""" Graph seeder — fetches REAL data from live public APIs and populates Neo4j. Data sources (all free, no auth): - ClinicalTrials.gov v2 API (NCT trial records) - RxNorm (NIH) (medication RxCUI codes) - ICD-10 CM (NLM) (diagnosis codes) - PubMed (NCBI) (supporting literature PMIDs) - Synthetic patients (500 realistic profiles matched to real trials) Run once to seed, or schedule periodically to stay current. """ import httpx import asyncio import time import random from neo4j_setup import neo4j_conn CTGOV_BASE = "https://clinicaltrials.gov/api/v2/studies" RXNORM_BASE = "https://rxnav.nlm.nih.gov/REST" ICD10_BASE = "https://clinicaltables.nlm.nih.gov/api/icd10cm/v3/search" PUBMED_BASE = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils" FDA_BASE = "https://api.fda.gov/drug" # Conditions to seed — expand as needed SEED_CONDITIONS = [ "breast cancer", "prostate cancer", "non-small cell lung cancer", "colorectal cancer", "ovarian cancer", "melanoma", "leukemia", "lymphoma", "glioblastoma", "pancreatic cancer", ] # Key oncology medications to pre-load SEED_MEDICATIONS = [ "trastuzumab", "pembrolizumab", "nivolumab", "osimertinib", "olaparib", "enzalutamide", "bevacizumab", "rituximab", "imatinib", "dabrafenib", "vemurafenib", "atezolizumab", "durvalumab", "cetuximab", "erlotinib", "capecitabine", ] # ICD-10 prefixes for oncology SEED_ICD10_PREFIXES = [ "C50", "C61", "C34", "C18", "C56", "C43", "C91", "C85", "C71", "C25", ] # ── Neo4j helpers ───────────────────────────────────────────────────────────── def upsert(query: str, params: dict | None = None): try: neo4j_conn.run_query(query, params or {}) except Exception as e: print(f" [neo4j] warn: {e}") def batch_upsert(queries: list[tuple[str, dict]]): for q, p in queries: upsert(q, p) # ── ClinicalTrials.gov ──────────────────────────────────────────────────────── async def fetch_trials_for_condition(client: httpx.AsyncClient, condition: str, page_size: int = 50) -> list[dict]: try: resp = await client.get(CTGOV_BASE, params={ "query.cond": condition, "filter.overallStatus": "RECRUITING", "pageSize": page_size, "format": "json", }, timeout=30) resp.raise_for_status() return resp.json().get("studies", []) except Exception as e: print(f" [ctgov] error for '{condition}': {e}") return [] def _extract_trial(study: dict, condition: str) -> dict | None: try: proto = study["protocolSection"] ident = proto["identificationModule"] status = proto.get("statusModule", {}) design = proto.get("designModule", {}) eligibility = proto.get("eligibilityModule", {}) desc = proto.get("descriptionModule", {}) sponsor = proto.get("sponsorCollaboratorsModule", {}) contacts = proto.get("contactsLocationsModule", {}) outcomes = proto.get("outcomesModule", {}) phases = design.get("phases", ["N/A"]) locations = contacts.get("locations", []) return { "nct_id": ident["nctId"], "title": ident.get("briefTitle", "")[:200], "status": status.get("overallStatus", "UNKNOWN"), "phase": phases[0] if phases else "N/A", "condition": condition, "brief_summary": desc.get("briefSummary", "")[:1000], "eligibility_criteria": eligibility.get("eligibilityCriteria", "")[:2000], "min_age": eligibility.get("minimumAge", ""), "max_age": eligibility.get("maximumAge", ""), "sex": eligibility.get("sex", "ALL"), "enrollment": design.get("enrollmentInfo", {}).get("count", 0), "start_date": status.get("startDateStruct", {}).get("date", ""), "completion_date": status.get("completionDateStruct", {}).get("date", ""), "sponsor": sponsor.get("leadSponsor", {}).get("name", "")[:100], "primary_outcomes": [o.get("measure", "")[:100] for o in outcomes.get("primaryOutcomes", [])[:3]], "location_count": len(locations), "locations": [ { "facility": loc.get("facility", "")[:100], "city": loc.get("city", ""), "state": loc.get("state", ""), "country": loc.get("country", "US"), "lat": loc.get("geoPoint", {}).get("lat"), "lon": loc.get("geoPoint", {}).get("lon"), } for loc in locations[:10] ], } except Exception as e: return None async def seed_trials(client: httpx.AsyncClient) -> int: print("\n[1/5] Seeding clinical trials from ClinicalTrials.gov...") total = 0 for condition in SEED_CONDITIONS: studies = await fetch_trials_for_condition(client, condition) print(f" {condition}: {len(studies)} trials fetched") for study in studies: trial = _extract_trial(study, condition) if not trial: continue # Upsert trial node upsert(""" MERGE (t:Trial {id: $nct_id}) SET t += { title: $title, status: $status, phase: $phase, condition: $condition, brief_summary: $brief_summary, eligibility_criteria: $eligibility_criteria, min_age: $min_age, max_age: $max_age, sex: $sex, enrollment: $enrollment, start_date: $start_date, completion_date: $completion_date, sponsor: $sponsor, location_count: $location_count, source: 'clinicaltrials.gov', updated_at: datetime() } """, trial) # Upsert Condition → Trial relationship upsert(""" MERGE (c:ConditionNode {name: $condition}) WITH c MATCH (t:Trial {id: $nct_id}) MERGE (c)-[:HAS_TRIAL]->(t) """, {"condition": condition, "nct_id": trial["nct_id"]}) # Upsert study sites for loc in trial["locations"]: if loc.get("lat") and loc.get("lon"): upsert(""" MERGE (s:StudySite {facility: $facility, city: $city, state: $state}) SET s += {country: $country, lat: $lat, lon: $lon, source: 'clinicaltrials.gov'} WITH s MATCH (t:Trial {id: $nct_id}) MERGE (t)-[:CONDUCTED_AT]->(s) """, {**loc, "nct_id": trial["nct_id"]}) total += 1 await asyncio.sleep(0.5) # Rate limit courtesy print(f" Total trials seeded: {total}") return total # ── RxNorm (NIH) — Medications ──────────────────────────────────────────────── async def fetch_rxcui(client: