| """ |
| HC-ONC-007: Leukemia Synthetic Dataset — Simulation Engine |
| XpertSystems.ai | SKU: HC-ONC-007 | Version 1.0.0 | April 2026 |
| HIPAA-Safe Synthetic Data — No Real Patient Records |
| |
| Generates biologically consistent leukemia patient cohort covering: |
| - AML, ALL, CML, CLL, APL subtypes with WHO 2022 classification |
| - ELN 2022 cytogenetic/molecular risk stratification |
| - Treatment protocols: 7+3, HyperCVAD, Venetoclax+AZA, ATRA+ATO, TKI |
| - MRD response trajectory modeling |
| - Bone marrow transplant (HSCT) outcomes calibrated to CIBMTR benchmarks |
| |
| Usage: |
| python hc_onc_007_simulation_engine.py --n_patients 25000 --seed 42 --output_dir ./output |
| """ |
|
|
| import argparse |
| import os |
| import json |
| import numpy as np |
| import pandas as pd |
| from scipy.stats import weibull_min |
| from datetime import datetime, timedelta |
| import uuid |
|
|
| |
| |
| |
|
|
| SUBTYPE_DIST = { |
| "AML": 0.35, "ALL": 0.25, "CML": 0.20, "CLL": 0.15, "APL": 0.05 |
| } |
|
|
| |
| ELN_RISK_DIST = { |
| "Favorable": 0.30, "Intermediate": 0.40, "Adverse": 0.30 |
| } |
|
|
| |
| CR_RATE = { |
| "AML_Favorable": 0.82, "AML_Intermediate": 0.65, "AML_Adverse": 0.45, |
| "ALL_Pediatric": 0.95, "ALL_Adult": 0.85, "APL": 0.97, |
| "CML": None, |
| "CLL": None, |
| } |
|
|
| |
| OS_LAMBDA = { |
| "AML_Favorable": 60.0, "AML_Intermediate": 36.0, "AML_Adverse": 14.0, |
| "ALL_Pediatric": 110.0, "ALL_Adult": 40.0, "APL": 88.0, |
| "CML": 130.0, "CLL": 120.0 |
| } |
|
|
| |
| AGVHD_RATE = {"MRD": 0.30, "MUD": 0.45, "MMUD": 0.55, "Haploidentical": 0.48, "Cord_Blood": 0.35} |
|
|
| |
| NRM_RATE = {"MRD": 0.10, "MUD": 0.17, "MMUD": 0.22, "Haploidentical": 0.18, "Cord_Blood": 0.20} |
|
|
| INDUCTION_AML_REGIMENS = [ |
| "7+3_Daunorubicin", "7+3_Idarubicin", "CPX-351", "Venetoclax+AZA", |
| "Venetoclax+LDAC", "Midostaurin+7+3", "Enasidenib+AZA", "Ivosidenib+AZA", |
| "Glasdegib+LDAC", "BSC" |
| ] |
| INDUCTION_ALL_REGIMENS = [ |
| "CALGB_10403", "HyperCVAD_A_B", "BFM_Protocol", "GRAALL_2005", |
| "UKALL14", "Blinatumomab+Chemo", "Inotuzumab_Ozogamicin+Chemo", |
| "Dasatinib+Steroids", "Ponatinib+Chemo" |
| ] |
| CML_REGIMENS = [ |
| "Imatinib_400mg", "Dasatinib_100mg", "Nilotinib_300mg_BID", |
| "Bosutinib_400mg", "Ponatinib_45mg", "Asciminib_40mg_BID" |
| ] |
| CLL_REGIMENS = [ |
| "Watch_and_Wait", "FCR", "BR", "Ibrutinib", "Acalabrutinib", |
| "Venetoclax_Obinutuzumab", "Venetoclax_Rituximab", "Pirtobrutinib", |
| "Lisocabtagene_Maraleucel" |
| ] |
|
|
| DONOR_TYPES = ["MRD", "MUD", "MMUD", "Haploidentical", "Cord_Blood"] |
| DONOR_DIST = [0.35, 0.38, 0.12, 0.10, 0.05] |
|
|
| CONDITIONING = { |
| "MAC": ["BuCy", "TBI_Cy", "FluBu4", "Cy_TBI"], |
| "RIC": ["FluBu2", "FluMel140", "TBI_Flu_200"], |
| "NMA": ["Flu_Cy_TBI_200", "Flu_Cy"] |
| } |
|
|
| GVHD_PROPHYLAXIS = [ |
| "Tacrolimus_MTX", "CsA_MTX", "PT_Cy_Tacrolimus_MMF", |
| "PT_Cy_Alone", "ATG_CsA_MTX" |
| ] |
|
|
| |
| |
| |
|
|
| def clamp(val, lo, hi): |
| return max(lo, min(hi, val)) |
|
|
| def weibull_sample(rng, k, lam, size=1): |
| """Sample from Weibull distribution.""" |
| return weibull_min.rvs(k, scale=lam, size=size, random_state=rng) |
|
|
| def choice(rng, options, probs=None): |
| probs = np.array(probs) if probs else None |
| if probs is not None: |
| probs = probs / probs.sum() |
| return rng.choice(options, p=probs) |
|
|
| def lognorm(rng, mu, sigma, lo, hi): |
| v = np.exp(rng.normal(mu, sigma)) |
| return clamp(v, lo, hi) |
|
|
| |
| |
| |
|
|
| def generate_demographics(rng, n): |
| """Generate demographic attributes.""" |
| subtypes = rng.choice(list(SUBTYPE_DIST.keys()), size=n, p=list(SUBTYPE_DIST.values())) |
|
|
| ages = [] |
| for st in subtypes: |
| if st == "ALL": |
| |
| if rng.random() < 0.55: |
| a = rng.normal(6, 3) |
| else: |
| a = rng.normal(35, 12) |
| elif st == "CLL": |
| a = rng.normal(68, 9) |
| elif st == "CML": |
| a = rng.normal(52, 14) |
| elif st == "APL": |
| a = rng.normal(44, 13) |
| else: |
| a = rng.normal(62, 14) |
| ages.append(clamp(a, 0.5, 95)) |
|
|
| ages = np.array(ages) |
| sex = rng.choice(["Male", "Female"], size=n, p=[0.55, 0.45]) |
| race = rng.choice( |
| ["White", "Black", "Hispanic", "Asian", "Other"], |
| size=n, p=[0.65, 0.12, 0.13, 0.08, 0.02] |
| ) |
| ecog = rng.choice([0, 1, 2, 3, 4], size=n, p=[0.25, 0.35, 0.25, 0.12, 0.03]) |
|
|
| return pd.DataFrame({ |
| "patient_id": [str(uuid.UUID(int=rng.integers(0, 2**128))) for _ in range(n)], |
| "leukemia_type": subtypes, |
| "age_at_diagnosis": np.round(ages, 1), |
| "pediatric_flag": (ages < 18).astype(int), |
| "sex": sex, |
| "race": race, |
| "ecog_performance_status": ecog, |
| }) |
|
|
| |
| |
| |
|
|
| def generate_disease_characteristics(rng, demo): |
| n = len(demo) |
| rows = [] |
|
|
| for i, row in demo.iterrows(): |
| st = row["leukemia_type"] |
| age = row["age_at_diagnosis"] |
| d = {} |
|
|
| |
| if st == "AML": |
| d["who_2022_classification"] = choice(rng, [ |
| "AML_NPM1", "AML_FLT3_ITD", "AML_CEBPA_biallelic", |
| "AML_NUP98", "AML_t8_21", "AML_inv16", "AML_MRC", |
| "AML_TP53", "AML_IDH2", "AML_IDH1", "AML_NOS" |
| ]) |
| d["fab_classification"] = choice(rng, ["M0","M1","M2","M3","M4","M5","M6","M7"], |
| [0.05,0.15,0.25,0.08,0.20,0.15,0.05,0.07]) |
| elif st == "APL": |
| d["who_2022_classification"] = "APL_PML_RARA" |
