EMEWS-PhysiBoSS calibration and parameter sweep data for PI3K-MEK and AKT-MEK inhibitor combinations in a multiscale AGS cancer cell line model

Parameter exploration dataset with the PhysiBoSS AGS model This dataset contains computational simulation results from a multiscale modeling study of schedule-dependent drug synergy in AGS gastric adenocarcinoma cells. Throughout this dataset, inhibitor abbreviations are used as follows: PI3Ki (PI3K inhibitor), MEKi (MEK inhibitor), and AKTi (AKT inhibitor). The data were generated using the EMEWS framework to perform high-throughput parameter calibration and systematic parameter sweeps of a PhysiBoSS-based multiscale agent-based model. The dataset includes: (1) Single-drug calibration results using CMA-ES and Genetic Algorithm optimization for PI3Ki, MEKi, and AKTi, which identified optimal parameter sets enabling the model to reproduce experimental growth curves under single and synergistic drug combination perturbations; (2) Dose-response experiments exploring drug concentration effects across 100 concentration points; (3) Synergy sweep experiments testing 5000 parameter combinations for PI3K-MEK and AKT-MEK drug combinations in order to study the capability of the model of recovering the experimentally-validated drug synergies in this system; and (4) Drug timing and diffusion experiments in 2D and 3D microenvironments that revealed asymmetric temporal windows of efficacy. Key findings from these simulations include: the model calibrated only on single-drug data successfully predicted combination outcomes without combination-specific training; PI3K/AKT-first administration sequences produce optimal synergy; and MEK-first sequences show declining efficacy when the second drug is delayed beyond ~6 hours. All simulations integrate Boolean signaling networks (75-node AGS-specific network) with three-dimensional agent-based modeling to bridge molecular-scale signaling dynamics, cellular-scale fate decisions, and tissue-scale population dynamics. This dataset enables reproducibility, model validation, extended parameter analysis, and methodological development for other drug-cell line systems. For the publication, all plots and analysis have been derived from these datasets. NOTE: Full code can be found in https://github.com/bsc-life/ags_synergy_paper/tree/main Data structure Within each experiment folder, the data are organized as follows: Main Summary File:- final_summary_*.csv: Contains all parameters employed in the EMEWS simulations, including individual replicate and generation information (for CMA-ES and GA algorithms), along with the assessment metric (typically RMSE in the last column). This file contains the complete dataset for all parameter combinations tested. Filtered Top-Performing Sets:Each experiment folder also contains pre-filtered subsets sorted by lowest metric value (RMSE):- Top N files: top_10.csv, top_20.csv, top_100.csv, top_200.csv - contain the N best-performing parameter sets- Top percentage files: top_1p.csv, top_5p.csv, top_10p.csv, top_25p.csv - contain the top 1%, 5%, 10%, and 25% of parameter sets by performance Additional Files:- final_summary_*_param_distribution_*.json: Parameter distribution statistics for top-performing sets Calibration experiments (CMA-ES and GA) For each drug (PI3Ki, MEKi or AKTi), we perform both a parameter optimization based on two different strategies: GA (Genetic Algorithm) or CMA-ES (Covariance Matrix Adaptation). Here, "18p" refers to 18 parameters used for this parametrization experiment. The following experiments can be found in this dataset: PI3Ki_CMA-0704-1815-18p_delayed_transient_rmse_postdrug_25gen PI3Ki_GA-0704-1815-18p_delayed_transient_rmse_postdrug_25gen MEKi_CMA-0704-1815-18p_delayed_transient_rmse_postdrug_25gen MEKi_GA-0704-1815-18p_delayed_transient_rmse_postdrug_25gen AKTi_CMA-0704-1815-18p_delayed_transient_rmse_postdrug_25gen AKTi_GA-0704-1815-18p_delayed_transient_rmse_postdrug_25gen Parameter sweep experiments These EMEWS experiments are uniform or random sampled points from a previously defined parameter space. Here, we find different types of experiments: Dose-Response Experiments (dose_response/): Systematic exploration of drug concentration effects across 100 concentration points for PI3Ki, MEKi, and AKTi. Organized by drug in subfolders:- PI3Ki/: DR_CURVE_PI3Ki-sweep-*-100-drug_threshold-1p- MEKi/: DR_CURVE_MEKi-sweep-*-100-drug_threshold-1p- AKTi/: DR_CURVE_AKTi-sweep-*-100-drug_threshold-1p Synergy Sweep Experiments (synergy_sweeps/): Parameter space exploration for drug combination simulations, testing 5000 parameter combinations sampled from the top-performing single-drug calibrations. Organized by drug combination and experiment type: PI3K-MEK synergy sweeps:- PI3K_MEK_combination/: synergy_sweep-pi3k_mek-*-18p_transient_delayed_uniform_5k_10p- PI3K_MEK_single_drug_PI3K/: synergy_sweep-pi3k_mek-*-18p_PI3K_transient_delayed_uniform_5k_10p- PI3K_MEK_single_drug_MEK/: synergy_sweep-pi3k_mek-*-18p_MEK_transient_delayed_uniform_5k_10p AKT-MEK synergy sweeps:- AKT_MEK_combination/: synergy_sweep-akt_mek-*-18p_transient_delayed_uniform_5k_10p- AKT_MEK_combination_postdrug/: synergy_sweep-akt_mek-*-18p_transient_delayed_uniform_postdrug_RMSE_5k- AKT_MEK_single_drug_AKT/: synergy_sweep-akt_mek-*-18p_AKT_transient_delayed_uniform_5k_singledrug- AKT_MEK_single_drug_MEK/: synergy_sweep-akt_mek-*-18p_MEK_transient_delayed_uniform_5k_singledrug Drug Timing Experiments (drug_timing/): Systematic investigation of how drug administration timing and spatial diffusion affect therapeutic outcomes in both 2D and 3D microenvironments. Organized by spatial dimension:- 2D/: synergy_sweep-akt_mek-*-4p_2D_drugtiming_synonly_consensus_hybrid_20, synergy_sweep-pi3k_mek-*-4p_2D_drugtiming_synonly_consensus_hybrid_20- 3D/: synergy_sweep-akt_mek-*-4p_3D_drugtiming_synonly_consensus_hybrid_20, synergy_sweep-pi3k_mek-*-4p_3D_drugtiming_synonly_consensus_hybrid_20