httpx.AsyncClient, drug_name: str) -> list[dict]: try: resp = await client.get(f"{RXNORM_BASE}/drugs.json", params={"name": drug_name}, timeout=15) resp.raise_for_status() d = resp.json() groups = d.get("drugGroup", {}).get("conceptGroup", []) results = [] for grp in groups: tty = grp.get("tty", "") for concept in grp.get("conceptProperties", [])[:3]: results.append({ "rxcui": concept.get("rxcui", ""), "name": concept.get("name", ""), "tty": tty, "search_name": drug_name, }) return results[:5] # Top 5 except Exception as e: print(f" [rxnorm] error for '{drug_name}': {e}") return [] async def seed_medications(client: httpx.AsyncClient) -> int: print("\n[2/5] Seeding medications from RxNorm...") total = 0 for drug_name in SEED_MEDICATIONS: concepts = await fetch_rxcui(client, drug_name) for concept in concepts[:1]: # Primary concept only upsert(""" MERGE (m:Medication {rxcui: $rxcui}) SET m += { name: $name, tty: $tty, generic_name: $search_name, source: 'rxnorm', updated_at: datetime() } """, concept) total += 1 print(f" {drug_name}: {len(concepts)} RxCUI concepts") await asyncio.sleep(0.2) print(f" Total medications seeded: {total}") return total # ── ICD-10 CM (NLM) — Diagnoses ────────────────────────────────────────────── async def fetch_icd10(client: httpx.AsyncClient, prefix: str) -> list[dict]: try: resp = await client.get(ICD10_BASE, params={ "sf": "code,name", "terms": prefix, "maxList": 20, }, timeout=15) resp.raise_for_status() data = resp.json() if not data or len(data) < 4: return [] return [{"code": item[0], "name": item[1]} for item in data[3]] except Exception as e: print(f" [icd10] error for '{prefix}': {e}") return [] async def seed_diagnoses(client: httpx.AsyncClient) -> int: print("\n[3/5] Seeding diagnoses from ICD-10 CM...") total = 0 for prefix in SEED_ICD10_PREFIXES: codes = await fetch_icd10(client, prefix) for item in codes: upsert(""" MERGE (d:Diagnosis {code: $code}) SET d += {name: $name, source: 'icd10cm', updated_at: datetime()} """, item) total += 1 # Link ICD prefix → condition names for matching condition_map = { "C50": "breast cancer", "C61": "prostate cancer", "C34": "non-small cell lung cancer", "C18": "colorectal cancer", "C56": "ovarian cancer", "C43": "melanoma", "C91": "leukemia", "C85": "lymphoma", "C71": "glioblastoma", "C25": "pancreatic cancer", } if prefix in condition_map: upsert(""" MATCH (d:Diagnosis) WHERE d.code STARTS WITH $prefix MATCH (c:ConditionNode {name: $condition}) MERGE (d)-[:MAPS_TO_CONDITION]->(c) """, {"prefix": prefix, "condition": condition_map[prefix]}) print(f" ICD-10 {prefix}: {len(codes)} codes") await asyncio.sleep(0.2) print(f" Total diagnoses seeded: {total}") return total # ── PubMed (NCBI) — Supporting Literature ──────────────────────────────────── async def fetch_pubmed_ids(client: httpx.AsyncClient, condition: str, count: int = 5) -> list[str]: try: resp = await client.get(f"{PUBMED_BASE}/esearch.fcgi", params={ "db": "pubmed", "term": f"clinical trial {condition} treatment[Title/Abstract]", "retmax": count, "retmode": "json", "sort": "relevance", }, timeout=15) resp.raise_for_status() return resp.json()["esearchresult"]["idlist"] except Exception as e: print(f" [pubmed] error for '{condition}': {e}") return [] async def fetch_pubmed_summary(client: httpx.AsyncClient, pmid: str) -> dict | None: try: resp = await client.get(f"{PUBMED_BASE}/esummary.fcgi", params={ "db": "pubmed", "id": pmid, "retmode": "json", }, timeout=15) resp.raise_for_status() result = resp.json()["result"] if pmid not in result: return None r = result[pmid] return { "pmid": pmid, "title": r.get("title", "")[:200], "source": r.get("source", ""), "pub_date": r.get("pubdate", ""), "authors": ", ".join(a.get("name", "") for a in r.get("authors", [])[:3]), } except Exception as e: return None async def seed_literature(client: httpx.AsyncClient) -> int: print("\n[4/5] Seeding supporting literature from PubMed...") total = 0 for condition in SEED_CONDITIONS[:5]: # Top 5 conditions to keep fast pmids = await fetch_pubmed_ids(client, condition) for pmid in pmids: summary = await fetch_pubmed_summary(client, pmid) if not summary: continue upsert(""" MERGE (p:Publication {pmid: $pmid}) SET p += { title: $title, journal: $source, pub_date: $pub_date, authors: $authors, source: 'pubmed', updated_at: datetime() } WITH p MATCH (c:ConditionNode {name: $condition}) MERGE (p)-[:SUPPORTS_RESEARCH_ON]->(c) """, {**summary, "condition": condition}) total += 1 print(f" {condition}: {len(pmids)} publications linked") await asyncio.sleep(0.3) print(f" Total publications seeded: {total}") return total # ── Biomarkers (static — curated from COSMIC / NCIT) ───────────────────────── # Expand seed conditions to 20 oncology types SEED_CONDITIONS = [ "breast cancer", "prostate cancer", "non-small cell lung cancer", "colorectal cancer", "ovarian cancer", "melanoma", "leukemia", "lymphoma", "glioblastoma", "pancreatic cancer", "bladder cancer", "renal cell carcinoma", "thyroid cancer", "multiple myeloma", "endometrial cancer", "cervical cancer", "gastric cancer", "hepatocellular carcinoma", "head and neck cancer", "sarcoma", ] CURATED_BIOMARKERS = [ # Breast cancer {"id": "HER2_POS", "name": "HER2 Positive", "gene": "ERBB2", "loinc": "85319-2", "condition": "breast cancer"}, {"id": "HER2_NEG", "name": "HER2 Negative", "gene": "ERBB2", "loinc": "85319-2", "condition": "breast cancer"}, {"id": "BRCA1_MUT", "name": "BRCA1 Pathogenic Variant", "gene": "BRCA1", "loinc": "21636-6", "condition": "breast cancer"}, {"id": "BRCA2_MUT", "name": "BRCA2 Pathogenic Variant", "gene": "BRCA2", "loinc": "21637-4", "condition": "breast