| d["fab_classification"] = "M3" |
| elif st == "ALL": |
| is_ped = age < 18 |
| d["who_2022_classification"] = choice(rng, |
| ["B-ALL_ETV6_RUNX1", "B-ALL_Hyperdiploidy", "B-ALL_Hypodiploidy", |
| "B-ALL_BCR_ABL1", "B-ALL_KMT2A", "B-ALL_NOS", |
| "T-ALL_NOS", "T-ALL_ETP"], |
| [0.25 if is_ped else 0.05, |
| 0.25 if is_ped else 0.10, |
| 0.02, 0.05 if is_ped else 0.25, |
| 0.08, 0.20, 0.10, 0.05]) |
| d["fab_classification"] = choice(rng, ["L1","L2","L3"], [0.50,0.40,0.10]) |
| elif st == "CML": |
| d["who_2022_classification"] = choice(rng, |
| ["CML_CP", "CML_AP", "CML_BP_Myeloid", "CML_BP_Lymphoid"], |
| [0.85, 0.08, 0.05, 0.02]) |
| d["fab_classification"] = "N/A" |
| else: |
| d["who_2022_classification"] = choice(rng, |
| ["CLL_IGHV_Mutated", "CLL_IGHV_Unmutated", "SLL"], [0.50, 0.40, 0.10]) |
| d["fab_classification"] = "N/A" |
|
|
| d["disease_phase"] = choice(rng, |
| ["Newly_Diagnosed", "Relapsed", "Refractory", "Relapsed_Refractory", "MRD_Positive"], |
| [0.70, 0.12, 0.08, 0.07, 0.03]) |
| d["de_novo_vs_secondary"] = choice(rng, |
| ["de_novo", "secondary_to_MDS", "therapy_related"], |
| [0.80, 0.12, 0.08]) if st in ["AML","APL"] else "N/A" |
|
|
| |
| if st in ["AML","APL"]: |
| d["blast_pct_bone_marrow"] = clamp(rng.normal(55, 20), 20, 98) |
| d["blast_pct_peripheral_blood"] = clamp(d["blast_pct_bone_marrow"] * rng.uniform(0.3, 0.9), 0, 95) |
| d["wbc_k_ul"] = lognorm(rng, 3.2, 1.1, 0.5, 400) |
| elif st == "ALL": |
| d["blast_pct_bone_marrow"] = clamp(rng.normal(75, 18), 25, 99) |
| d["blast_pct_peripheral_blood"] = clamp(d["blast_pct_bone_marrow"] * rng.uniform(0.4, 0.95), 0, 99) |
| d["wbc_k_ul"] = lognorm(rng, 3.5, 1.3, 0.5, 500) |
| elif st == "CML": |
| d["blast_pct_bone_marrow"] = clamp(rng.normal(8, 5), 0, 19) |
| d["blast_pct_peripheral_blood"] = clamp(rng.normal(4, 3), 0, 15) |
| d["wbc_k_ul"] = lognorm(rng, 5.0, 0.9, 20, 500) |
| else: |
| d["blast_pct_bone_marrow"] = "N/A" |
| d["blast_pct_peripheral_blood"] = "N/A" |
| d["wbc_k_ul"] = lognorm(rng, 4.2, 0.8, 5, 300) |
|
|
| d["hemoglobin_g_dl"] = round(clamp(rng.normal(8.5, 2.1), 4.0, 16.0), 1) |
| d["platelets_k_ul"] = round(lognorm(rng, 4.8, 0.9, 5, 800), 0) |
| d["ldh_u_l"] = round(lognorm(rng, 6.2, 0.7, 100, 8000), 0) |
| d["bone_marrow_cellularity_pct"] = int(clamp(rng.normal(80, 12), 40, 100)) |
| d["auer_rods_flag"] = int(st == "AML" and rng.random() < 0.30) |
| d["splenomegaly_flag"] = int( |
| (st == "CML" and rng.random() < 0.55) or |
| (st == "CLL" and rng.random() < 0.35) or |
| (st in ["AML","ALL"] and rng.random() < 0.15) |
| ) |
| d["lymphadenopathy_flag"] = int( |
| (st == "CLL" and rng.random() < 0.65) or |
| (st == "ALL" and rng.random() < 0.40) or |
| (st in ["AML","APL"] and rng.random() < 0.05) |
| ) |
| d["cns_involvement_flag"] = int( |
| (st == "ALL" and rng.random() < 0.15) or |
| (st == "AML" and rng.random() < 0.03) |
| ) |
|
|
| rows.append(d) |
|
|
| return pd.DataFrame(rows, index=demo.index) |
|
|
| |
| |
| |
|
|
| def generate_molecular(rng, demo): |
| n = len(demo) |
| rows = [] |
|
|
| for i, row in demo.iterrows(): |
| st = row["leukemia_type"] |
| age = row["age_at_diagnosis"] |
| d = {} |
|
|
| |
| d["t_8_21_flag"] = int(st == "AML" and rng.random() < 0.07) |
| d["inv_16_flag"] = int(st == "AML" and rng.random() < 0.05) |
| d["t_15_17_pml_rara_flag"] = int(st == "APL") |
| d["t_9_22_bcr_abl1_flag"] = int( |
| st == "CML" or |
| (st == "ALL" and rng.random() < (0.05 if age < 18 else 0.25)) |
| ) |
| d["bcr_abl1_transcript"] = ( |
| "p210" if st == "CML" else |
| "p190" if (st == "ALL" and d["t_9_22_bcr_abl1_flag"]) else "None" |
| ) |
| d["inv_3_flag"] = int(st == "AML" and rng.random() < 0.03) |
| d["del_7_monosomy7_flag"] = int(st == "AML" and rng.random() < 0.08) |
| d["del_5q_monosomy5_flag"] = int(st == "AML" and rng.random() < 0.07) |
| d["complex_karyotype_flag"] = int(st == "AML" and rng.random() < 0.12) |
| d["monosomal_karyotype_flag"] = int(st == "AML" and rng.random() < 0.08) |
|
|
| d["t_4_11_kmt2a_flag"] = int(st == "ALL" and age < 2 and rng.random() < 0.70) |
| d["t_12_21_etv6_runx1_flag"] = int(st == "ALL" and age < 18 and rng.random() < 0.25) |
| d["t_1_19_tcf3_pbx1_flag"] = int(st == "ALL" and rng.random() < 0.05) |
| d["hypodiploidy_flag"] = int(st == "ALL" and rng.random() < 0.05) |
| d["hyperdiploidy_flag"] = int(st == "ALL" and age < 18 and rng.random() < 0.25) |
| d["ikzf1_deletion_flag"] = int(st == "ALL" and rng.random() < 0.15) |
|
|
| |
| cbf_aml = d["t_8_21_flag"] or d["inv_16_flag"] |
| d["npm1_mutation"] = choice(rng, ["WT", "Mutant"], |
| [0.65, 0.35] if st == "AML" else [1.0, 0.0]) |
| d["flt3_itd_status"] = choice(rng, |
| ["Negative", "Low_Allelic_Ratio", "High_Allelic_Ratio"], |
| [0.72, 0.14, 0.14] if st == "AML" else [1.0, 0.0, 0.0]) |
| d["flt3_itd_allelic_ratio"] = ( |
| round(lognorm(rng, -0.5, 0.7, 0.01, 10.0), 3) |
| if d["flt3_itd_status"] != "Negative" else 0.0 |
| ) |
| d["flt3_itd_high_ratio"] = int(d["flt3_itd_allelic_ratio"] >= 0.5) |
| d["flt3_tkd_d835_status"] = choice(rng, ["Negative","Positive"], |
| [0.93, 0.07] if st == "AML" else [1.0, 0.0]) |
|
|
| d["idh1_mutation"] = choice(rng, ["WT","R132H","R132C","R132S"], |
| [0.92, 0.05, 0.02, 0.01] if st == "AML" else [1.0, 0.0, 0.0, 0.0]) |
| d["idh2_mutation"] = choice(rng, ["WT","R140Q","R172K","R172M"], |
