src/gemeo/seed_aura.py

"""Seed an empty Aura instance with the minimal KG Gemeo needs.

Without external Disease/Phenotype/Gene nodes, the cohort, subgraph,
drug repurposing, family, reverse-pheno, and PCDT-compliance heads all
return empty (graceful degrade). This script populates a curated rare-
disease subset (~20 diseases, ~80 phenotypes, ~25 genes, plus a few
synthetic confirmed-dx PatientSpaces for cohort retrieval) so the SOTA
layer can demonstrate non-trivial behaviour.

Run:  python -m gemeo.seed_aura

Idempotent: uses MERGE for every node. Safe to re-run.
"""
from __future__ import annotations
import asyncio
import logging
import sys

logger = logging.getLogger("gemeo.seed_aura")
logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(name)s] %(message)s")


# ─── Curated minimal KG (pediatric / Brazilian rare disease focus) ────────

DISEASES = [
    # AT
    {"orpha": "100", "name": "Ataxia-telangiectasia", "name_pt": "Ataxia-telangiectasia",
     "cid10": "G11.3", "inheritance": "autosomal recessive",
     "prevalence": "1-9/1000000", "summary": "Progressive cerebellar ataxia, immunodeficiency, lymphoid malignancy risk."},
    # NPC
    {"orpha": "646", "name": "Niemann-Pick disease type C", "name_pt": "Doença de Niemann-Pick tipo C",
     "cid10": "E75.2", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "Lysosomal lipid storage; vertical supranuclear gaze palsy, hepatosplenomegaly."},
    # Gaucher
    {"orpha": "355", "name": "Gaucher disease", "name_pt": "Doença de Gaucher",
     "cid10": "E75.2", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "GBA deficiency; hepatosplenomegaly, bone pain, anemia."},
    # Fabry
    {"orpha": "324", "name": "Fabry disease", "name_pt": "Doença de Fabry",
     "cid10": "E75.2", "inheritance": "X-linked recessive",
     "prevalence": "1-9/100000", "summary": "GLA deficiency; acroparesthesias, angiokeratomas, renal/cardiac involvement."},
    # Pompe
    {"orpha": "365", "name": "Glycogen storage disease due to acid maltase deficiency", "name_pt": "Doença de Pompe",
     "cid10": "E74.0", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "GAA deficiency; hypotonia, cardiomyopathy in infants."},
    # MPS I
    {"orpha": "579", "name": "Mucopolysaccharidosis type 1", "name_pt": "MPS I (Hurler-Scheie)",
     "cid10": "E76.0", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "IDUA deficiency; coarse facies, dysostosis multiplex."},
    # SMA
    {"orpha": "70", "name": "Proximal spinal muscular atrophy type 1", "name_pt": "Atrofia Muscular Espinhal tipo 1",
     "cid10": "G12.0", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "SMN1 loss; severe infantile hypotonia, respiratory failure."},
    # Wilson
    {"orpha": "905", "name": "Wilson disease", "name_pt": "Doença de Wilson",
     "cid10": "E83.0", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "ATP7B deficiency; hepatic, neurologic, Kayser-Fleischer rings."},
    # DMD
    {"orpha": "98896", "name": "Duchenne muscular dystrophy", "name_pt": "Distrofia muscular de Duchenne",
     "cid10": "G71.0", "inheritance": "X-linked recessive",
     "prevalence": "1-9/100000", "summary": "DMD gene mutation; progressive muscle weakness, Gowers sign."},
    # CF
    {"orpha": "586", "name": "Cystic fibrosis", "name_pt": "Fibrose cística",
     "cid10": "E84", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "CFTR; recurrent infections, pancreatic insufficiency."},
    # PKU
    {"orpha": "716", "name": "Phenylketonuria", "name_pt": "Fenilcetonúria",
     "cid10": "E70.0", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "PAH deficiency; intellectual disability if untreated."},
    # Marfan
    {"orpha": "558", "name": "Marfan syndrome", "name_pt": "Síndrome de Marfan",
     "cid10": "Q87.4", "inheritance": "autosomal dominant",
     "prevalence": "1-9/100000", "summary": "FBN1; aortic dilation, ectopia lentis, tall stature."},
    # NF1
    {"orpha": "636", "name": "Neurofibromatosis type 1", "name_pt": "Neurofibromatose tipo 1",
     "cid10": "Q85.0", "inheritance": "autosomal dominant",
     "prevalence": "1-9/10000", "summary": "NF1; café-au-lait, neurofibromas."},
    # Rett
    {"orpha": "778", "name": "Rett syndrome", "name_pt": "Síndrome de Rett",
     "cid10": "F84.2", "inheritance": "X-linked dominant",
     "prevalence": "1-9/100000", "summary": "MECP2; regression, stereotypies, mostly females."},
    # Friedreich ataxia
    {"orpha": "95", "name": "Friedreich ataxia", "name_pt": "Ataxia de Friedreich",
     "cid10": "G11.1", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "FXN GAA expansion; gait ataxia, cardiomyopathy, diabetes."},
    # Pelizaeus-Merzbacher
    {"orpha": "702", "name": "Pelizaeus-Merzbacher disease", "name_pt": "Doença de Pelizaeus-Merzbacher",
     "cid10": "E75.2", "inheritance": "X-linked recessive",
     "prevalence": "1-9/1000000", "summary": "PLP1; nystagmus, hypotonia, leukodystrophy."},
    # Hunter MPS II
    {"orpha": "580", "name": "Mucopolysaccharidosis type 2", "name_pt": "MPS II (Hunter)",
     "cid10": "E76.1", "inheritance": "X-linked recessive",
     "prevalence": "1-9/100000", "summary": "IDS deficiency; coarse facies, hepatosplenomegaly."},
    # SCID
    {"orpha": "183660", "name": "Severe combined immunodeficiency", "name_pt": "Imunodeficiência combinada grave",
     "cid10": "D81.1", "inheritance": "various",
     "prevalence": "1-9/100000", "summary": "Multiple genes; severe T/B/NK deficiency, opportunistic infections."},
    # CAH
    {"orpha": "90794", "name": "Congenital adrenal hyperplasia due to 21-hydroxylase deficiency", "name_pt": "Hiperplasia adrenal congênita",
     "cid10": "E25.0", "inheritance": "autosomal recessive",
     "prevalence": "1-9/100000", "summary": "CYP21A2; salt-wasting, virilization."},
    # Hereditary hemolytic spherocytosis (added since H histórico familiar)
    {"orpha": "822", "name": "Hereditary spherocytosis", "name_pt": "Esferocitose hereditária",
     "cid10": "D58.0", "inheritance": "autosomal dominant",
     "prevalence": "1-9/10000", "summary": "Membrane defect; hemolysis, splenomegaly, jaundice."},
]