cancer"}, {"id": "PIK3CA_MUT", "name": "PIK3CA Mutation", "gene": "PIK3CA", "loinc": "82457-4", "condition": "breast cancer"}, {"id": "TP53_MUT", "name": "TP53 Mutation", "gene": "TP53", "loinc": "21637-4", "condition": "breast cancer"}, {"id": "ER_POS", "name": "Estrogen Receptor Positive", "gene": "ESR1", "loinc": "85310-1", "condition": "breast cancer"}, {"id": "PR_POS", "name": "Progesterone Receptor Positive", "gene": "PGR", "loinc": "85321-8", "condition": "breast cancer"}, {"id": "TNBC", "name": "Triple Negative Breast Cancer", "gene": "ERBB2/ESR1/PGR", "loinc": "85319-2", "condition": "breast cancer"}, # Lung {"id": "EGFR_L858R", "name": "EGFR L858R Mutation", "gene": "EGFR", "loinc": "81704-9", "condition": "non-small cell lung cancer"}, {"id": "EGFR_DEL19", "name": "EGFR Exon 19 Deletion", "gene": "EGFR", "loinc": "81704-9", "condition": "non-small cell lung cancer"}, {"id": "EGFR_T790M", "name": "EGFR T790M Resistance Mutation", "gene": "EGFR", "loinc": "81704-9", "condition": "non-small cell lung cancer"}, {"id": "ALK_FUSION", "name": "ALK Gene Fusion", "gene": "ALK", "loinc": "81695-9", "condition": "non-small cell lung cancer"}, {"id": "ROS1_FUSION", "name": "ROS1 Gene Fusion", "gene": "ROS1", "loinc": "81696-7", "condition": "non-small cell lung cancer"}, {"id": "MET_EX14", "name": "MET Exon 14 Skipping", "gene": "MET", "loinc": "82139-8", "condition": "non-small cell lung cancer"}, {"id": "KRAS_G12C", "name": "KRAS G12C Mutation", "gene": "KRAS", "loinc": "81434-5", "condition": "non-small cell lung cancer"}, {"id": "PDL1_HIGH", "name": "PD-L1 TPS ≥50%", "gene": "CD274", "loinc": "73977-1", "condition": "non-small cell lung cancer"}, {"id": "PDL1_LOW", "name": "PD-L1 TPS 1-49%", "gene": "CD274", "loinc": "73977-1", "condition": "non-small cell lung cancer"}, {"id": "PDL1_NEG", "name": "PD-L1 TPS <1%", "gene": "CD274", "loinc": "73977-1", "condition": "non-small cell lung cancer"}, # Prostate {"id": "PSA_ELEVATED","name": "PSA Elevated (>4 ng/mL)", "gene": "KLK3", "loinc": "2857-1", "condition": "prostate cancer"}, {"id": "PTEN_LOSS", "name": "PTEN Loss", "gene": "PTEN", "loinc": "21637-4", "condition": "prostate cancer"}, {"id": "AR_V7", "name": "Androgen Receptor Splice Variant 7", "gene": "AR", "loinc": "82145-5", "condition": "prostate cancer"}, # Colorectal {"id": "MSI_H", "name": "Microsatellite Instability-High", "gene": "MLH1/MSH2", "loinc": "85077-6", "condition": "colorectal cancer"}, {"id": "MSS", "name": "Microsatellite Stable", "gene": "MLH1/MSH2", "loinc": "85077-6", "condition": "colorectal cancer"}, {"id": "KRAS_WT", "name": "KRAS Wild-Type", "gene": "KRAS", "loinc": "21637-4", "condition": "colorectal cancer"}, {"id": "BRAF_V600E", "name": "BRAF V600E Mutation", "gene": "BRAF", "loinc": "81287-7", "condition": "colorectal cancer"}, {"id": "NRAS_MUT", "name": "NRAS Mutation", "gene": "NRAS", "loinc": "82143-0", "condition": "colorectal cancer"}, # Melanoma {"id": "BRAF_V600K", "name": "BRAF V600K Mutation", "gene": "BRAF", "loinc": "81287-7", "condition": "melanoma"}, {"id": "TMB_HIGH", "name": "Tumor Mutational Burden High (≥10)", "gene": "TMB", "loinc": "94076-7", "condition": "melanoma"}, {"id": "NRAS_MEL", "name": "NRAS Mutation (Melanoma)", "gene": "NRAS", "loinc": "82143-0", "condition": "melanoma"}, # GBM {"id": "IDH1_R132H", "name": "IDH1 R132H Mutation", "gene": "IDH1", "loinc": "82140-6", "condition": "glioblastoma"}, {"id": "IDH1_WT", "name": "IDH1 Wild-Type", "gene": "IDH1", "loinc": "82140-6", "condition": "glioblastoma"}, {"id": "MGMT_METH", "name": "MGMT Promoter Methylation", "gene": "MGMT", "loinc": "85319-2", "condition": "glioblastoma"}, {"id": "EGFR_AMP", "name": "EGFR Amplification", "gene": "EGFR", "loinc": "81704-9", "condition": "glioblastoma"}, # Leukemia / Lymphoma {"id": "BCR_ABL1", "name": "BCR-ABL1 Fusion (Philadelphia Chr)", "gene": "BCR/ABL1", "loinc": "33899-6", "condition": "leukemia"}, {"id": "FLT3_ITD", "name": "FLT3 Internal Tandem Duplication", "gene": "FLT3", "loinc": "82144-8", "condition": "leukemia"}, {"id": "NPM1_MUT", "name": "NPM1 Mutation", "gene": "NPM1", "loinc": "82147-1", "condition": "leukemia"}, {"id": "CD20_POS", "name": "CD20 Positive", "gene": "MS4A1", "loinc": "85080-0", "condition": "lymphoma"}, {"id": "EZH2_MUT", "name": "EZH2 Mutation", "gene": "EZH2", "loinc": "82148-9", "condition": "lymphoma"}, # New conditions {"id": "FGFR3_MUT", "name": "FGFR3 Mutation", "gene": "FGFR3", "loinc": "82150-5", "condition": "bladder cancer"}, {"id": "VHL_LOSS", "name": "VHL Gene Loss", "gene": "VHL", "loinc": "82151-3", "condition": "renal cell carcinoma"}, {"id": "MTOR_MUT", "name": "mTOR Pathway Mutation", "gene": "MTOR", "loinc": "82152-1", "condition": "renal cell carcinoma"}, {"id": "BRAF_THYROID","name": "BRAF V600E (Thyroid)", "gene": "BRAF", "loinc": "81287-7", "condition": "thyroid cancer"}, {"id": "RET_FUSION", "name": "RET Gene Fusion", "gene": "RET", "loinc": "82153-9", "condition": "thyroid cancer"}, {"id": "NTRK_FUSION", "name": "NTRK Gene Fusion", "gene": "NTRK1/2/3", "loinc": "82154-7", "condition": "thyroid cancer"}, {"id": "WHSC1_MUT", "name": "MMSET/WHSC1 Mutation", "gene": "NSD2", "loinc": "82155-4", "condition": "multiple myeloma"}, {"id": "CDKN2A_LOSS", "name": "CDKN2A Loss", "gene": "CDKN2A", "loinc": "82156-2", "condition": "multiple myeloma"}, {"id": "POLE_MUT", "name": "POLE Mutation", "gene": "POLE", "loinc": "82157-0", "condition": "endometrial cancer"}, {"id": "CTNNB1_MUT", "name": "CTNNB1 Mutation", "gene": "CTNNB1", "loinc": "82158-8", "condition": "endometrial cancer"}, {"id": "HPV_POS", "name": "HPV