| [0.87, 0.07, 0.04, 0.02] if st == "AML" else [1.0, 0.0, 0.0, 0.0]) |
| d["dnmt3a_mutation"] = choice(rng, ["WT","R882H","R882C","Other"], |
| [0.75, 0.12, 0.08, 0.05] if st == "AML" else [1.0, 0.0, 0.0, 0.0]) |
| d["tet2_mutation"] = int(st == "AML" and rng.random() < 0.20) |
| d["asxl1_mutation"] = int(st == "AML" and rng.random() < 0.15) |
| d["runx1_mutation"] = int(st == "AML" and rng.random() < 0.12) |
| d["tp53_mutation"] = int( |
| (st == "AML" and rng.random() < 0.10) or |
| (st == "CLL" and rng.random() < 0.10) |
| ) |
| d["cebpa_mutation"] = choice(rng, ["WT","Biallelic","Monoallelic"], |
| [0.90, 0.05, 0.05] if st == "AML" else [1.0, 0.0, 0.0]) |
| d["sf3b1_mutation"] = int(st == "AML" and rng.random() < 0.08) |
| d["kit_mutation"] = int(cbf_aml and rng.random() < 0.25) |
|
|
| |
| d["ighv_mutated_flag"] = int(st == "CLL" and rng.random() < 0.50) |
| d["del_17p_flag"] = int(st == "CLL" and rng.random() < 0.08) |
| d["del_11q_flag"] = int(st == "CLL" and rng.random() < 0.12) |
| d["del_13q_flag"] = int(st == "CLL" and rng.random() < 0.50) |
| d["trisomy_12_flag"] = int(st == "CLL" and rng.random() < 0.15) |
| d["notch1_mutation_cll"] = int(st == "CLL" and rng.random() < 0.12) |
| d["btk_c481s_flag"] = int(st == "CLL" and rng.random() < 0.05) |
|
|
| |
| d["jak_stat_mutation"] = int(st == "ALL" and rng.random() < 0.12) |
| d["notch1_fbxw7_mutation"] = int(st == "ALL" and rng.random() < 0.55) |
|
|
| |
| d["mrd_method"] = choice(rng, |
| ["Flow_Cytometry", "PCR", "NGS_MRD", "Not_Assessed"], |
| [0.45, 0.30, 0.15, 0.10]) |
|
|
| |
| if st in ["AML", "APL"]: |
| if st == "APL" or d["t_8_21_flag"] or d["inv_16_flag"] or d["cebpa_mutation"] == "Biallelic": |
| d["eln_2022_risk_category"] = "Favorable" |
| elif d["tp53_mutation"] or d["complex_karyotype_flag"] or d["monosomal_karyotype_flag"] or d["asxl1_mutation"] or d["runx1_mutation"]: |
| d["eln_2022_risk_category"] = "Adverse" |
| elif d["flt3_itd_high_ratio"] and d["npm1_mutation"] == "WT": |
| d["eln_2022_risk_category"] = "Adverse" |
| elif d["npm1_mutation"] == "Mutant": |
| d["eln_2022_risk_category"] = "Favorable" if d["flt3_itd_status"] == "Negative" else "Intermediate" |
| else: |
| d["eln_2022_risk_category"] = "Intermediate" |
| elif st == "CML": |
| d["eln_2022_risk_category"] = "N/A" |
| d["sokal_score"] = choice(rng, ["Low","Intermediate","High"], [0.40, 0.40, 0.20]) |
| d["eutos_score"] = choice(rng, ["Low","High"], [0.70, 0.30]) |
| elif st == "CLL": |
| d["eln_2022_risk_category"] = "N/A" |
| d["cll_rai_stage"] = choice(rng, [0,1,2,3,4], [0.30,0.25,0.20,0.15,0.10]) |
| d["cll_binet_stage"] = choice(rng, ["A","B","C"], [0.55,0.30,0.15]) |
| else: |
| d["eln_2022_risk_category"] = "N/A" |
| d["nccn_all_risk"] = choice(rng, ["Standard","High","Very_High"], |
| [0.30, 0.45, 0.25] if age >= 18 else [0.55, 0.35, 0.10]) |
|
|
| |
| for f in ["sokal_score","eutos_score","cll_rai_stage","cll_binet_stage","nccn_all_risk"]: |
| if f not in d: |
| d[f] = "N/A" |
|
|
| rows.append(d) |
|
|
| return pd.DataFrame(rows, index=demo.index) |
|
|
| |
| |
| |
|
|
| def generate_treatment(rng, demo, mol): |
| rows = [] |
|
|
| for i, row in demo.iterrows(): |
| st = row["leukemia_type"] |
| age = row["age_at_diagnosis"] |
| ecog = row["ecog_performance_status"] |
| m = mol.loc[i] |
| d = {} |
|
|
| d["treatment_intent"] = ( |
| "Palliative" if ecog >= 3 else |
| "Watch_and_Wait" if (st == "CLL" and m.get("cll_rai_stage", 0) in [0, 1]) else |
| "Curative" |
| ) |
|
|
| if st == "AML": |
| flt3_pos = m["flt3_itd_status"] != "Negative" or m["flt3_tkd_d835_status"] == "Positive" |
| idh1_pos = m["idh1_mutation"] != "WT" |
| idh2_pos = m["idh2_mutation"] != "WT" |
| fit = ecog <= 2 and age <= 75 |
|
|
| if not fit: |
| d["induction_regimen"] = choice(rng, ["Venetoclax+AZA","Venetoclax+LDAC","BSC"], [0.55,0.30,0.15]) |
| elif flt3_pos: |
| d["induction_regimen"] = "Midostaurin+7+3" |
| elif idh1_pos: |
| d["induction_regimen"] = choice(rng, ["Ivosidenib+AZA","7+3_Idarubicin"], [0.40, 0.60]) |
| elif idh2_pos: |
| d["induction_regimen"] = choice(rng, ["Enasidenib+AZA","7+3_Idarubicin"], [0.35, 0.65]) |
| elif m.get("eln_2022_risk_category") == "Adverse": |
| d["induction_regimen"] = choice(rng, ["CPX-351","7+3_Idarubicin"], [0.45, 0.55]) |
| else: |
| d["induction_regimen"] = choice(rng, ["7+3_Daunorubicin","7+3_Idarubicin"], [0.40, 0.60]) |
|
|
| d["consolidation_regimen"] = choice(rng, |
| ["HiDAC_3g_m2", "Intermediate_AraC", "Azacitidine_Maintenance", "None"], |
| [0.50, 0.25, 0.15, 0.10]) |
| d["maintenance_flag"] = int(d["consolidation_regimen"] == "Azacitidine_Maintenance" or |
| (flt3_pos and rng.random() < 0.50)) |
| d["maintenance_regimen"] = "Gilteritinib" if (flt3_pos and d["maintenance_flag"]) else ( |
| "Azacitidine" if d["maintenance_flag"] else "None") |
| d["induction_cycles"] = int(rng.choice([1, 2], p=[0.70, 0.30])) |
| d["consolidation_cycles"] = int(rng.choice([1, 2, 3, 4], p=[0.15, 0.35, 0.35, 0.15])) |
| d["targeted_agent"] = ( |
| "Midostaurin" if d["induction_regimen"] == "Midostaurin+7+3" else |
| "Ivosidenib" if "Ivosidenib" in d["induction_regimen"] else |
| "Enasidenib" if "Enasidenib" in d["induction_regimen"] else |
| "Venetoclax" if "Venetoclax" in d["induction_regimen"] else "None" |
| ) |
| d["apl_atra_flag"] = 0 |
|
|
| elif st == "APL": |
