Representative simulation and analysis code for Near-atomistic simulations reveal the molecular principles that control chromatin structure and phase separation.
This repository contains the code needed to inspect the principal computational workflows without duplicating every production directory or the very large simulation trajectories.
simulations/long_fibres/: the three 108-nucleosome systems (0 Ac, patterned 5 x 12 Ac, and block 54 Ac) plus their shared continuation script.simulations/bulk_tetramers/: 25 bp and 30 bp direct-coexistence preparation examples, their compact HREX seed inputs, and production-continuation scripts.simulations/hrex_12mer_25bp/: a representative Debye-length Hamiltonian replica-exchange setup and MPI multi-GPU driver.analysis/: cleaned notebooks used for Figs. 2-8 and the long-fibre contact-generation script.
OpenCGChromatin is maintained separately at https://github.com/CollepardoLab/OpenCGChromatin. The version associated with the paper will be archived and cited by DOI with the final release of this repository.
Create the analysis and simulation environment with:
conda env create -f environment.yml
conda activate russell-2026Install the archived paper version of OpenCGChromatin into that environment before running the simulation examples.
The notebooks do not include the multi-gigabyte production trajectories. Their configuration cells use /path/to/analysis_data as a deliberately non-functional placeholder. Replace DATA_ROOT or ROOT_DIR with the location of the corresponding trajectories, topologies, processed tables, and model-parameter files before running an analysis.
Run the notebooks from the analysis directory so their relative paths resolve consistently:
cd analysis
jupyter labGenerated figures and intermediate tables are written below analysis/outputs/ and are ignored by Git.
The bulk-tetramer preparation scripts include the required topology and 81 HREX seed configurations under each condition's inputs/ directory. Run prepare_coexistence.py from a separate output directory; generated simulation files are intentionally excluded. The production scripts start from the resulting modified_positions.pdb and automatically resume from checkpoint_nvt.chk on subsequent runs.
The HREX directory provides one representative 25 bp, 12-nucleosome example. The paper used the same protocol across the linker-length series described in the Methods. Within an existing SLURM allocation, launch one MPI rank per GPU with run_mpi_gpus.sh; set GPUS_PER_NODE if the allocation does not provide four GPUs per node.
The long-fibre scripts define the initial 20 microsecond production segment for each condition. Run the shared continue_simulation.py from a condition's output directory containing model_out.pdb and checkpoint.chk to append another 20 microsecond segment. These simulations were extended through checkpointed continuation jobs to the aggregate durations reported in the Methods; cluster-specific submission files are not included.
This repository is released under the MIT License.