Positive", "gene": "HPV", "loinc": "21440-3", "condition": "cervical cancer"}, {"id": "ERBB2_GC", "name": "HER2 Amplification (Gastric)", "gene": "ERBB2", "loinc": "85319-2", "condition": "gastric cancer"}, {"id": "HBV_POS", "name": "Hepatitis B Virus Positive", "gene": "HBV", "loinc": "16933-4", "condition": "hepatocellular carcinoma"}, {"id": "TERT_MUT", "name": "TERT Promoter Mutation", "gene": "TERT", "loinc": "82159-6", "condition": "hepatocellular carcinoma"}, {"id": "PIK3CA_HNC", "name": "PIK3CA Mutation (H&N)", "gene": "PIK3CA", "loinc": "82457-4", "condition": "head and neck cancer"}, {"id": "HPV_HNSC", "name": "HPV-Positive HNSCC", "gene": "HPV", "loinc": "21440-3", "condition": "head and neck cancer"}, {"id": "CDK4_AMP", "name": "CDK4 Amplification", "gene": "CDK4", "loinc": "82160-4", "condition": "sarcoma"}, {"id": "MDM2_AMP", "name": "MDM2 Amplification", "gene": "MDM2", "loinc": "82161-2", "condition": "sarcoma"}, ] def seed_biomarkers() -> int: print("\n[5/5] Seeding biomarkers (curated from COSMIC/NCIT)...") for bm in CURATED_BIOMARKERS: upsert(""" MERGE (b:Biomarker {id: $id}) SET b += {name: $name, gene: $gene, loinc: $loinc, source: 'curated', updated_at: datetime()} WITH b MERGE (c:ConditionNode {name: $condition}) MERGE (b)-[:RELEVANT_TO]->(c) """, bm) print(f" {len(CURATED_BIOMARKERS)} biomarkers seeded and linked to conditions") return len(CURATED_BIOMARKERS) # ── Eligibility relationships ───────────────────────────────────────────────── def derive_eligibility_relationships(): print("\n[+] Deriving eligibility relationships...") upsert("MATCH (d:Diagnosis)-[:MAPS_TO_CONDITION]->(c:ConditionNode)-[:HAS_TRIAL]->(t:Trial) MERGE (d)-[:ELIGIBLE_FOR]->(t)") upsert("MATCH (b:Biomarker)-[:RELEVANT_TO]->(c:ConditionNode)-[:HAS_TRIAL]->(t:Trial) MERGE (b)-[:MAY_QUALIFY_FOR]->(t)") print(" Eligibility relationships derived.") # ══════════════════════════════════════════════════════════════════════════════ # Synthetic Patient Engine — 100 k clinically-informed personas # Distributions based on: SEER 2023, TCGA biomarker atlas, ASCO guidelines, # US Census 2020 demographics, ACS Cancer Facts & Figures 2024. # ══════════════════════════════════════════════════════════════════════════════ # ── Name pools (US Census racial/ethnic proportions) ───────────────────────── _NAMES_F_WHITE = ["Emma","Olivia","Ava","Isabella","Sophia","Charlotte","Amelia","Mia","Harper", "Evelyn","Abigail","Emily","Elizabeth","Avery","Ella","Madison","Scarlett", "Victoria","Grace","Chloe","Penelope","Riley","Lily","Eleanor","Hannah", "Lillian","Addison","Aubrey","Ellie","Stella","Natalie","Leah","Hazel", "Violet","Audrey","Claire","Lucy","Anna","Samantha","Katherine"] _NAMES_F_BLACK = ["Aaliyah","Amara","Destiny","Imani","Jasmine","Keisha","Layla","Maya","Naomi", "Nia","Raven","Serena","Tamara","Unique","Zora","Aisha","Brianna","Crystal", "Diamond","Essence","Faith","Genesis","Heaven","India","Jade","Kiara","Lashonda", "Monique","Nadia","Precious","Quiana","Regina","Shanice","Tiffany","Whitney"] _NAMES_F_HISPANIC = ["Sofia","Camila","Valentina","Isabella","Daniela","Fernanda","Gabriela","Lucia", "Maria","Ana","Carmen","Diana","Elena","Gloria","Iris","Jessica","Laura", "Linda","Margarita","Natalia","Paola","Rosa","Sandra","Teresa","Veronica", "Ximena","Yolanda","Adriana","Beatriz","Carolina","Esperanza","Francisca"] _NAMES_F_ASIAN = ["Aiko","Mei","Yuki","Sakura","Hana","Yuna","Ji-Young","Soo-Jin","Lan","Linh", "Nguyen","Phuong","Priya","Divya","Ananya","Kavya","Shreya","Sanjana", "Hui","Xin","Ying","Fang","Jing","Li","Min","Qian","Wei","Xue","Yan","Zhen"] _NAMES_M_WHITE = ["Liam","Noah","William","James","Oliver","Benjamin","Elijah","Lucas","Mason", "Logan","Alexander","Ethan","Jacob","Michael","Daniel","Henry","Jackson", "Sebastian","Aiden","Matthew","Samuel","David","Joseph","Carter","Owen", "Wyatt","John","Jack","Luke","Dylan","Grayson","Levi","Isaac","Gabriel"] _NAMES_M_BLACK = ["Andre","DeShawn","Darius","Elijah","Isaiah","Jamal","Jaylen","Jordan","Kendrick", "Malik","Marcus","Marquise","Nathaniel","Omari","Quincy","Rashad","Roderick", "Terrence","Trevon","Xavier","Zion","Aaron","Calvin","Damon","Ernest","Frederick", "Gerald","Harold","Ivan","Jerome","Kenneth","Leonard","Maurice","Nelson"] _NAMES_M_HISPANIC = ["Santiago","Mateo","Alejandro","Sebastian","Diego","Carlos","Miguel","Andres", "Fernando","Jose","Luis","Manuel","Marco","Mario","Pablo","Rafael","Ricardo", "Roberto","Rodrigo","Victor","Alberto","Arturo","Cesar","Eduardo","Ernesto", "Francisco","Guillermo","Hector","Ignacio","Javier","Juan","Lorenzo","Oscar"] _NAMES_M_ASIAN = ["Wei","Ming","Jian","Yang","Hao","Lei","Tao","Xiao","Yong","Jun","Ryu","Kenji", "Hiroshi","Takashi","Yuto","Min-Jun","Seo-Jun","Ji-Ho","Arjun","Rahul","Vikram", "Suresh","Rajesh","Anil","Vijay","Amit","Nikhil","Rohan","Kiran","Sanjay"] _LAST_NAMES_WHITE = ["Smith","Johnson","Williams","Brown","Jones","Miller","Davis","Wilson","Anderson", "Thomas","Taylor","Moore","Jackson","Martin","Lee","Thompson","White","Harris", "Clark","Lewis","Robinson","Walker","Young","Allen","King","Wright","Scott", "Green","Adams","Nelson","Baker","Hall","Campbell","Mitchell","Carter","Roberts"] _LAST_NAMES_BLACK = ["Williams","Johnson","Jones","Brown","Davis","Wilson","Thomas","Taylor","Moore", "Jackson","Harris","Thompson","White","Robinson","Walker","King","Green","Adams", "Baker","Hall","Carter","Mitchell","Peele","Banks","Bell","Boyd","Brooks","Bryant", "Byrd","Chambers","Coleman","Collins","Cooper","Crawford","Dixon","Edwards"] _LAST_NAMES_HISPANIC = ["Garcia","Rodriguez","Martinez","Hernandez","Lopez","Gonzalez","Perez","Sanchez", "Ramirez","Torres","Flores","Rivera","Gomez","Diaz","Reyes","Morales","Cruz", "Gutierrez","Ortiz","Chavez","Ramos","Romero","Vargas","Castillo","Jimenez", "Moreno","Alvarez","Mendoza","Ruiz","Aguilar","Vega","Castro","Medina"] _LAST_NAMES_ASIAN = ["Wang","Li","Zhang","Liu","Chen","Yang","Huang","Zhao","Wu","Zhou","Kim","Park", "Lee","Choi","Jung","Nguyen","Tran","Le","Pham","Hoang","Patel","Shah","Kumar", "Singh","Sharma","Gupta","Mehta","Kapoor","Nair","Reddy","Iyer","Rao","Joshi"] # Ethnic distribution approximating US cancer patient demographics (ACS 2024) _ETHNICITY_GROUPS = [ ("White", 0.60, _NAMES_F_WHITE, _NAMES_M_WHITE, _LAST_NAMES_WHITE), ("Black or African American", 0.13, _NAMES_F_BLACK, _NAMES_M_BLACK, _LAST_NAMES_BLACK), ("Hispanic or Latino", 0.14, _NAMES_F_HISPANIC, _NAMES_M_HISPANIC, _LAST_NAMES_HISPANIC), ("Asian", 0.07, _NAMES_F_ASIAN, _NAMES_M_ASIAN, _LAST_NAMES_ASIAN), ("American Indian or Alaska Native", 0.03, _NAMES_F_WHITE, _NAMES_M_WHITE, _LAST_NAMES_WHITE), ("Native Hawaiian or Pacific Islander", 0.01, _NAMES_F_ASIAN, _NAMES_M_ASIAN, _LAST_NAMES_ASIAN), ("Other / Multiracial", 0.02, _NAMES_F_WHITE, _NAMES_M_WHITE, _LAST_NAMES_WHITE), ] _ETH_NAMES = [(e[0], e[2], e[3], e[4]) for e in _ETHNICITY_GROUPS] _ETH_WEIGHTS = [e[1] for e in _ETHNICITY_GROUPS] # City pool weighted by US metropolitan population (2020 Census) _CITIES = [ ("New York","NY",0.060),("Los Angeles","CA",0.045),("Chicago","IL",0.033), ("Houston","TX",0.027),("Phoenix","AZ",0.020),("Philadelphia","PA",0.018), ("San Antonio","TX",0.016),("San Diego","CA",0.016),("Dallas","TX",0.015), ("San Jose","CA",0.013),("Austin","TX",0.013),("Jacksonville","FL",0.011), ("Fort Worth","TX",0.010),("Columbus","OH",0.010),("Charlotte","NC",0.010), ("Indianapolis","IN",0.009),("San Francisco","CA",0.009),("Seattle","WA",0.009), ("Denver","CO",0.009),("Nashville","TN",0.009),("Boston","MA",0.009), ("Baltimore","MD",0.008),("Louisville","KY",0.007),("Portland","OR",0.007), ("Las Vegas","NV",0.007),("Milwaukee","WI",0.006),("Albuquerque","NM",0.006), ("Tucson","AZ",0.006),("Fresno","CA",0.005),("Sacramento","CA",0.005), ("Atlanta","GA",0.009),("Kansas City","MO",0.005),("Omaha","NE",0.004), ("Raleigh","NC",0.005),("Cleveland","OH",0.005),("Minneapolis","MN",0.006), ("Miami","FL",0.008),("Tampa","FL",0.007),("New Orleans","LA",0.005), ("Pittsburgh","PA",0.006),("Memphis","TN",0.005),("Richmond","VA",0.004), ("Birmingham","AL",0.004),("Salt Lake City","UT",0.004),("Hartford","CT",0.004), ("Buffalo","NY",0.004),("Rochester","NY",0.003),("Providence","RI",0.003), ("Des Moines","IA",0.003),("Little Rock","AR",0.003),("Madison","WI",0.003), ] _CITY_NAMES = [(c[0], c[1]) for c in _CITIES] _CITY_WEIGHTS = [c[2] for c in _CITIES] # Comorbidity prevalence in US oncology patients (literature-based) _COMORBIDITY_POOL = [ ("Type 2 Diabetes", 0.18), ("Hypertension", 0.42), ("Coronary Artery Disease",0.09), ("COPD", 0.08), ("Chronic Kidney Disease", 0.12), ("Obesity (BMI>30)", 0.36), ("Depression/Anxiety", 0.22), ("Hypothyroidism", 0.07), ("Atrial Fibrillation", 0.05), ("Osteoporosis", 0.06), ] # Insurance status (US cancer patient distribution, KFF 2023) _INSURANCE = [ ("Private/Employer", 0.48), ("Medicare", 0.30), ("Medicaid", 0.14), ("Uninsured", 0.05), ("VA/Military", 0.03), ] _INS_LABELS = [i[0] for i in _INSURANCE] _INS_WEIGHTS = [i[1] for i in _INSURANCE] # ECOG score distribution varies by condition severity _ECOG_BY_CONDITION: dict[str, list[float]] = { # [P(0), P(1), P(2), P(3)] "breast cancer": [0.35, 0.40, 0.18, 0.07], "prostate cancer": [0.30, 0.40, 0.20, 0.10], "non-small cell lung cancer": [0.20, 0.38, 0.28, 0.14], "colorectal cancer": [0.28, 0.40, 0.22, 0.10], "ovarian cancer": [0.25, 0.40, 0.25, 0.10], "melanoma": [0.40, 0.38, 0.15, 0.07], "leukemia": [0.25, 0.38, 0.25, 0.12], "lymphoma": [0.28, 0.40, 0.22, 0.10], "glioblastoma": [0.15, 0.35, 0.30, 0.20], "pancreatic cancer": [0.15, 0.32, 0.33, 0.20], "bladder cancer": [0.28, 0.40, 0.22, 0.10], "renal cell carcinoma": [0.32, 0.40, 0.20, 0.08], "thyroid cancer": [0.50, 0.35, 0.12, 0.03], "multiple myeloma": [0.22, 0.38, 0.28, 0.12], "endometrial cancer": [0.30, 0.40, 0.22, 0.08], "cervical cancer": [0.25, 0.40, 0.25, 0.10], "gastric cancer": [0.18, 0.35, 0.30, 0.17], "hepatocellular carcinoma": [0.15, 0.32, 0.33, 0.20], "head and neck cancer": [0.20, 0.38, 0.28, 0.14], "sarcoma": [0.30, 0.40, 0.22, 0.08], } # ── Condition profiles (SEER-weighted) ─────────────────────────────────────── # count_weight → how many of the 100 k total patients come from this condition # biomarker_prevalences → {biomarker_id: probability} (TCGA / literature) _CONDITION_PROFILES: dict[str, dict] = { "breast cancer": { "icd10_prefix": "C50", "sex": "FEMALE", "count_weight": 0.155, "age_range": (25, 82), "age_mode": 62, "stages": ["I","II","III","IV"], "stage_weights": [0.28, 0.32, 0.25, 0.15], "biomarker_prevalences": { "ER_POS":0.75,"PR_POS":0.65,"HER2_POS":0.17,"HER2_NEG":0.83, "TNBC":0.12,"BRCA1_MUT":0.05,"BRCA2_MUT":0.04, "PIK3CA_MUT":0.35,"TP53_MUT":0.28, }, "med_pool": ["trastuzumab","bevacizumab","capecitabine","olaparib","pembrolizumab"], "prior_chemo_rate": 0.65, }, "non-small cell lung cancer": { "icd10_prefix": "C34", "sex": "ALL", "count_weight": 0.130, "age_range": (40, 84), "age_mode": 68, "stages": ["I","II","III","IV"], "stage_weights": [0.09, 0.12, 0.28, 0.51], "biomarker_prevalences": { "EGFR_L858R":0.08,"EGFR_DEL19":0.09,"EGFR_T790M":0.05, "ALK_FUSION":0.04,"ROS1_FUSION":0.02,"MET_EX14":0.03, "KRAS_G12C":0.13,"PDL1_HIGH":0.28,"PDL1_LOW":0.30,"PDL1_NEG":0.42, }, "med_pool": ["osimertinib","pembrolizumab","nivolumab","erlotinib","atezolizumab","durvalumab"], "prior_chemo_rate": 