| d["induction_regimen"] = choice(rng, ["ATRA_ATO_APL0406","ATRA_ATO_APML4","ATRA_Daunorubicin_Sanz"], [0.50, 0.25, 0.25]) |
| d["apl_atra_flag"] = 1 |
| d["consolidation_regimen"] = choice(rng, ["ATRA_ATO_Consolidation","ATRA_Chemo"], [0.60, 0.40]) |
| d["maintenance_flag"] = int(rng.random() < 0.70) |
| d["maintenance_regimen"] = "ATRA_6MP_MTX" if d["maintenance_flag"] else "None" |
| d["induction_cycles"] = 1 |
| d["consolidation_cycles"] = int(rng.choice([2, 3], p=[0.30, 0.70])) |
| d["targeted_agent"] = "ATO+ATRA" |
|
|
| elif st == "ALL": |
| ph_pos = m.get("t_9_22_bcr_abl1_flag", 0) == 1 |
| if ph_pos: |
| d["induction_regimen"] = choice(rng, ["Dasatinib+Steroids","Ponatinib+Chemo","HyperCVAD_Dasatinib"], [0.40, 0.30, 0.30]) |
| d["targeted_agent"] = "Ponatinib" if "Ponatinib" in d["induction_regimen"] else "Dasatinib" |
| elif age < 18: |
| d["induction_regimen"] = choice(rng, ["CALGB_10403","BFM_Protocol","COG_AALL0434"], [0.40, 0.35, 0.25]) |
| d["targeted_agent"] = "None" |
| elif m.get("cns_involvement_flag", 0): |
| d["induction_regimen"] = "HyperCVAD_A_B" |
| d["targeted_agent"] = "None" |
| else: |
| d["induction_regimen"] = choice(rng, |
| ["CALGB_10403","HyperCVAD_A_B","GRAALL_2005","UKALL14", |
| "Blinatumomab+Chemo","Inotuzumab_Ozogamicin+Chemo"], |
| [0.25, 0.25, 0.15, 0.15, 0.10, 0.10]) |
| d["targeted_agent"] = ( |
| "Blinatumomab" if "Blinatumomab" in d["induction_regimen"] else |
| "Inotuzumab" if "Inotuzumab" in d["induction_regimen"] else "None" |
| ) |
| d["consolidation_regimen"] = choice(rng, ["Consolidation_A_B_C","Maintenance_Only","None"], [0.60, 0.30, 0.10]) |
| d["maintenance_flag"] = 1 |
| d["maintenance_regimen"] = "6MP_MTX_Vincristine_Prednisone" |
| d["induction_cycles"] = 1 |
| d["consolidation_cycles"] = int(rng.choice([2, 3, 4], p=[0.25, 0.45, 0.30])) |
| d["apl_atra_flag"] = 0 |
|
|
| elif st == "CML": |
| phase = m.get("who_2022_classification","CML_CP") |
| if "BP" in str(phase): |
| d["induction_regimen"] = choice(rng, ["Ponatinib_45mg","Dasatinib_Chemo","Asciminib"], [0.40, 0.35, 0.25]) |
| else: |
| d["induction_regimen"] = choice(rng, CML_REGIMENS, [0.25, 0.25, 0.20, 0.10, 0.10, 0.10]) |
| d["targeted_agent"] = d["induction_regimen"].split("_")[0] |
| d["tki_generation"] = ( |
| "First" if "Imatinib" in d["induction_regimen"] else |
| "Second" if any(x in d["induction_regimen"] for x in ["Dasatinib","Nilotinib","Bosutinib"]) else |
| "Third" if "Ponatinib" in d["induction_regimen"] else "STAMP" |
| ) |
| d["consolidation_regimen"] = "N/A" |
| d["maintenance_flag"] = 0 |
| d["maintenance_regimen"] = "Ongoing_TKI" |
| d["induction_cycles"] = "Continuous" |
| d["consolidation_cycles"] = 0 |
| d["apl_atra_flag"] = 0 |
|
|
| else: |
| rai = m.get("cll_rai_stage", 0) |
| unfit = ecog >= 2 or age >= 75 |
| if rai in [0, 1]: |
| d["induction_regimen"] = "Watch_and_Wait" |
| d["targeted_agent"] = "None" |
| elif unfit: |
| d["induction_regimen"] = choice(rng, ["Venetoclax_Obinutuzumab","Acalabrutinib","Ibrutinib"], [0.45, 0.30, 0.25]) |
| d["targeted_agent"] = d["induction_regimen"].split("_")[0] |
| elif m.get("del_17p_flag", 0) or m.get("tp53_mutation", 0): |
| d["induction_regimen"] = choice(rng, ["Ibrutinib","Acalabrutinib","Venetoclax_Rituximab"], [0.40, 0.35, 0.25]) |
| d["targeted_agent"] = d["induction_regimen"].split("_")[0] |
| elif m.get("ighv_mutated_flag", 0): |
| d["induction_regimen"] = choice(rng, ["FCR","Venetoclax_Obinutuzumab","Venetoclax_Rituximab"], [0.35, 0.35, 0.30]) |
| d["targeted_agent"] = d["induction_regimen"].split("_")[0] |
| else: |
| d["induction_regimen"] = choice(rng, ["Ibrutinib","Acalabrutinib","Venetoclax_Obinutuzumab"], [0.35, 0.35, 0.30]) |
| d["targeted_agent"] = d["induction_regimen"].split("_")[0] |
|
|
| d["consolidation_regimen"] = "N/A" |
| d["maintenance_flag"] = 0 |
| d["maintenance_regimen"] = "Ongoing_BTKi" if "rutinib" in d["induction_regimen"] else "None" |
| d["induction_cycles"] = int(rng.choice([6, 12, 24], p=[0.30, 0.40, 0.30])) |
| d["consolidation_cycles"] = 0 |
| d["apl_atra_flag"] = 0 |
|
|
| d["total_treatment_months"] = int(clamp(rng.normal(24, 8), 6, 48)) |
| d["clinical_trial_enrollment_flag"] = int(rng.random() < 0.25) |
| d["trial_phase"] = choice(rng, ["Phase_I","Phase_II","Phase_III"], [0.10, 0.40, 0.50]) if d["clinical_trial_enrollment_flag"] else "None" |
|
|
| rows.append(d) |
|
|
| return pd.DataFrame(rows, index=demo.index) |
|
|
| |
| |
| |
|
|
| def generate_remission(rng, demo, mol, treat): |
| rows = [] |
|
|
| for i, row in demo.iterrows(): |
| st = row["leukemia_type"] |
| age = row["age_at_diagnosis"] |
| m = mol.loc[i] |
| t = treat.loc[i] |
| d = {} |
|
|
| if st in ["AML","APL"]: |
| eln = m.get("eln_2022_risk_category","Intermediate") |
| key = f"AML_{eln}" if st == "AML" else "APL" |
| base_cr = CR_RATE.get(key, 0.60) |
| d["cr_achieved_flag"] = int(rng.random() < base_cr) |
| d["cri_achieved_flag"] = int(not d["cr_achieved_flag"] and rng.random() < 0.15) |
| d["pr_achieved_flag"] = int(not d["cr_achieved_flag"] and not d["cri_achieved_flag"] and rng.random() < 0.10) |
| d["time_to_cr_days"] = int(clamp(rng.normal(28, 7), 14, 60)) if d["cr_achieved_flag"] else None |
| d["cytogenetic_remission_flag"] = int(d["cr_achieved_flag"] and rng.random() < 0.85) |
| d["molecular_remission_flag"] = int(d["cytogenetic_remission_flag"] and rng.random() < 0.75) |