# HPO terms (curated, with PT-BR labels for the most clinically common)
PHENOTYPES = [
    ("HP:0001250", "Seizure", "Convulsão"),
    ("HP:0001251", "Ataxia", "Ataxia"),
    ("HP:0001260", "Dysarthria", "Disartria"),
    ("HP:0001009", "Telangiectasia", "Telangiectasia"),
    ("HP:0033106", "Increased serum alpha-fetoprotein", "AFP sérica elevada"),
    ("HP:0002720", "Decreased circulating IgA level", "IgA baixa"),
    ("HP:0006532", "Recurrent pneumonia", "Pneumonia recorrente"),
    ("HP:0002121", "Cerebellar atrophy", "Atrofia cerebelar"),
    ("HP:0000639", "Nystagmus", "Nistagmo"),
    ("HP:0001270", "Motor delay", "Atraso motor"),
    ("HP:0001249", "Intellectual disability", "Deficiência intelectual"),
    ("HP:0002240", "Hepatomegaly", "Hepatomegalia"),
    ("HP:0001744", "Splenomegaly", "Esplenomegalia"),
    ("HP:0000505", "Visual impairment", "Comprometimento visual"),
    ("HP:0001392", "Abnormal liver", "Alteração hepática"),
    ("HP:0001627", "Abnormal heart morphology", "Alteração cardíaca"),
    ("HP:0001638", "Cardiomyopathy", "Cardiomiopatia"),
    ("HP:0002715", "Abnormality of the immune system", "Imunodeficiência"),
    ("HP:0007354", "Amyotrophic lateral sclerosis-like", "Síndrome ELA-like"),
    ("HP:0002493", "Upper motor neuron dysfunction", "Disfunção do neurônio motor superior"),
    ("HP:0001257", "Spasticity", "Espasticidade"),
    ("HP:0001284", "Areflexia", "Arreflexia"),
    ("HP:0008936", "Muscular hypotonia of the trunk", "Hipotonia troncular"),
    ("HP:0003236", "Elevated circulating creatine kinase", "CK elevada"),
    ("HP:0001626", "Abnormal vasculature", "Alteração vascular"),
    ("HP:0001392", "Liver dysfunction", "Disfunção hepática"),
    ("HP:0002910", "Elevated hepatic transaminase", "Transaminases elevadas"),
    ("HP:0007360", "Aplasia/Hypoplasia of the cerebellum", "Hipoplasia cerebelar"),
    ("HP:0001263", "Global developmental delay", "Atraso global do desenvolvimento"),
    ("HP:0000252", "Microcephaly", "Microcefalia"),
    ("HP:0001876", "Pancytopenia", "Pancitopenia"),
    ("HP:0001873", "Thrombocytopenia", "Plaquetopenia"),
    ("HP:0001508", "Failure to thrive", "Baixo ganho ponderal"),
    ("HP:0002013", "Vomiting", "Vômitos"),
    ("HP:0002643", "Neonatal respiratory distress", "Insuficiência respiratória neonatal"),
    ("HP:0011675", "Arrhythmia", "Arritmia"),
    ("HP:0001417", "X-linked inheritance", "Herança ligada ao X"),
    ("HP:0001425", "Heterogeneous", "Heterogêneo"),
    ("HP:0010972", "Anemia of inadequate production", "Anemia hipoproliferativa"),
    ("HP:0001873", "Hemolytic anemia", "Anemia hemolítica"),
    ("HP:0001081", "Cholelithiasis", "Colelitíase"),
    ("HP:0006530", "Interstitial pulmonary disease", "Pneumopatia intersticial"),
    ("HP:0010301", "Spinal cord compression", "Compressão medular"),
    ("HP:0040315", "B-cell lymphoma", "Linfoma de células B"),
    ("HP:0010785", "T-cell lymphoma", "Linfoma de células T"),
]

