"""synthetic_generator.py Generate synthetic Seahorse-style plate CSVs covering 4 protocols. Design choices (seed=42, reproducible): - 4 protocols: Mito Stress, Glycolysis Stress, FAO Assay, ATP Rate Assay - 5 cell types: PMH (primary mouse hep), HepG2, AML12, HepaRG, LX-2 (HSC) - 5 drugs: resmetirom, pegozafermin, efruxifermin, lanifibranor, vehicle - 24 wells per plate (6 conditions x 4 replicate wells) - 12-15 timepoints with 4 injection steps per Mito Stress / Glyc Stress - baseline + cell-type baseline + drug-specific delta + injection response Outputs CSVs to data/synthetic/ relative to this file: - mito_stress.csv - glycolysis_stress.csv - fao_assay.csv - atp_rate.csv The schema (long-form): plate_id, well, group, cell_type, drug, dose, substrate, measurement, time_min, injection, ocr, ecar, ppr """ from __future__ import annotations import os import csv import math import random from dataclasses import dataclass from typing import Dict, List, Tuple import numpy as np SEED = 42 # directory containing this script HERE = os.path.dirname(os.path.abspath(__file__)) OUT_DIR = os.path.join(HERE, "synthetic") # --------------------------------------------------------------------------- # baseline matrices # --------------------------------------------------------------------------- CELL_BASELINES: Dict[str, Tuple[float, float]] = { # cell_type: (basal_ocr_pmol_O2_min, basal_ecar_mpH_min) "PMH": (180.0, 25.0), # primary mouse hep — high OXPHOS "AML12": (140.0, 35.0), # mouse hep line "HepG2": ( 90.0, 75.0), # warburg-ish "HepaRG": (160.0, 45.0), # diff hep line "LX-2": ( 70.0, 95.0), # HSC — glycolytic } # Drug effects (multiplicative on basal OCR / ECAR) DRUG_EFFECT: Dict[str, Tuple[float, float]] = { # drug: (ocr_mult, ecar_mult) "vehicle": (1.00, 1.00), "resmetirom": (1.15, 0.95), # THR-beta -> beta-ox up "pegozafermin": (1.20, 0.90), # FGF21 analog -> FAO up "efruxifermin": (1.18, 0.92), # FGF21 analog "lanifibranor": (1.10, 1.05), # pan-PPAR } CELL_TYPES = list(CELL_BASELINES.keys()) DRUGS = list(DRUG_EFFECT.keys()) # --------------------------------------------------------------------------- # protocol templates (injection sequences + multiplicative response factors) # --------------------------------------------------------------------------- @dataclass class Protocol: name: str injections: List[str] # phase labels per measurement # phase -> (ocr_factor, ecar_factor) multiplier relative to baseline response: Dict[str, Tuple[float, float]] def _make_phase_labels(phase_sequence: List[Tuple[str, int]]) -> List[str]: """Expand [(label, n_meas), ...] into a flat label list.""" out: List[str] = [] for lab, n in phase_sequence: out.extend([lab] * n) return out PROTOCOLS: Dict[str, Protocol] = { "mito_stress": Protocol( name="Mito Stress", injections=_make_phase_labels([ ("baseline", 3), ("Oligomycin", 3), ("FCCP", 3), ("Rot/AA", 3), ]), response={ "baseline": (1.00, 1.00), "Oligomycin": (0.45, 1.30), # ATP-linked OCR drops, ECAR comp up "FCCP": (1.95, 0.95), # maximal OCR "Rot/AA": (0.10, 0.70), # non-mito floor }, ), "glycolysis_stress": Protocol( name="Glycolysis Stress", injections=_make_phase_labels([ ("baseline", 3), ("Glucose", 3), ("Oligomycin", 3), ("2-DG", 3), ]), response={ "baseline": (1.00, 0.20), # no glucose -> ECAR floor "Glucose": (1.00, 1.00), # glucose triggers glycolysis "Oligomycin": (0.50, 1.60), # glycolytic capacity "2-DG": (0.50, 0.15), # ECAR back to non-glyc floor }, ), "fao_assay": Protocol( name="FAO Assay", injections=_make_phase_labels([ ("baseline", 3), ("Oligomycin", 3), ("FCCP", 3), ("Rot/AA", 3), ]), # FAO assay uses palmitate-BSA substrate; baselines higher for FAO-competent cells response={ "baseline": (1.00, 1.00), "Oligomycin": (0.40, 1.25), "FCCP": (1.85, 0.90), "Rot/AA": (0.12, 0.70), }, ), "atp_rate": Protocol( name="ATP Rate Assay", injections=_make_phase_labels([ ("baseline", 3), ("Oligomycin", 3), ("Rot/AA", 3), ]), response={ "baseline": (1.00, 1.00), "Oligomycin": (0.45, 1.30), "Rot/AA": (0.10, 