| d["mrd_negativity_flag"] = int(d["molecular_remission_flag"] and rng.random() < 0.80) |
| d["mrd_conversion_cycle"] = int(rng.choice([1,2,3,4], p=[0.40,0.35,0.15,0.10])) if d["mrd_negativity_flag"] else None |
| d["hematologic_relapse_flag"] = int(d["cr_achieved_flag"] and rng.random() < (0.20 if eln == "Favorable" else 0.45 if eln == "Intermediate" else 0.65)) |
| d["molecular_relapse_flag"] = int(d["mrd_negativity_flag"] and rng.random() < 0.25) |
| d["time_to_relapse_months"] = round(weibull_sample(rng, 1.5, 18 if eln=="Favorable" else 12)[0], 1) if d["hematologic_relapse_flag"] else None |
| d["relapse_site"] = choice(rng, ["Bone_Marrow","Extramedullary","CNS","Peripheral_Blood"], [0.70,0.15,0.08,0.07]) if d["hematologic_relapse_flag"] else "None" |
| d["salvage_cr_rate"] = round(rng.uniform(0.30, 0.45) if st=="AML" else rng.uniform(0.40, 0.65), 2) if d["hematologic_relapse_flag"] else None |
| |
| for f in ["bcr_abl1_is_pct","ccyr_achieved_flag","mmr_achieved_flag","dmr_achieved_flag","sokal_score_val","tki_resistance_mechanism"]: |
| d[f] = "N/A" |
| |
| for f in ["cll_iwcll_response","bruton_resistance_flag"]: |
| d[f] = "N/A" |
|
|
| elif st == "ALL": |
| is_ped = age < 18 |
| key = "ALL_Pediatric" if is_ped else "ALL_Adult" |
| base_cr = CR_RATE[key] |
| d["cr_achieved_flag"] = int(rng.random() < base_cr) |
| d["cri_achieved_flag"] = int(not d["cr_achieved_flag"] and rng.random() < 0.08) |
| d["pr_achieved_flag"] = int(not d["cr_achieved_flag"] and not d["cri_achieved_flag"] and rng.random() < 0.05) |
| d["time_to_cr_days"] = int(clamp(rng.normal(28, 7), 14, 56)) if d["cr_achieved_flag"] else None |
| d["cytogenetic_remission_flag"] = int(d["cr_achieved_flag"] and rng.random() < 0.80) |
| d["molecular_remission_flag"] = int(d["cytogenetic_remission_flag"] and rng.random() < 0.70) |
| d["mrd_negativity_flag"] = int(d["molecular_remission_flag"] and rng.random() < 0.85) |
| d["mrd_conversion_cycle"] = int(rng.choice([1,2,3], p=[0.45,0.35,0.20])) if d["mrd_negativity_flag"] else None |
| d["hematologic_relapse_flag"] = int(d["cr_achieved_flag"] and rng.random() < (0.05 if is_ped else 0.40)) |
| d["molecular_relapse_flag"] = int(d["mrd_negativity_flag"] and rng.random() < 0.20) |
| d["time_to_relapse_months"] = round(weibull_sample(rng, 1.5, 24 if is_ped else 14)[0], 1) if d["hematologic_relapse_flag"] else None |
| d["relapse_site"] = choice(rng, ["Bone_Marrow","CNS","Testicular","Peripheral_Blood"], [0.65,0.20,0.05,0.10]) if d["hematologic_relapse_flag"] else "None" |
| d["salvage_cr_rate"] = round(rng.uniform(0.25, 0.50), 2) if d["hematologic_relapse_flag"] else None |
| for f in ["bcr_abl1_is_pct","ccyr_achieved_flag","mmr_achieved_flag","dmr_achieved_flag","sokal_score_val","tki_resistance_mechanism"]: |
| d[f] = "N/A" |
| for f in ["cll_iwcll_response","bruton_resistance_flag"]: |
| d[f] = "N/A" |
|
|
| elif st == "CML": |
| d["cr_achieved_flag"] = "N/A" |
| d["cri_achieved_flag"] = "N/A" |
| d["pr_achieved_flag"] = "N/A" |
| d["time_to_cr_days"] = "N/A" |
| d["cytogenetic_remission_flag"] = "N/A" |
| d["molecular_remission_flag"] = "N/A" |
| d["mrd_negativity_flag"] = "N/A" |
| d["mrd_conversion_cycle"] = "N/A" |
| d["hematologic_relapse_flag"] = int(rng.random() < 0.10) |
| d["molecular_relapse_flag"] = int(rng.random() < 0.15) |
| d["time_to_relapse_months"] = round(weibull_sample(rng, 1.8, 36)[0], 1) if d["hematologic_relapse_flag"] else None |
| d["relapse_site"] = "Peripheral_Blood" if d["hematologic_relapse_flag"] else "None" |
| d["salvage_cr_rate"] = "N/A" |
| |
| tki_gen = t.get("tki_generation", "Second") |
| mmr_base = 0.70 if tki_gen == "First" else 0.80 if tki_gen == "Second" else 0.75 |
| d["ccyr_achieved_flag"] = int(rng.random() < (mmr_base + 0.10)) |
| d["mmr_achieved_flag"] = int(rng.random() < mmr_base) |
| d["dmr_achieved_flag"] = int(d["mmr_achieved_flag"] and rng.random() < 0.40) |
| d["bcr_abl1_is_pct"] = round(lognorm(rng, -2.0, 1.5, 0.001, 100) if d["mmr_achieved_flag"] else lognorm(rng, 0.5, 1.0, 0.1, 100), 4) |
| d["tki_resistance_mechanism"] = choice(rng, ["T315I","E255K","F317L","Compound_Mutation","None"], [0.30,0.20,0.15,0.10,0.25]) if d["molecular_relapse_flag"] else "None" |
| for f in ["cll_iwcll_response","bruton_resistance_flag"]: |
| d[f] = "N/A" |
|
|
| else: |
| d["cr_achieved_flag"] = int(rng.random() < 0.25) |
| d["cri_achieved_flag"] = "N/A" |
| d["pr_achieved_flag"] = int(not d["cr_achieved_flag"] and rng.random() < 0.55) |
| d["time_to_cr_days"] = "N/A" |
| d["cytogenetic_remission_flag"] = "N/A" |
| d["molecular_remission_flag"] = "N/A" |
| d["mrd_negativity_flag"] = int(d["cr_achieved_flag"] and rng.random() < 0.60) |
| d["mrd_conversion_cycle"] = "N/A" |
| d["hematologic_relapse_flag"] = int(rng.random() < 0.25) |
| d["molecular_relapse_flag"] = "N/A" |
| d["time_to_relapse_months"] = round(weibull_sample(rng, 1.4, 36)[0], 1) if d["hematologic_relapse_flag"] else None |
| d["relapse_site"] = "Peripheral_Blood" if d["hematologic_relapse_flag"] else "None" |
| d["salvage_cr_rate"] = "N/A" |
| d["cll_iwcll_response"] = choice(rng, ["CR","MRD_Negative_CR","PR","SD","PD"], [0.20,0.15,0.45,0.10,0.10]) |
| d["bruton_resistance_flag"] = int(m.get("btk_c481s_flag", 0)) |
| for f in ["bcr_abl1_is_pct","ccyr_achieved_flag","mmr_achieved_flag","dmr_achieved_flag","sokal_score_val","tki_resistance_mechanism"]: |