# Genes
GENES = [
    ("ATM", "11q22.3"),
    ("NPC1", "18q11.2"),
    ("NPC2", "14q24.3"),
    ("GBA", "1q22"),
    ("GLA", "Xq22.1"),
    ("GAA", "17q25.3"),
    ("IDUA", "4p16.3"),
    ("IDS", "Xq28"),
    ("SMN1", "5q13.2"),
    ("SMN2", "5q13.2"),
    ("ATP7B", "13q14.3"),
    ("DMD", "Xp21.2"),
    ("CFTR", "7q31.2"),
    ("PAH", "12q23.2"),
    ("FBN1", "15q21.1"),
    ("NF1", "17q11.2"),
    ("MECP2", "Xq28"),
    ("FXN", "9q21.11"),
    ("PLP1", "Xq22.2"),
    ("CYP21A2", "6p21.33"),
    ("ANK1", "8p11.21"),
    ("SPTB", "14q23.3"),
    ("IL2RG", "Xq13.1"),
    ("ADA", "20q13.12"),
    ("RAG1", "11p12"),
    ("RAG2", "11p12"),
]

# Disease ↔ Phenotype edges (with frequency)
DISEASE_PHENOTYPES = [
    # AT (ORPHA:100)
    ("100", "HP:0001251", 0.95),  # ataxia
    ("100", "HP:0001009", 0.85),  # telangiectasia
    ("100", "HP:0033106", 0.95),  # AFP elevated
    ("100", "HP:0002720", 0.80),  # IgA low
    ("100", "HP:0006532", 0.85),  # recurrent pneumonia
    ("100", "HP:0002121", 0.90),  # cerebellar atrophy
    ("100", "HP:0001260", 0.80),  # dysarthria
    ("100", "HP:0002715", 0.85),  # immunodeficiency
    ("100", "HP:0040315", 0.25),  # B-cell lymphoma (lifetime)
    ("100", "HP:0010785", 0.20),  # T-cell lymphoma
    ("100", "HP:0006530", 0.50),  # interstitial pulm disease (older)
    # NPC (ORPHA:646)
    ("646", "HP:0001251", 0.85),
    ("646", "HP:0002240", 0.70),  # hepatomegaly
    ("646", "HP:0001744", 0.80),  # splenomegaly
    ("646", "HP:0001249", 0.75),  # ID
    ("646", "HP:0001260", 0.60),
    ("646", "HP:0000505", 0.70),  # vertical gaze
    # Gaucher (ORPHA:355)
    ("355", "HP:0002240", 0.85),
    ("355", "HP:0001744", 0.95),
    ("355", "HP:0001873", 0.70),  # thrombocytopenia
    # Fabry (ORPHA:324)
    ("324", "HP:0001626", 0.80),
    ("324", "HP:0001638", 0.70),
    # Pompe (ORPHA:365)
    ("365", "HP:0001638", 0.95),
    ("365", "HP:0008936", 0.95),
    ("365", "HP:0003236", 0.90),
    # MPS I (ORPHA:579)
    ("579", "HP:0002240", 0.85),
    ("579", "HP:0001744", 0.70),
    # SMA-1 (ORPHA:70)
    ("70", "HP:0008936", 0.99),
    ("70", "HP:0001284", 0.90),
    ("70", "HP:0001270", 0.99),
    # Wilson (ORPHA:905)
    ("905", "HP:0002910", 0.85),
    ("905", "HP:0001392", 0.90),
    # DMD (ORPHA:98896)
    ("98896", "HP:0001270", 0.95),
    ("98896", "HP:0003236", 0.95),
    ("98896", "HP:0001638", 0.60),
    # CF (ORPHA:586)
    ("586", "HP:0006532", 0.95),
    ("586", "HP:0001508", 0.75),
    # PKU (ORPHA:716)
    ("716", "HP:0001249", 0.85),
    # Friedreich (ORPHA:95)
    ("95", "HP:0001251", 0.95),
    ("95", "HP:0001638", 0.70),
    # Spherocytosis (ORPHA:822)
    ("822", "HP:0001873", 0.95),  # hemolytic anemia
    ("822", "HP:0001744", 0.80),
    ("822", "HP:0001081", 0.50),
]

