OpenMpipi_for_IDRs

This repository contains Python scripts for OpenMpipi implementation for intrinsically disordered proteins simulating biomolecular condensates with the residue-resolution coarse-grained force field (OpenMpipi). The scripts are designed for GPU-accelerated, direct-coexistence simulations of charge-rich, intrinsically disordered proteins (e.g., hnRNPA1-LCD, DDX4-NTD), along with post-processing tools for phase diagrams, critical temperatures, density profiles, and performance benchmarking.

---

Repository structure

• slab_formation.py
  Two-stage direct-coexistence OpenMM protocol using OpenMpipi:

  1. Build compact initial chains and tile multiple proteins on a 3D grid.
  2. Compress the system via NPT to a target density.
  3. Stretch the box along one axis to form a slab and run NVT production.

• energy_decomposition_nvt_runs.py
  Continues an Mpipi-Recharged simulation from an equilibrated equi_model.pdb using OpenMM + CUDA (mixed precision), writing an trajectory.xtc and state outputs.

• density_profiles.py
  Uses MDTraj to load .xtc trajectories and corresponding .pdb files, wraps and centres coordinates, and computes 1D density profiles along the slab axis (x) over multiple temperatures. Outputs profiles to .csv and plots.

• critical_temperature.py
  Reads pre-computed coexistence densities from .csv, fits a binodal, and extracts the critical temperature Tc (and fit parameters) using non-linear least-squares. Optionally generates publication-quality plots.

---

Requirements

• Python 3.9+
• OpenMM (GPU build; CUDA recommended)
• The OpenMpipi / Mpipi-Recharged Python package (local or installed)
• Python libraries:
  • numpy, pandas, scipy, matplotlib
  • mdtraj (for density_profiles.py)

You can install typical dependencies with:

conda create -n openmpipi python=3.10
conda activate openmpipi
conda install -c conda-forge openmm mdtraj numpy scipy pandas matplotlib
# then install/clone your OpenMpipi package


Note: The original OpenMpipi is available at https://github.com/CollepardoLab/OpenMpipi