| d[f] = "N/A" |
|
|
| rows.append(d) |
|
|
| return pd.DataFrame(rows, index=demo.index) |
|
|
| |
| |
| |
|
|
| def generate_bmt(rng, demo, mol, treat, remission): |
| rows = [] |
|
|
| for i, row in demo.iterrows(): |
| st = row["leukemia_type"] |
| age = row["age_at_diagnosis"] |
| m = mol.loc[i] |
| t = treat.loc[i] |
| r = remission.loc[i] |
| d = {} |
|
|
| |
| eln = m.get("eln_2022_risk_category", "Intermediate") |
| age_eligible = age <= 75 |
| if st == "AML": |
| bmt_indicated = (eln in ["Intermediate","Adverse"]) and age_eligible and str(r.get("cr_achieved_flag","0")) == "1" |
| elif st == "ALL": |
| bmt_indicated = age_eligible and str(r.get("cr_achieved_flag","0")) == "1" |
| elif st == "APL": |
| bmt_indicated = str(r.get("hematologic_relapse_flag","0")) == "1" |
| elif st == "CML": |
| bmt_indicated = str(r.get("tki_resistance_mechanism","None")) not in ["None","N/A"] |
| else: |
| bmt_indicated = False |
|
|
| d["bmt_indicated_flag"] = int(bmt_indicated) |
| d["bmt_performed_flag"] = int(bmt_indicated and rng.random() < 0.70) |
|
|
| if d["bmt_performed_flag"]: |
| d["bmt_timing"] = choice(rng, |
| ["CR1","CR2","Upfront_High_Risk","Relapsed_Refractory"], |
| [0.55, 0.25, 0.12, 0.08]) |
| d["bmt_type"] = choice(rng, ["Allogeneic","Autologous","Haploidentical"], [0.75, 0.15, 0.10]) |
| if d["bmt_type"] in ["Allogeneic","Haploidentical"]: |
| d["donor_type"] = "Haploidentical" if d["bmt_type"] == "Haploidentical" else choice(rng, ["MRD","MUD","MMUD"], [0.45, 0.42, 0.13]) |
| d["hla_match_grade"] = "5-6/10" if d["donor_type"] == "Haploidentical" else choice(rng, ["10/10","9/10","8/10"], [0.60, 0.30, 0.10]) |
| else: |
| d["donor_type"] = "Autologous" |
| d["hla_match_grade"] = "N/A" |
|
|
| d["donor_recipient_sex_mismatch_flag"] = int(rng.random() < 0.25) |
| d["stem_cell_source"] = choice(rng, ["Peripheral_Blood","Bone_Marrow","Cord_Blood"], [0.75, 0.20, 0.05]) |
| age_fit = age >= 60 or m.get("eln_2022_risk_category","") == "Adverse" |
| d["conditioning_intensity"] = choice(rng, ["Myeloablative_MAC","Reduced_Intensity_RIC","Non_Myeloablative_NMA"], |
| [0.35, 0.50, 0.15] if age >= 55 else [0.70, 0.25, 0.05]) |
| cond_type = d["conditioning_intensity"].split("_")[0].split("_")[0] |
| if "Mac" in d["conditioning_intensity"] or "Myeloab" in d["conditioning_intensity"]: |
| d["conditioning_regimen"] = choice(rng, CONDITIONING["MAC"]) |
| elif "RIC" in d["conditioning_intensity"]: |
| d["conditioning_regimen"] = choice(rng, CONDITIONING["RIC"]) |
| else: |
| d["conditioning_regimen"] = choice(rng, CONDITIONING["NMA"]) |
|
|
| d["gvhd_prophylaxis"] = choice(rng, GVHD_PROPHYLAXIS) |
| d["graft_cd34_x10e6_kg"] = round(lognorm(rng, 1.8, 0.4, 1.0, 15.0), 2) |
| d["engraftment_day_neutrophil"] = int(clamp(rng.normal(14, 4), 8, 35)) |
| d["platelet_engraftment_day"] = int(clamp(rng.normal(18, 5), 10, 45)) |
| d["primary_graft_failure_flag"] = int(rng.random() < 0.05) |
| d["secondary_graft_failure_flag"] = int(not d["primary_graft_failure_flag"] and rng.random() < 0.03) |
|
|
| if d["bmt_type"] in ["Allogeneic","Haploidentical"] and not d["primary_graft_failure_flag"]: |
| donor = d["donor_type"] |
| agvhd_p = AGVHD_RATE.get(donor, 0.40) |
| |
| if "PT_Cy" in d["gvhd_prophylaxis"]: |
| agvhd_p *= 0.65 |
|
|
| d["acute_gvhd_flag"] = int(rng.random() < agvhd_p) |
| d["acute_gvhd_grade"] = choice(rng, ["None","I","II","III","IV"], |
| [1-agvhd_p, agvhd_p*0.30, agvhd_p*0.40, agvhd_p*0.20, agvhd_p*0.10]) if d["acute_gvhd_flag"] else "None" |
| d["acute_gvhd_organs"] = choice(rng, ["Skin","Gut","Liver","Multi_organ"], [0.45, 0.30, 0.10, 0.15]) if d["acute_gvhd_flag"] else "None" |
| d["chronic_gvhd_flag"] = int(rng.random() < 0.40) |
| d["chronic_gvhd_severity"] = choice(rng, ["Mild","Moderate","Severe"], [0.45, 0.35, 0.20]) if d["chronic_gvhd_flag"] else "None" |
| d["chronic_gvhd_organs"] = choice(rng, ["Skin","Mouth","Eye","Lung","Liver","Multi"], [0.30,0.20,0.15,0.10,0.10,0.15]) if d["chronic_gvhd_flag"] else "None" |
| nrm_p = NRM_RATE.get(donor, 0.15) |
| d["non_relapse_mortality_flag"] = int(rng.random() < nrm_p) |
| else: |
| for f in ["acute_gvhd_flag","acute_gvhd_grade","acute_gvhd_organs", |
| "chronic_gvhd_flag","chronic_gvhd_severity","chronic_gvhd_organs", |
| "non_relapse_mortality_flag"]: |
| d[f] = 0 if "flag" in f else "None" if f.endswith("_flag")==False else 0 |
|
|
| |
| is_mac = "Mac" in d["conditioning_intensity"] or "Myeloab" in d["conditioning_intensity"] |
| d["vod_flag"] = int(rng.random() < (0.12 if is_mac else 0.04)) |
| d["vod_severity"] = choice(rng, ["Mild","Moderate","Severe","Very_Severe"], [0.40,0.30,0.20,0.10]) if d["vod_flag"] else "None" |
| d["cmv_reactivation_flag"] = int(rng.random() < 0.50) |
| d["ebv_reactivation_flag"] = int(rng.random() < 0.18) |
| d["invasive_fungal_infection_flag"] = int(rng.random() < (0.10 if is_mac else 0.04)) |
| d["bacterial_infection_flag"] = int(rng.random() < 0.40) |
| d["idiopathic_pneumonia_flag"] = int(rng.random() < 0.05) |
| d["thrombotic_microangiopathy_flag"] = int(rng.random() < 0.07) |
| d["day_100_mortality_flag"] = int( |
| d["primary_graft_failure_flag"] or |
| (d["vod_flag"] and d["vod_severity"] in ["Severe","Very_Severe"] and rng.random() < 0.50) or |
| (d.get("acute_gvhd_grade","None") in ["III","IV"] and rng.random() < 0.35) or |
