are informative, not predictive.
• Comparing phenotype profiles across candidate targets.

Do not use when:

• You need human genetic constraint (pLI, LOEUF) or disease associations (GWAS, ClinVar) -> target-genetic-evidence.
• You need tissue expression data -> tissue-expression-profile (multi-tissue) or target-genetic-evidence (GTEx whole-blood).
• You need predicted regulatory variant effects -> regulatory-variant-effect.
• You need protein structure confidence -> protein-structure-confidence.

2. Preconditions

• Gene input: a gene symbol (e.g. TP53, BRCA1). MGI resolves symbols to mouse orthologs; HPO resolves to NCBI Gene IDs.
• No authentication needed -- MGI, the Alliance of Genome Resources, and HPO are all public APIs.
• MGI resolution path: the tool resolves gene symbols through MGI's marker search, scrapes the MGI accession ID from the marker detail page, then fetches phenotype annotations from the Alliance of Genome Resources REST API. This multi-step resolution means that genes without MGI marker records or without Alliance phenotype annotations are refused with a remedy message.
• HPO resolution path: the tool resolves gene symbols through HPO's gene search API to an NCBI Gene ID, then fetches phenotype and disease annotations.
• Run pkde doctor before first use.

3. Tool invocations

Run search before analyze. analyze reads from disk and summarises without network access.

The .phenotype-<source>.json is the verbatim combined API response. The .artifact.json is the structured pkde.phenotype.v1 artifact with parsed phenotype records. For MGI, each record carries the MP ontology term, allele type, and PMID references. For HPO, each record carries the HPO term ID, and the artifact summary includes gene-level disease associations.

All output options: --json for machine-readable output, --quiet for paths only, --out to override the default output directory.

4. Interpretation contract

Verdicts

phenotype analyze produces:

• phenotypes_found -- one or more phenotype annotations exist for the gene in the queried source. The analysis reports the count of phenotypes, unique terms, and the top phenotype terms.
• no_phenotypes -- no phenotype annotations found.

MGI phenotype annotations

MGI/Alliance phenotype data describes observable consequences of gene disruption in mouse -- knockout, knockin, and other allele types. Each annotation carries:

• phenotype_term: the Mammalian Phenotype (MP) ontology term (e.g. "embryonic lethality", "abnormal heart morphology").
• mp_id: the MP ontology identifier.
• allele_type: the allele symbol (describes the type of genetic modification).
• references: PMID references for the annotation.

Multiple phenotype terms for one gene are normal -- a knockout may produce many observable phenotypes. The number and severity of phenotypes is informative but does not by itself predict human outcomes.

HPO phenotype annotations

HPO maps genes to human phenotype terms through disease-gene associations. These are ontology-derived associations, not direct experimental observations. Each annotation carries:

• phenotype_term: the HPO term name (e.g. "Intellectual disability", "Short stature").
• hpo_id: the HPO identifier.

The artifact summary includes disease_associations -- the diseases through which the gene-phenotype link was established. This context is essential: an HPO term associated with a gene through a rare Mendelian disease may not be relevant to a common-disease drug target.

Mandatory relays

Check mandatory_relays in the .analysis.json. Every relay code present must be satisfied in the finding.

Consequence rules

• MGI phenotypes found: report the count and the most prominent phenotype terms. Carry the model_organism_not_human qualifier. State the organism (Mus musculus) and the allele types observed. Embryonic lethality in mouse knockouts is particularly informative for safety assessment -- but the relay applies: lethal in mouse does not mean toxic in human pharmacological inhibition.
• HPO phenotypes found: report the phenotype terms and the disease associations through which they were established. HPO terms linked through a rare Mendelian disease are evidence of gene-disease biology, but the clinical relevance to a common-disease program depends on the pathway, not on the ontology association alone.
• No phenotypes: report as "no phenotype annotations in [source] for this gene." This may mean the gene has not been knocked out in mouse (for MGI) or has no mapped HPO terms. It is missing data, not evidence of no phenotype.
• Both sources: when both MGI and HPO data are available, report them separately. Mouse phenotypes are experimental observations; HPO associations are ontology-derived. Concordance between them (e.g. both sources reporting cardiac phenotypes) strengthens the evidence but does not prove the human phenotype will match.

Cross-disciplinary consequences

• Mouse knockout lethality or severe organ phenotypes are material to safety assessment regardless of which specialist fetched the data. The model_organism_not_human relay ensures these are reported as model organism observations, not human safety predictions.
• Phenotype data complements genetic constraint evidence (gnomAD) and disease association evidence (GWAS, ClinVar). A constrained gene with lethal mouse knockouts and strong GWAS associations presents a different target profile than one with only GWAS associations.
• HPO disease associations can identify indication-relevant biology that GWAS alone may miss -- particularly for rare diseases where GWAS power is insufficient.
• The absence of mouse phenotype data for a gene does not mean the gene can be safely modulated -- it means the experiment has not been done or has not been annotated in MGI.

What this section produces

Following this contract produces a Layer 1 finding in findings/ that cites the .analysis.json, satisfies the model_organism_not_human relay when it fires, names the organism, and does not carry mouse phenotypes into human safety predictions. Everything the tools emitted stays under raw/genomics/.

5. Failure modes and anti-fabrication guard

Do not compute these values yourself. The value is whatever the tool emitted. If the tool did not run, there is no value. If a required tool is unavailable, report the task as blocked. Do not estimate, and do not proceed on an assumed result.

Named pathologies

• Carrying mouse knockout phenotypes into human safety claims. A lethal knockout in mouse does not mean the drug will be toxic. The model_organism_not_human relay exists because this is the most consequential misreading. Species differences in gene function, compensatory mechanisms, and the difference between genetic knockout and pharmacological inhibition all limit the inference.
• Treating HPO associations as direct experimental evidence. HPO maps genes to phenotype terms through disease associations in the ontology. These are curated associations, not experimental observations. A gene associated with "Intellectual disability" via a rare syndrome does not mean modulating the gene causes cognitive impairment.
• Ignoring the disease context of HPO associations. An HPO term linked through a rare autosomal recessive condition has different relevance to a common-disease drug program than one linked through an autosomal dominant condition affecting the same pathway.
• Treating no phenotype data as "gene is safe to modulate." No MGI phenotypes means the knockout has not been reported, not that it produces no phenotype. Many genes have not been knocked out in mouse or have not had their phenotypes deposited in MGI.
• Conflating phenotype count with severity. A gene with 50 phenotype annotations may have many mild or overlapping phenotypes. A gene with 3 annotations including embryonic lethality is a more significant safety signal. Phenotype terms must be read, not counted.
• Blending MGI and HPO results into a single phenotype profile. MGI provides mouse experimental data; HPO provides human ontology associations. They use different ontologies (MP vs HP), describe different organisms, and represent different types of evidence. Report them separately and note concordance or discordance.
• Reporting a gene resolution failure as "no phenotypes." MGI resolution goes through marker search and accession ID scraping. A failure at any resolution step is a tool error (refused with remedy), not a phenotype finding. Do not report "no phenotypes" when the tool could not resolve the gene.