0.55, }, "prostate cancer": { "icd10_prefix": "C61", "sex": "MALE", "count_weight": 0.095, "age_range": (45, 86), "age_mode": 67, "stages": ["I","II","III","IV"], "stage_weights": [0.18, 0.28, 0.28, 0.26], "biomarker_prevalences": { "PSA_ELEVATED":0.90,"BRCA2_MUT":0.05,"PTEN_LOSS":0.25,"AR_V7":0.20, }, "med_pool": ["enzalutamide","bevacizumab","olaparib","pembrolizumab"], "prior_chemo_rate": 0.40, }, "colorectal cancer": { "icd10_prefix": "C18", "sex": "ALL", "count_weight": 0.085, "age_range": (35, 82), "age_mode": 65, "stages": ["I","II","III","IV"], "stage_weights": [0.18, 0.26, 0.30, 0.26], "biomarker_prevalences": { "MSI_H":0.10,"MSS":0.90,"KRAS_WT":0.42, "BRAF_V600E":0.08,"NRAS_MUT":0.05,"KRAS_G12C":0.04, }, "med_pool": ["bevacizumab","cetuximab","capecitabine","pembrolizumab"], "prior_chemo_rate": 0.60, }, "melanoma": { "icd10_prefix": "C43", "sex": "ALL", "count_weight": 0.055, "age_range": (20, 80), "age_mode": 57, "stages": ["I","II","III","IV"], "stage_weights": [0.30, 0.28, 0.22, 0.20], "biomarker_prevalences": { "BRAF_V600E":0.45,"BRAF_V600K":0.06,"TMB_HIGH":0.35,"NRAS_MEL":0.20, }, "med_pool": ["pembrolizumab","nivolumab","dabrafenib","vemurafenib","ipilimumab"], "prior_chemo_rate": 0.30, }, "bladder cancer": { "icd10_prefix": "C67", "sex": "ALL", "count_weight": 0.045, "age_range": (45, 85), "age_mode": 69, "stages": ["I","II","III","IV"], "stage_weights": [0.28, 0.24, 0.26, 0.22], "biomarker_prevalences": { "FGFR3_MUT":0.20,"PDL1_HIGH":0.22,"TMB_HIGH":0.15,"TP53_MUT":0.30, }, "med_pool": ["pembrolizumab","atezolizumab","nivolumab","erdafitinib"], "prior_chemo_rate": 0.45, }, "renal cell carcinoma": { "icd10_prefix": "C64", "sex": "ALL", "count_weight": 0.042, "age_range": (40, 82), "age_mode": 64, "stages": ["I","II","III","IV"], "stage_weights": [0.25, 0.20, 0.25, 0.30], "biomarker_prevalences": { "VHL_LOSS":0.55,"MTOR_MUT":0.15,"PDL1_HIGH":0.18, }, "med_pool": ["pembrolizumab","nivolumab","bevacizumab","sunitinib"], "prior_chemo_rate": 0.25, }, "lymphoma": { "icd10_prefix": "C85", "sex": "ALL", "count_weight": 0.042, "age_range": (20, 80), "age_mode": 58, "stages": ["I","II","III","IV"], "stage_weights": [0.20, 0.25, 0.30, 0.25], "biomarker_prevalences": { "CD20_POS":0.85,"EZH2_MUT":0.22,"TMB_HIGH":0.12,"PDL1_HIGH":0.15, }, "med_pool": ["rituximab","pembrolizumab","nivolumab"], "prior_chemo_rate": 0.55, }, "endometrial cancer": { "icd10_prefix": "C54", "sex": "FEMALE", "count_weight": 0.038, "age_range": (40, 82), "age_mode": 63, "stages": ["I","II","III","IV"], "stage_weights": [0.50, 0.15, 0.20, 0.15], "biomarker_prevalences": { "MSI_H":0.25,"POLE_MUT":0.07,"CTNNB1_MUT":0.30,"TP53_MUT":0.25,"PIK3CA_MUT":0.35, }, "med_pool": ["pembrolizumab","bevacizumab","olaparib","capecitabine"], "prior_chemo_rate": 0.40, }, "leukemia": { "icd10_prefix": "C91", "sex": "ALL", "count_weight": 0.035, "age_range": (18, 82), "age_mode": 55, "stages": ["I","II","III","IV"], "stage_weights": [0.25, 0.25, 0.28, 0.22], "biomarker_prevalences": { "BCR_ABL1":0.30,"FLT3_ITD":0.25,"NPM1_MUT":0.30,"TP53_MUT":0.15, }, "med_pool": ["imatinib","rituximab","pembrolizumab"], "prior_chemo_rate": 0.60, }, "pancreatic cancer": { "icd10_prefix": "C25", "sex": "ALL", "count_weight": 0.033, "age_range": (40, 82), "age_mode": 68, "stages": ["I","II","III","IV"], "stage_weights": [0.05, 0.12, 0.28, 0.55], "biomarker_prevalences": { "KRAS_G12C":0.07,"BRCA2_MUT":0.06,"TP53_MUT":0.55,"MSI_H":0.02, }, "med_pool": ["capecitabine","erlotinib","olaparib"], "prior_chemo_rate": 0.50, }, "thyroid cancer": { "icd10_prefix": "C73", "sex": "FEMALE", "count_weight": 0.030, "age_range": (20, 75), "age_mode": 47, "stages": ["I","II","III","IV"], "stage_weights": [0.55, 0.20, 0.15, 0.10], "biomarker_prevalences": { "BRAF_THYROID":0.45,"RET_FUSION":0.08,"NTRK_FUSION":0.05, }, "med_pool": ["pembrolizumab","dabrafenib","vemurafenib"], "prior_chemo_rate": 0.15, }, "multiple myeloma": { "icd10_prefix": "C90", "sex": "ALL", "count_weight": 0.025, "age_range": (45, 84), "age_mode": 67, "stages": ["I","II","III","IV"], "stage_weights": [0.20, 0.28, 0.30, 0.22], "biomarker_prevalences": { "WHSC1_MUT":0.20,"CDKN2A_LOSS":0.30,"TP53_MUT":0.15, }, "med_pool": ["pembrolizumab","rituximab","bevacizumab"], "prior_chemo_rate": 0.65, }, "gastric cancer": { "icd10_prefix": "C16", "sex": "ALL", "count_weight": 0.018, "age_range": (35, 82), "age_mode": 65, "stages": ["I","II","III","IV"], "stage_weights": [0.10, 0.20, 0.35, 0.35], "biomarker_prevalences": { "ERBB2_GC":0.15,"MSI_H":0.10,"PDL1_HIGH":0.20,"TP53_MUT":0.40, }, "med_pool": ["trastuzumab","pembrolizumab","nivolumab","capecitabine"], "prior_chemo_rate": 0.55, }, "ovarian cancer": { "icd10_prefix": "C56", "sex": "FEMALE", "count_weight": 0.018, "age_range": (35, 80), "age_mode": 62, "stages": ["I","II","III","IV"], "stage_weights": [0.12, 0.14, 0.40, 0.34], "biomarker_prevalences": { "BRCA1_MUT":0.12,"BRCA2_MUT":0.08,"TP53_MUT":0.60,"PIK3CA_MUT":0.08, }, "med_pool": ["olaparib","bevacizumab","pembrolizumab"], "prior_chemo_rate": 0.75, }, "hepatocellular carcinoma": { "icd10_prefix": "C22", "sex": "ALL", "count_weight": 0.015, "age_range": (35, 80), "age_mode": 62, "stages": ["I","II","III","IV"], "stage_weights": [0.10, 0.18, 0.32, 0.40], "biomarker_prevalences": { "HBV_POS":0.25,"TERT_MUT":0.55,"TP53_MUT":0.20,"CTNNB1_MUT":0.25, }, "med_pool": ["pembrolizumab","nivolumab","bevacizumab","atezolizumab"], "prior_chemo_rate": 0.35, }, "glioblastoma": { "icd10_prefix": "C71", "sex": "ALL", "count_weight": 0.012, "age_range": (30, 76), "age_mode": 