# Disease ↔ Gene
DISEASE_GENES = [
    ("100", "ATM", "pathogenic"),
    ("646", "NPC1", "pathogenic"),
    ("646", "NPC2", "pathogenic"),
    ("355", "GBA", "pathogenic"),
    ("324", "GLA", "pathogenic"),
    ("365", "GAA", "pathogenic"),
    ("579", "IDUA", "pathogenic"),
    ("580", "IDS", "pathogenic"),
    ("70", "SMN1", "pathogenic"),
    ("905", "ATP7B", "pathogenic"),
    ("98896", "DMD", "pathogenic"),
    ("586", "CFTR", "pathogenic"),
    ("716", "PAH", "pathogenic"),
    ("558", "FBN1", "pathogenic"),
    ("636", "NF1", "pathogenic"),
    ("778", "MECP2", "pathogenic"),
    ("95", "FXN", "pathogenic"),
    ("702", "PLP1", "pathogenic"),
    ("90794", "CYP21A2", "pathogenic"),
    ("822", "ANK1", "pathogenic"),
    ("822", "SPTB", "pathogenic"),
    ("183660", "IL2RG", "pathogenic"),
    ("183660", "ADA", "pathogenic"),
    ("183660", "RAG1", "pathogenic"),
]

# Synthetic confirmed-dx PatientSpaces (so cohort retrieval surfaces real exemplars)
# Each is a confirmed AT/NPC/Gaucher/etc patient with a small HPO panel.
SYNTHETIC_PATIENTS = [
    {"space_id": "synth-at-001", "orpha": "100", "state": "SP", "hpos": ["HP:0001251", "HP:0001009", "HP:0033106", "HP:0006532"], "age": 8},
    {"space_id": "synth-at-002", "orpha": "100", "state": "RJ", "hpos": ["HP:0001251", "HP:0033106", "HP:0002715"], "age": 10},
    {"space_id": "synth-at-003", "orpha": "100", "state": "MG", "hpos": ["HP:0001251", "HP:0001009", "HP:0002121", "HP:0006532"], "age": 7},
    {"space_id": "synth-at-004", "orpha": "100", "state": "BA", "hpos": ["HP:0001251", "HP:0033106", "HP:0040315"], "age": 14},
    {"space_id": "synth-at-005", "orpha": "100", "state": "SP", "hpos": ["HP:0001251", "HP:0001009", "HP:0006530"], "age": 12},
    {"space_id": "synth-npc-001", "orpha": "646", "state": "SP", "hpos": ["HP:0001251", "HP:0001744", "HP:0002240"], "age": 9},
    {"space_id": "synth-npc-002", "orpha": "646", "state": "RS", "hpos": ["HP:0001251", "HP:0001249", "HP:0000505"], "age": 11},
    {"space_id": "synth-gaucher-001", "orpha": "355", "state": "SP", "hpos": ["HP:0002240", "HP:0001744", "HP:0001873"], "age": 6},
    {"space_id": "synth-fabry-001", "orpha": "324", "state": "PR", "hpos": ["HP:0001626", "HP:0001638"], "age": 16},
    {"space_id": "synth-pompe-001", "orpha": "365", "state": "SP", "hpos": ["HP:0001638", "HP:0008936"], "age": 1},
    {"space_id": "synth-sma1-001", "orpha": "70", "state": "RJ", "hpos": ["HP:0008936", "HP:0001284", "HP:0002643"], "age": 1},
    {"space_id": "synth-dmd-001", "orpha": "98896", "state": "MG", "hpos": ["HP:0001270", "HP:0003236"], "age": 5},
]


async def _q(cypher, params=None):
    from space_graph import _safe_query
    return await _safe_query(cypher, params or {})


async def seed():
    print("Seeding Aura with curated rare-disease KG...")