0.70), }, ), } # --------------------------------------------------------------------------- # substrate matrices for FAO assay # --------------------------------------------------------------------------- # for FAO assay we vary substrate per well group; basal_ocr multiplier SUBSTRATE_OCR_MULT: Dict[str, float] = { "BSA": 1.00, "Palmitate-BSA": 1.30, "Palmitate+Etomoxir": 0.85, "Glutamine": 1.05, "Glutamine+BPTES": 0.80, "Glucose": 1.10, "Glucose+UK5099": 0.90, } # --------------------------------------------------------------------------- # plate generator # --------------------------------------------------------------------------- def _well_name(i: int) -> str: # 96-well: A1..H12 ; we only use up to 24 wells (A1-B12) rows = "ABCDEFGH" row = i // 12 col = i % 12 + 1 return f"{rows[row]}{col}" def _build_conditions(protocol_key: str) -> List[Dict[str, str]]: """Return list of well-condition dicts for a plate. Mito/Glyc/ATP: 5 cell types x 5 drugs -> too many; use a fixed cohort: 6 conditions: (HepG2,vehicle), (HepG2,resmetirom), (AML12,vehicle), (AML12,pegozafermin), (PMH, lanifibranor), (LX-2, efruxifermin) x 4 replicate wells = 24 wells FAO: 6 substrate conditions for HepG2+vehicle, 4 replicates each. """ if protocol_key == "fao_assay": substrates = ["BSA", "Palmitate-BSA", "Palmitate+Etomoxir", "Glutamine", "Glutamine+BPTES", "Glucose+UK5099"] return [ {"cell_type": "HepG2", "drug": "vehicle", "dose": "0", "substrate": s, "group": f"HepG2|vehicle|{s}"} for s in substrates ] base_conditions = [ ("HepG2", "vehicle", "0"), ("HepG2", "resmetirom", "1uM"), ("AML12", "vehicle", "0"), ("AML12", "pegozafermin", "100ng/mL"), ("PMH", "lanifibranor", "1uM"), ("LX-2", "efruxifermin", "100ng/mL"), ] return [ {"cell_type": c, "drug": d, "dose": dose, "substrate": "Glucose" if protocol_key == "glycolysis_stress" else "Standard", "group": f"{c}|{d}"} for c, d, dose in base_conditions ] def generate_plate_csv(protocol_key: str, out_path: str, seed: int = SEED) -> int: """Generate one plate CSV; returns number of rows written.""" rng = np.random.default_rng(seed + hash(protocol_key) % 10000) proto = PROTOCOLS[protocol_key] conditions = _build_conditions(protocol_key) replicates = 4 n_meas = len(proto.injections) rows: List[Dict[str, object]] = [] well_idx = 0 for cond in conditions: for rep in range(replicates): well = _well_name(well_idx) well_idx += 1 cell = cond["cell_type"] drug = cond["drug"] sub = cond["substrate"] base_ocr, base_ecar = CELL_BASELINES[cell] ocr_mult, ecar_mult = DRUG_EFFECT[drug] base_ocr *= ocr_mult base_ecar *= ecar_mult # substrate effect (FAO assay) if protocol_key == "fao_assay": base_ocr *= SUBSTRATE_OCR_MULT.get(sub, 1.0) # per-well biological noise base_ocr *= rng.normal(1.0, 0.06) base_ecar *= rng.normal(1.0, 0.08) for m_idx, phase in enumerate(proto.injections, start=1): of, ef = proto.response[phase] ocr_val = base_ocr * of * rng.normal(1.0, 0.04) ecar_val = base_ecar * ef * rng.normal(1.0, 0.05) # ensure non-negative ocr_val = max(0.0, ocr_val) ecar_val = max(0.0, ecar_val) time_min = m_idx * 7.0 # ~7 min per measurement rows.append({ "plate_id": protocol_key, "well": well, "group": cond["group"], "cell_type": cell, "drug": drug, "dose": cond["dose"], "substrate": sub, "measurement": m_idx, "time_min": round(time_min, 2), "injection": phase, "ocr": round(ocr_val, 3), "ecar": round(ecar_val, 3), "ppr": round(ecar_val * 0.8, 3), # rough proxy }) fields = list(rows[0].keys()) with open(out_path, "w", newline="") as f: w = csv.DictWriter(f, fieldnames=fields) w.writeheader() for r in rows: w.writerow(r) return len(rows) def main() -> None: os.makedirs(OUT_DIR, exist_ok=True) targets = [ ("mito_stress", "mito_stress.csv"), ("glycolysis_stress", "glycolysis_stress.csv"), ("fao_assay", "fao_assay.csv"), ("atp_rate", "atp_rate.csv"), ] total = 0 for key, fname in targets: path = os.path.join(OUT_DIR, fname) n = generate_plate_csv(key, path) total += n print(f"[gen] {fname}: {n} rows") print(f"[done] wrote {total} rows to {OUT_DIR}") if __name__ == "__main__": main()