62, "stages": ["III","IV"], "stage_weights": [0.28, 0.72], "biomarker_prevalences": { "IDH1_WT":0.90,"IDH1_R132H":0.10,"MGMT_METH":0.45, "EGFR_AMP":0.40,"TP53_MUT":0.25, }, "med_pool": ["bevacizumab","pembrolizumab"], "prior_chemo_rate": 0.70, }, "head and neck cancer": { "icd10_prefix": "C10", "sex": "ALL", "count_weight": 0.012, "age_range": (30, 80), "age_mode": 60, "stages": ["I","II","III","IV"], "stage_weights": [0.10, 0.15, 0.30, 0.45], "biomarker_prevalences": { "HPV_HNSC":0.60,"PIK3CA_HNC":0.25,"PDL1_HIGH":0.20,"TP53_MUT":0.45, }, "med_pool": ["pembrolizumab","nivolumab","cetuximab"], "prior_chemo_rate": 0.55, }, "cervical cancer": { "icd10_prefix": "C53", "sex": "FEMALE", "count_weight": 0.008, "age_range": (20, 72), "age_mode": 48, "stages": ["I","II","III","IV"], "stage_weights": [0.28, 0.25, 0.25, 0.22], "biomarker_prevalences": { "HPV_POS":0.99,"PDL1_HIGH":0.25,"PIK3CA_MUT":0.25, }, "med_pool": ["pembrolizumab","bevacizumab","nivolumab"], "prior_chemo_rate": 0.50, }, "sarcoma": { "icd10_prefix": "C49", "sex": "ALL", "count_weight": 0.007, "age_range": (15, 75), "age_mode": 45, "stages": ["I","II","III","IV"], "stage_weights": [0.20, 0.25, 0.30, 0.25], "biomarker_prevalences": { "CDK4_AMP":0.20,"MDM2_AMP":0.18,"TP53_MUT":0.25, }, "med_pool": ["pembrolizumab","nivolumab","bevacizumab"], "prior_chemo_rate": 0.45, }, } random.seed(42) # reproducible synthetic data def _parse_age(age_str: str) -> int | None: if not age_str: return None try: return int(age_str.split()[0]) except Exception: return None def _skewed_age(age_range: tuple[int, int], mode: int) -> int: """Triangle-distributed age reflecting real incidence peak.""" lo, hi = age_range mode = max(lo, min(hi, mode)) return int(random.triangular(lo, hi, mode)) def _pick_biomarkers(prevalences: dict[str, float], rng: random.Random) -> list[str]: """Independent Bernoulli draw per biomarker based on literature prevalence.""" return [bm for bm, p in prevalences.items() if rng.random() < p] def _pick_comorbidities(rng: random.Random, age: int) -> list[str]: """Age-scaled comorbidity draw.""" scale = 1.0 + max(0, (age - 50)) * 0.015 # comorbidities rise ~1.5% per year after 50 return [c for c, p in _COMORBIDITY_POOL if rng.random() < min(p * scale, 0.95)] def _generate_patient(pid: str, condition: str, profile: dict, seq: int, rng: random.Random) -> dict: sex_raw = profile["sex"] sex = rng.choice(["MALE","FEMALE"]) if sex_raw == "ALL" else sex_raw age = _skewed_age(profile["age_range"], profile["age_mode"]) stage = rng.choices(profile["stages"], weights=profile["stage_weights"])[0] ecog_weights = _ECOG_BY_CONDITION.get(condition, [0.28, 0.40, 0.22, 0.10]) ecog = rng.choices([0, 1, 2, 3], weights=ecog_weights)[0] eth_group = rng.choices(_ETH_NAMES, weights=_ETH_WEIGHTS)[0] ethnicity, names_f, names_m, last_names = eth_group first = rng.choice(names_f if sex == "FEMALE" else names_m) last = rng.choice(last_names) city, state = rng.choices(_CITY_NAMES, weights=_CITY_WEIGHTS)[0] insurance = rng.choices(_INS_LABELS, weights=_INS_WEIGHTS)[0] biomarkers = _pick_biomarkers(profile["biomarker_prevalences"], rng) comorbidities = _pick_comorbidities(rng, age) med_pool = profile["med_pool"] n_med = min(rng.randint(1, 2), len(med_pool)) medications = rng.sample(med_pool, n_med) prior_chemo = rng.random() < profile.get("prior_chemo_rate", 0.5) prior_radiation = rng.random() < 0.35 prior_surgery = rng.random() < 0.50 prior_lines = rng.randint(0, 3) if prior_chemo else 0 return { "id": pid, "name": f"{first} {last}", "age": age, "sex": sex, "stage": stage, "ecog": ecog, "condition": condition, "icd10_prefix": profile["icd10_prefix"], "city": city, "state": state, "ethnicity": ethnicity, "insurance": insurance, "biomarkers": biomarkers, "medications": medications, "comorbidities": comorbidities, "prior_chemo": prior_chemo, "prior_radiation": prior_radiation, "prior_surgery": prior_surgery, "prior_lines_of_therapy": prior_lines, "source": "synthetic_v2", } # ── Batch write helpers ─────────────────────────────────────────────────────── _BATCH_SIZE = 500 def _batch_write_patients(patients: list[dict]) -> None: neo4j_conn.run_query(""" UNWIND $patients AS p MERGE (n:Patient {id: p.id}) SET n += { name: p.name, age: p.age, sex: p.sex, stage: p.stage, ecog: p.ecog, condition: p.condition, icd10_prefix: p.icd10_prefix, city: p.city, state: p.state, ethnicity: p.ethnicity, insurance: p.insurance, biomarkers: p.biomarkers, medications: p.medications, comorbidities: p.comorbidities, prior_chemo: p.prior_chemo, prior_radiation: p.prior_radiation, prior_surgery: p.prior_surgery, prior_lines_of_therapy: p.prior_lines_of_therapy, source: p.source, updated_at: datetime() } """, {"patients": patients}) def _batch_write_biomarker_links(links: list[dict]) -> None: neo4j_conn.run_query(""" UNWIND $links AS l MATCH (p:Patient {id: l.pid}) MATCH (b:Biomarker {id: l.bm_id}) MERGE (p)-[:HAS_BIOMARKER]->(b) """, {"links": links}) def _batch_write_diagnosis_links(links: list[dict]) -> None: # links already have resolved diagnosis_code (exact match, no scan needed) neo4j_conn.run_query(""" UNWIND $links AS l MATCH (p:Patient {id: l.pid}) MATCH (d:Diagnosis {code: l.diagnosis_code}) MERGE (p)-[:HAS_DIAGNOSIS]->(d) """, {"links": links}) def _batch_write_eligibility(edges: list[dict]) -> None: neo4j_conn.run_query(""" UNWIND $edges AS e MATCH (p:Patient {id: e.pid}) MATCH (t:Trial {id: e.tid}) MERGE (p)-[r:ELIGIBLE_FOR]->(t) SET r.score = e.score, r.matched_at = datetime() """, {"edges": edges}) # ── Main patient seeder ─────────────────────────────────────────────────────── def seed_patients_and_eligibility(total_patients: int = 100_000) -> int: print(f"\n[6/6] Generating {total_patients:,} clinically-informed synthetic patients...") print(" (SEER incidence weights · TCGA biomarker prevalence · US Census demographics)") # Pre-load trials grouped by condition trial_rows = neo4j_conn.run_query(""" MATCH (t:Trial {status: 'RECRUITING'}) RETURN t.id AS id, t.condition AS condition, t.sex AS sex, t.min_age AS min_age, t.max_age AS max_age """) trials_by_condition: dict[str, list[dict]] = {} for row in (trial_rows or []): cond = (row.get("condition") or "").lower().strip() trials_by_condition.setdefault(cond, []).append(row) # Calculate per-condition counts from SEER weights total_weight = sum(p["count_weight"] for p in _CONDITION_PROFILES.values()) condition_counts = { cond: max(1, round(total_patients * prof["count_weight"] / total_weight)) for cond, prof in _CONDITION_PROFILES.items() } # Adjust rounding error so we hit exactly total_patients allocated = sum(condition_counts.values()) diff = total_patients - allocated largest = max(condition_counts, key=lambda c: condition_counts[c]) condition_counts[largest] += diff # Pre-load one canonical Diagnosis code per ICD-10 prefix all_prefixes = list({p["icd10_prefix"] for p in _CONDITION_PROFILES.values()}) dx_canon: dict[str, str] = {} for prefix in all_prefixes: rows = neo4j_conn.run_query( "MATCH (d:Diagnosis) WHERE d.code STARTS WITH $p RETURN d.code AS code ORDER BY d.code LIMIT 1", {"p": prefix} ) if rows: dx_canon[prefix] = rows[0]["code"] # Check existing patients per condition to allow resume existing_rows = neo4j_conn.run_query(""" MATCH (p:Patient) WHERE p.source = 'synthetic_v2' RETURN p.condition AS condition, count(p) AS cnt """) existing_by_condition: dict[str, int] = { r["condition"]: r["cnt"] for r in (existing_rows or []) if r.get("condition") } rng = random.Random(42) grand_total = 0 grand_edges = 0 for condition, profile in _CONDITION_PROFILES.items(): icd_prefix = profile["icd10_prefix"] n = condition_counts[condition] already = existing_by_condition.get(condition, 0) condition_trials = trials_by_condition.get(condition, []) if already >= n: print(f" {condition}: {n:,} patients — already done, skipping") grand_total += n # advance RNG to stay deterministic for _ in range(n): rng.random() continue skip = already todo = n - skip print(f" {condition}: {n:,} patients ({len(condition_trials)} trials)" + (f" [resuming from {skip:,}]" if skip else "")) patient_batch: list[dict] = [] bm_links: list[dict] = [] dx_links: list[dict] = [] elig_edges: list[dict] = [] # Advance RNG past already-written patients so IDs/values stay consistent for _ in range(skip): rng.random() condition_written = 0 for i in range(skip, n): pid = f"P_{icd_prefix}_{grand_total + i + 1:06d}" p = _generate_patient(pid, condition, profile, i, rng) patient_batch.append(p) if icd_prefix in dx_canon: dx_links.append({"pid": pid, "diagnosis_code": dx_canon[icd_prefix]}) for bm in p["biomarkers"]: bm_links.append({"pid": pid, "bm_id": bm}) # Eligibility edges — apply sex/age/ECOG filters for trial in condition_trials: t_sex = (trial.get("sex") or "ALL").upper() t_min = _parse_age(trial.get("min_age") or "") t_max = _parse_age(trial.get("max_age") or "") if t_sex not in ("ALL", "BOTH", p["sex"]): continue if t_min is not None and p["age"] < t_min: continue if t_max is not None and p["age"] > t_max: continue if p["ecog"] > 2: continue base = rng.uniform(0.55, 0.90) bm_bonus = 0.08 if p["biomarkers"] else 0.0 score = round(min(base + bm_bonus, 0.99), 2) elig_edges.append({"pid": pid, "tid": trial["id"], "score": score}) condition_written += 1 # Flush batches if len(patient_batch) >= _BATCH_SIZE: _batch_write_patients(patient_batch) _batch_write_diagnosis_links(dx_links) if bm_links: _batch_write_biomarker_links(bm_links) if elig_edges: _batch_write_eligibility(elig_edges) grand_edges += len(elig_edges) patient_batch, dx_links, bm_links, elig_edges = [], [], [], [] # Flush remainder if patient_batch: _batch_write_patients(patient_batch) _batch_write_diagnosis_links(dx_links) if bm_links: _batch_write_biomarker_links(bm_links) if elig_edges: _batch_write_eligibility(elig_edges) grand_edges += len(elig_edges) grand_total += n print(f" ↳ wrote {condition_written:,} patients | total so far: {grand_total:,}/{total_patients:,} | edges: {grand_edges:,}") print(f"\n ✓ Total patients: {grand_total:,}") print(f" ✓ Total ELIGIBLE_FOR edges: {grand_edges:,}") return grand_total # ── Main entry point ────────────────────────────────────────────────────────── async def run_seeder(conditions: list[str] | None = None): start = time.time() print("=" * 60) print("ClinicalMatch AI — Graph Seeder v2") print("100 k synthetic patients · 20 oncology conditions") print("=" * 60) async with httpx.AsyncClient(headers={"User-Agent": "ClinicalMatchAI/2.0 (hackathon@research.org)"}) as client: n_trials = await seed_trials(client) n_meds = await seed_medications(client) n_dx = await seed_diagnoses(client) n_pubs = await seed_literature(client) n_bm = seed_biomarkers() derive_eligibility_relationships() n_patients = seed_patients_and_eligibility(total_patients=100_000) elapsed = time.time() - start print(f"\n{'=' * 60}") print(f"Seeding complete in {elapsed / 60:.1f} min") print(f" Trials: {n_trials}") print(f" Medications: {n_meds}") print(f" Diagnoses: {n_dx}") print(f" Publications: {n_pubs}") print(f" Biomarkers: {n_bm}") print(f" Patients: {n_patients:,}") print("=" * 60) def seed_sync(): asyncio.run(run_seeder()) if __name__ == "__main__": import sys conditions = sys.argv[1:] if len(sys.argv) > 1 else None asyncio.run(run_seeder(conditions))