    # 1. Diseases
    for d in DISEASES:
        await _q("""
        MERGE (x:Disease {orphaCode: $orpha})
        SET x.name = $name,
            x.cid10DescriptionPt = $name_pt,
            x.cid10 = $cid10,
            x.inheritance = $inheritance,
            x.prevalence = $prevalence,
            x.summary = $summary
        """, d)
    print(f"  ✓ {len(DISEASES)} Disease nodes")

    # 2. Phenotypes (HPO)
    for hpo, name, name_pt in PHENOTYPES:
        await _q("""
        MERGE (p:Phenotype {hpoId: $hpo})
        SET p.name = $name, p.name_pt = $name_pt
        """, {"hpo": hpo, "name": name, "name_pt": name_pt})
    print(f"  ✓ {len(PHENOTYPES)} Phenotype nodes")

    # 3. Genes
    for sym, loc in GENES:
        await _q("""
        MERGE (g:Gene {symbol: $sym})
        SET g.location = $loc
        """, {"sym": sym, "loc": loc})
    print(f"  ✓ {len(GENES)} Gene nodes")

    # 4. Disease ↔ Phenotype
    for orpha, hpo, freq in DISEASE_PHENOTYPES:
        await _q("""
        MATCH (d:Disease {orphaCode: $orpha})
        MATCH (p:Phenotype {hpoId: $hpo})
        MERGE (d)-[r:HAS_PHENOTYPE]->(p)
        SET r.frequency = $freq
        """, {"orpha": orpha, "hpo": hpo, "freq": freq})
    print(f"  ✓ {len(DISEASE_PHENOTYPES)} Disease-Phenotype edges")

    # 5. Disease ↔ Gene
    for orpha, sym, path in DISEASE_GENES:
        await _q("""
        MATCH (d:Disease {orphaCode: $orpha})
        MATCH (g:Gene {symbol: $sym})
        MERGE (d)-[r:ASSOCIATED_WITH]->(g)
        SET r.pathogenicity = $path
        """, {"orpha": orpha, "sym": sym, "path": path})
    print(f"  ✓ {len(DISEASE_GENES)} Disease-Gene edges")

    # 6. Synthetic PatientSpaces (so cohort works)
    for p in SYNTHETIC_PATIENTS:
        await _q("""
        MERGE (s:PatientSpace {space_id: $space_id})
        SET s.state = $state, s.age = $age, s.is_synthetic = true
        """, {"space_id": p["space_id"], "state": p["state"], "age": p["age"]})

        # Confirmed hypothesis pointing at the disease
        await _q("""
        MATCH (s:PatientSpace {space_id: $space_id})
        MERGE (h:SpaceHypothesis {hypothesis_id: $space_id + '-h0'})
        SET h.status = 'confirmed', h.orpha_code = $orpha, h.probability = 0.95
        MERGE (s)-[:HAS_HYPOTHESIS]->(h)
        """, p)
        await _q("""
        MATCH (h:SpaceHypothesis {hypothesis_id: $space_id + '-h0'})
        MATCH (d:Disease {orphaCode: $orpha})
        MERGE (h)-[:TARGETS]->(d)
        """, p)

        # SpaceEvent → Phenotypes
        for i, hpo in enumerate(p["hpos"]):
            eid = f"{p['space_id']}-e{i}"
            await _q("""
            MATCH (s:PatientSpace {space_id: $space_id})
            MERGE (e:SpaceEvent {event_id: $eid})
            SET e.event_type = 'symptom_onset', e.title = $hpo
            MERGE (s)-[:HAS_EVENT]->(e)
            """, {"space_id": p["space_id"], "eid": eid, "hpo": hpo})
            await _q("""
            MATCH (e:SpaceEvent {event_id: $eid})
            MATCH (p:Phenotype {hpoId: $hpo})
            MERGE (e)-[:INVOLVES]->(p)
            """, {"eid": eid, "hpo": hpo})
    print(f"  ✓ {len(SYNTHETIC_PATIENTS)} synthetic PatientSpaces seeded")

    # Verify
    counts = await _q("""
    MATCH (n) RETURN labels(n)[0] AS l, count(n) AS c
    ORDER BY c DESC LIMIT 20
    """)
    print("\nFinal Aura inventory:")
    for r in counts:
        print(f"  {r['l']}: {r['c']}")


if __name__ == "__main__":
    asyncio.run(seed())