ScooBDoob

Setup

Use Python 3.10 or newer. Run commands from this directory.

python -m pip install -r requirements.txt
python -m pytest tests -q

Contents

• scoobdoob/: conditioning, MSM construction, and analysis functions.
• make_*.py, paper_figures/: paper analyses and plots.
• analysis/, data/: sensitivity analyses and saved numerical results.
• tests/: numerical correctness checks.

Paper analyses

Generate the MB, Aib9 robustness, and p53 plots from the included results:

python run_cached_figures.py

Outputs are written to figures/. For a subset, use --only mb, --only aib9_robustness, or --only p53.

Analysis	Script
MB sampling comparisons	make_mb_panels.py, make_mb_unconditional.py
MB replicated statistics and parameter sweeps	mb_stats.py, mb_alpha_seeds.py, mb_ticalag_seeds.py
MB density-floor sensitivity	analysis/mb_floor_sensitivity/analyze.py
Aib9 sampling comparison	make_aib9_method_figure.py
Aib9 robustness and density sweep	validate_aib9.py, aib9_kappa_sweep.py
p53 discovery and preference	paper_figures/data/, paper_figures/plot/
p53 trajectory-held-out analysis	trajectory_cv_tilt.py, summarize_trajectory_cv_tilt.py

After preparing the inputs below, rerun the main analyses with:

python make_mb_panels.py --cache-only
python make_mb_unconditional.py --cache-only
python make_aib9_method_figure.py
python validate_aib9.py --robustness
python aib9_kappa_sweep.py
python paper_figures/data/p53_preference_data.py --splits 400

For the MB floor sweep, run analyze.py --seed N for seeds 0–19, then analyze.py --summarize. Use --validate for its exact finite-model check.

Data

MD trajectories are not distributed with this package. Small numerical results and docking scores are included; additional analysis inputs are listed in INPUTS.tsv.

Müller–Brown

Generate 12 overdamped Langevin trajectories with the paper settings (250,000 steps each, timestep 0.0002, inverse temperature 0.06, saving every five steps):

python -m scoobdoob.validation.mb_data --out data/mb_reactive

Aib9

Use the published Aib9 OpenMM workflow from Mehdi et al. (2022) to prepare and simulate the solvated peptide. The paper analyzes a continuous, approximately 223 ns trajectory at 500 K with a 2 fs timestep. Record the nine backbone phi and psi angles in radians every 0.2 ps using PLUMED.

Place COLVAR_phi and COLVAR_psi in data/aib9/, or set SCOOBDOOB_AIB9_DATA to their directory. Each file should contain time followed by nine angles. The loader takes every fifth sample, giving 1 ps spacing.

p53 Y220C

The supplied numerical results support plotting and docking-preference analysis. MSM refits require the feature and pocket-observable files listed in INPUTS.tsv. Use the simulation and feature definitions in the paper's Methods.

For held-out fits, use --k 75, 100, or 150 and --seed 0, 1, or 2. The observable/threshold pairs are site_drug/0.5, site_volume/300.865, and site_apolar/55.551. For example:

python trajectory_cv_tilt.py --k 100 --seed 0 --observable site_volume \
    --drug-thresh 300.865 --out data/p53/trajectory_cv_tilt_vol_k100_seed0.json
python summarize_trajectory_cv_tilt.py --pattern 'trajectory_cv_tilt_vol_k*_seed*.json' \
    --out data/p53/trajectory_cv_tilt_vol_robustness.json