| rng.random() < 0.05 |
| ) |
| d["day_100_chimerism_pct"] = round(clamp(rng.normal(97, 4), 80, 100), 1) if not d["primary_graft_failure_flag"] else round(clamp(rng.normal(60, 20), 10, 80), 1) |
| d["relapse_post_bmt_flag"] = int(not d["day_100_mortality_flag"] and rng.random() < 0.28) |
| d["time_to_relapse_post_bmt_months"] = round(weibull_sample(rng, 1.3, 14)[0], 1) if d["relapse_post_bmt_flag"] else None |
| d["donor_lymphocyte_infusion_flag"] = int(d["relapse_post_bmt_flag"] and rng.random() < 0.45) |
| d["second_bmt_flag"] = int(d["relapse_post_bmt_flag"] and rng.random() < 0.07) |
| d["one_yr_os_post_bmt"] = int(not d["day_100_mortality_flag"] and not d["non_relapse_mortality_flag"]) |
| d["two_yr_rfs_post_bmt"] = int(d["one_yr_os_post_bmt"] and not d["relapse_post_bmt_flag"] and rng.random() < 0.70) |
| else: |
| for f in ["bmt_timing","bmt_type","donor_type","hla_match_grade","donor_recipient_sex_mismatch_flag", |
| "stem_cell_source","conditioning_intensity","conditioning_regimen","gvhd_prophylaxis", |
| "graft_cd34_x10e6_kg","engraftment_day_neutrophil","platelet_engraftment_day", |
| "primary_graft_failure_flag","secondary_graft_failure_flag", |
| "acute_gvhd_flag","acute_gvhd_grade","acute_gvhd_organs", |
| "chronic_gvhd_flag","chronic_gvhd_severity","chronic_gvhd_organs", |
| "vod_flag","vod_severity","cmv_reactivation_flag","ebv_reactivation_flag", |
| "invasive_fungal_infection_flag","bacterial_infection_flag", |
| "idiopathic_pneumonia_flag","thrombotic_microangiopathy_flag", |
| "day_100_mortality_flag","non_relapse_mortality_flag","day_100_chimerism_pct", |
| "relapse_post_bmt_flag","time_to_relapse_post_bmt_months", |
| "donor_lymphocyte_infusion_flag","second_bmt_flag", |
| "one_yr_os_post_bmt","two_yr_rfs_post_bmt"]: |
| d[f] = "N/A" |
|
|
| rows.append(d) |
|
|
| return pd.DataFrame(rows, index=demo.index) |
|
|
| |
| |
| |
|
|
| def generate_toxicity(rng, demo, treat): |
| rows = [] |
| for i, row in demo.iterrows(): |
| st = row["leukemia_type"] |
| t = treat.loc[i] |
| reg = str(t.get("induction_regimen","")) |
| d = {} |
| d["febrile_neutropenia_flag"] = int(st in ["AML","APL","ALL"] and rng.random() < 0.85) |
| d["induction_mortality_flag"] = int( |
| (st == "AML" and rng.random() < 0.04) or |
| (st == "APL" and rng.random() < 0.02) or |
| (st == "ALL" and rng.random() < 0.02) |
| ) |
| d["fungal_infection_flag"] = int(rng.random() < 0.15 if st in ["AML","APL","ALL"] else rng.random() < 0.03) |
| d["tumor_lysis_syndrome_flag"] = int( |
| (st == "ALL" and rng.random() < 0.18) or |
| (st == "AML" and rng.random() < 0.08) |
| ) |
| d["tls_cairo_bishop_grade"] = int(rng.choice([1,2,3], p=[0.50,0.35,0.15])) if d["tumor_lysis_syndrome_flag"] else 0 |
| d["differentiation_syndrome_flag"] = int( |
| (st == "APL" and rng.random() < 0.22) or |
| ("IDH" in reg and rng.random() < 0.15) |
| ) |
| d["differentiation_syndrome_severity"] = choice(rng, ["Mild","Moderate","Severe"], [0.50,0.35,0.15]) if d["differentiation_syndrome_flag"] else "None" |
| d["anthracycline_cardiotoxicity_flag"] = int("7+3" in reg and rng.random() < 0.08) |
| d["lvef_post_treatment_pct"] = int(clamp(rng.normal(58, 8), 30, 75)) |
| d["peripheral_neuropathy_flag"] = int(st == "ALL" and rng.random() < 0.30) |
| d["hepatotoxicity_flag"] = int(rng.random() < 0.15) |
| d["mucositis_grade"] = int(rng.choice([0,1,2,3,4], p=[0.40,0.25,0.20,0.10,0.05])) |
| d["hemorrhagic_cystitis_flag"] = int(rng.random() < 0.08) |
| d["qol_score_baseline_fact_leu"] = int(clamp(rng.normal(130, 22), 60, 176)) |
| d["qol_score_end_of_induction"] = int(clamp(d["qol_score_baseline_fact_leu"] - rng.normal(20, 12), 40, 176)) |
| rows.append(d) |
| return pd.DataFrame(rows, index=demo.index) |
|
|
| |
| |
| |
|
|
| def generate_survival(rng, demo, mol, remission, bmt, tox): |
| rows = [] |
| for i, row in demo.iterrows(): |
| st = row["leukemia_type"] |
| age = row["age_at_diagnosis"] |
| m = mol.loc[i] |
| r = remission.loc[i] |
| b = bmt.loc[i] |
| tx = tox.loc[i] |
| d = {} |
|
|
| eln = str(m.get("eln_2022_risk_category","Intermediate")) |
| is_ped = age < 18 |
| induction_dead = tx.get("induction_mortality_flag", 0) |
| bmt_dead = str(b.get("day_100_mortality_flag","0")) in ["1","True"] |
| nrm_dead = str(b.get("non_relapse_mortality_flag","0")) in ["1","True"] |
|
|
| if st == "AML": |
| key = f"AML_{eln}" if eln in ["Favorable","Intermediate","Adverse"] else "AML_Intermediate" |
| elif st == "APL": |
| key = "APL" |
| elif st == "ALL": |
| key = "ALL_Pediatric" if is_ped else "ALL_Adult" |
| elif st == "CML": |
| key = "CML" |
| else: |
| key = "CLL" |
|
|
| lam = OS_LAMBDA.get(key, 36.0) |
| os_months = round(weibull_sample(rng, 1.4, lam)[0], 1) |
|
|
| |
| if induction_dead: |
| os_months = round(rng.uniform(0.5, 3.0), 1) |
| elif bmt_dead: |
| bmt_os = str(b.get("one_yr_os_post_bmt","N/A")) |
| if bmt_os == "0": |
| os_months = min(os_months, round(rng.uniform(2, 12), 1)) |
|
|
| d["overall_survival_months"] = os_months |
| d["event_free_survival_months"] = round(min(os_months, weibull_sample(rng, 1.5, lam * 0.7)[0]), 1) |
| d["leukemia_free_survival_months"] = round(min(d["event_free_survival_months"], weibull_sample(rng, 1.6, lam * 0.8)[0]), 1) |
| d["relapse_free_survival_months"] = round(min(d["leukemia_free_survival_months"], weibull_sample(rng, 1.7, lam * 0.9)[0]), 1) |
