Mode-Mixing MCMC Enhances Sampling Efficiency for G-Quadruplex Ligand Binding Pose Sampling under a Walltime-Normalized Benchmark

Benchmark data for MM-MCMC sampling efficiency evaluation on G-quadruplex ligand binding This dataset supports the manuscript: Tanigawa, M. & Iwaki, T. "Mode-Mixing MCMC Enhances Sampling Efficiency for G-Quadruplex Ligand Binding Pose Sampling under a Walltime-Normalized Benchmark." Journal of Chemical Information and Modeling (2026), under review. Overview Mode-Mixing MCMC (MM-MCMC) is an adaptive sampling strategy that learns the covariance structure of the sampled distribution and proposes collective rigid-body moves to accelerate conformational sampling. This repository contains benchmark data comparing MM-MCMC with standard Metropolis Monte Carlo under identical 60-minute walltime budgets. Key Results 4.9-fold geometric mean improvement in effective sample size per second (ESS/s) Ligand-dependent speedups ranging from 1.3× to 14.7× 108 independent simulations (6 ligands × 3 poses × 3 seeds × 2 methods) Contents data/benchmark_results.csv — Individual simulation results (108 rows) data/benchmark_summary.csv — Summary statistics by ligand src/mcmc_core.py — MM-MCMC implementation src/surrogate_potential.py — Surrogate energy function src/ess.py — Effective sample size estimation Methods Target: c-MYC promoter G-quadruplex (PDB: 6AU4) Ligands: DC-34, PhenDC3, Pyridostatin, BRACO-19, DAPI, Caffeine Surrogate potential: Coulomb (ε = 4r) + Lennard-Jones (AMBER ff14SB) ESS estimation: Initial monotone sequence estimator (Geyer, 1992)