|
|
| if induction_dead: |
| d["vital_status"] = "Dead_Treatment" |
| d["cause_of_death"] = "Organ_Failure" |
| elif bmt_dead or nrm_dead: |
| d["vital_status"] = "Dead_Treatment" |
| d["cause_of_death"] = choice(rng, ["GvHD","Infection","Organ_Failure"], [0.35,0.45,0.20]) |
| elif str(r.get("hematologic_relapse_flag","0")) == "1" and os_months < lam * 0.5: |
| d["vital_status"] = "Dead_Disease" |
| d["cause_of_death"] = "Relapse" |
| elif rng.random() < 0.65: |
| d["vital_status"] = "Alive" |
| d["cause_of_death"] = "None" |
| else: |
| d["vital_status"] = choice(rng, ["Dead_Disease","Dead_Other"], [0.75, 0.25]) |
| d["cause_of_death"] = choice(rng, ["Relapse","Infection","Secondary_Malignancy","Other"], |
| [0.60, 0.20, 0.10, 0.10]) |
|
|
| d["second_malignancy_flag"] = int(rng.random() < 0.03) |
| rows.append(d) |
|
|
| return pd.DataFrame(rows, index=demo.index) |
|
|
| |
| |
| |
|
|
| def generate_mrd_longitudinal(rng, demo, remission, n_timepoints=12): |
| """Generate monthly MRD assessments for a subset of patients.""" |
| records = [] |
| sample_ids = demo.sample(frac=0.40, random_state=42)["patient_id"].tolist() |
|
|
| for pid in sample_ids: |
| idx = demo[demo["patient_id"] == pid].index[0] |
| st = demo.loc[idx, "leukemia_type"] |
| r = remission.loc[idx] |
| cr = str(r.get("cr_achieved_flag", "0")) == "1" |
| mrd_neg = str(r.get("mrd_negativity_flag", "0")) == "1" |
|
|
| for month in range(1, n_timepoints + 1): |
| mrd_val = None |
| if cr and month <= 3: |
| mrd_val = round(lognorm(rng, -1.0, 1.2, 0.001, 10.0), 4) |
| elif mrd_neg and month > 3: |
| mrd_val = round(lognorm(rng, -3.5, 0.8, 0.0001, 0.05), 5) |
| elif cr: |
| mrd_val = round(lognorm(rng, -2.0, 1.0, 0.001, 1.0), 4) |
| records.append({ |
| "patient_id": pid, |
| "leukemia_type": st, |
| "assessment_month": month, |
| "mrd_value_pct": mrd_val, |
| "mrd_negative_flag": int(mrd_val is not None and mrd_val < 0.01), |
| "assessment_method": rng.choice(["Flow_Cytometry","PCR","NGS_MRD"]), |
| "wbc_k_ul": round(lognorm(rng, 2.3, 0.5, 1.0, 15.0), 1), |
| "hemoglobin_g_dl": round(clamp(rng.normal(11.5, 1.8), 7.0, 16.0), 1), |
| "platelets_k_ul": round(lognorm(rng, 5.0, 0.4, 50, 400), 0), |
| "neutrophils_k_ul": round(lognorm(rng, 1.2, 0.6, 0.1, 12.0), 2), |
| }) |
| return pd.DataFrame(records) |
|
|
| |
| |
| |
|
|
| def main(): |
| parser = argparse.ArgumentParser(description="HC-ONC-007 Leukemia Simulation Engine") |
| parser.add_argument("--n_patients", type=int, default=25000) |
| parser.add_argument("--seed", type=int, default=42) |
| parser.add_argument("--output_dir", type=str, default="./output") |
| parser.add_argument("--format", type=str, default="csv", choices=["csv","parquet","json"]) |
| args = parser.parse_args() |
|
|
| os.makedirs(args.output_dir, exist_ok=True) |
| rng = np.random.default_rng(args.seed) |
| n = args.n_patients |
|
|
| print(f"[HC-ONC-007] Generating {n:,} synthetic leukemia patients (seed={args.seed})...") |
|
|
| print(" [1/8] Demographics...") |
| demo = generate_demographics(rng, n) |
|
|
| print(" [2/8] Disease characteristics...") |
| disease = generate_disease_characteristics(rng, demo) |
|
|
| print(" [3/8] Cytogenetics & molecular markers...") |
| mol = generate_molecular(rng, demo) |
|
|
| print(" [4/8] Treatment protocols...") |
| treat = generate_treatment(rng, demo, mol) |
|
|
| print(" [5/8] Remission & response...") |
| remission = generate_remission(rng, demo, mol, treat) |
|
|
| print(" [6/8] Bone marrow transplant outcomes...") |
| bmt = generate_bmt(rng, demo, mol, treat, remission) |
|
|
| print(" [7/8] Toxicity...") |
| tox = generate_toxicity(rng, demo, treat) |
|
|
| print(" [8/8] Survival outcomes...") |
| survival = generate_survival(rng, demo, mol, remission, bmt, tox) |
|
|
| |
| primary = pd.concat([demo, disease, mol, treat, remission, bmt, tox, survival], axis=1) |
| primary["generation_timestamp"] = datetime.now().isoformat() |
| primary["sku"] = "HC-ONC-007" |
| primary["version"] = "1.0.0" |
| primary["seed"] = args.seed |
|
|
| out_path = os.path.join(args.output_dir, "hc_onc_007_primary_cohort.csv") |
| primary.to_csv(out_path, index=False) |
| print(f"\n Primary cohort: {len(primary):,} rows × {len(primary.columns)} columns → {out_path}") |
|
|
| |
| print(" Generating MRD longitudinal data...") |
| mrd_long = generate_mrd_longitudinal(rng, demo, remission) |
| mrd_path = os.path.join(args.output_dir, "hc_onc_007_mrd_longitudinal.csv") |
| mrd_long.to_csv(mrd_path, index=False) |
| print(f" MRD longitudinal: {len(mrd_long):,} records → {mrd_path}") |
|
|
| |
| print("\n" + "="*60) |
| print("HC-ONC-007 Generation Complete") |
| print("="*60) |
| print(f" Patients: {len(primary):,}") |
| print(f" Columns: {len(primary.columns)}") |
| print(f" Subtype distribution:") |
| for st, cnt in primary["leukemia_type"].value_counts().items(): |
| print(f" {st}: {cnt:,} ({cnt/len(primary)*100:.1f}%)") |
| bmt_done = (primary["bmt_performed_flag"] == 1).sum() |
| print(f" BMT performed: {bmt_done:,} ({bmt_done/len(primary)*100:.1f}%)") |
| cr_rate = (primary["cr_achieved_flag"].astype(str).isin(["1","1.0"])).mean() |
| print(f" Overall CR rate: {cr_rate*100:.1f}%") |
| print(f" Output: {args.output_dir}/") |
|
|
| if __name__ == "__main__": |
| main() |
|
|
