Skip to content

cMD: barnase–barstar from a deposited crystal structure

On hold

Protein–protein complexes are on hold as of 2026-10-13 and are not part of the current release. This page was last run against 0.6.1. See On hold.

Tested against md-tools 0.6.1. Every command and every number on this page comes from a run executed as written, at that version.

Ordinary molecular dynamics on barnase–barstar (PDB 1BRS), the ribonuclease and its inhibitor, from the deposited crystal structure to 10 ns of production. This is the first protein–protein tutorial: two chains, an interface, crystal waters, residues with missing side-chain atoms, and PROPKA3 protonation. Every command below was run exactly as written and every number is copied from the files that run produced (one NVIDIA RTX 3080, CUDA, mixed precision).

Building took 16 s; minimisation, equilibration and 10 ns of production took 40 min.

Read cMD: paracetamol first; this page explains only what is new.

Starts from a built system. Do the system page first: it downloads 1BRS, chooses assembly 3, assigns protonation and writes build/built.xml, built.pdb and built.log.

1. Generate and run

A 30,000-atom protein complex gets longer equilibration than the shipped example's 10 ps stages. cMD.config at the dataset root:

protocol: cMD
solvent: explicit

stages:
  minimization_iterations: 5000
  restrained_nvt_steps: 50000      # 100 ps
  restrained_npt_steps: 50000      # 100 ps
  unrestrained_npt_steps: 50000    # 100 ps
  production_steps: 5000000        # 10 ns at 2 fs

reporting:
  crd_printout_solute: 5000        # 10 ps
  info_printout: 5000
  checkpoint_printout: 50000

cd ..                                            # 1BRS/
md-openmm build-md -odir ./cMD-run1 --config cMD.config
cd cMD-run1
CUDA_DEVICE_ORDER=PCI_BUS_ID CUDA_VISIBLE_DEVICES=8 ./run.sh

It finished in 40 min 42 s:

run.sh: all stages reported completion

2. Result

From cMD-run1/cMD.out:

Averages
  over                         1000 report(s) in mdout.csv
  Temperature (K)              mean 300.31   rms fluctuation 1.6943
  Density (g/mL)               mean 1.02756   rms fluctuation 0.00277362
  Speed (ns/day)               mean 356.876   rms fluctuation 12.0968

The structure of the complex over the run, from barnase_barstar_analysis.py (run it from 1BRS/):

frames 1000, time 310-10300 ps
complex CA RMSD (A): mean 0.84, last 1 ns 0.97, max 1.22
chain index 0 (108 CA): CA RMSD mean 0.72 A, last 1 ns 0.85 A
chain index 1 (89 CA): CA RMSD mean 0.73 A, last 1 ns 0.79 A
interface heavy-atom contacts <4.5 A: start 311, mean 269, last 1 ns 218
fraction of initial contacts kept: mean 0.63, last 1 ns 0.50

Both proteins stay within 1 Å of the prepared structure and the complex as a whole within 1.3 Å, so the fold and the assembly are stable. The interface is the looser number: it starts at 311 heavy-atom contacts, averages 269 over the run and is at 218 in the last nanosecond, with about half the individual starting contacts still made — side chains at an interface move even when the complex does not.

The contact counts are one realisation, and they are the numbers to trust least here

An earlier run of this page — a different build, 66 waters and 0.8 % of box volume apart — gave 297 mean and 286 last-nanosecond contacts against the 269 and 218 above, from the same commands. The CA RMSDs barely moved between the two; the contact counts moved by a quarter. That is what a single 10 ns trajectory of a flexible interface is worth: enough to say the complex did not come apart, not enough to quantify how the interface loosens. For that, run several independent replicates and report a spread.

Image the trajectory before measuring anything between the chains

The solute trajectory holds the chains as the periodic simulation holds them, so barnase and barstar can sit in different periodic images. Measured naively, this run shows a complex RMSD of 25 Å and 2700 "contacts". The analysis script makes each molecule whole and puts barstar in the image nearest barnase first (mdtraj.Trajectory.image_molecules).

What 10 ns shows, and what it does not. It shows that the prepared complex -- the chosen assembly, the built side chains, the PROPKA protonation -- is stable in explicit water on this timescale. It says nothing about binding affinity, and it does not sample the slower motions of the interface; for either, the run length and the analysis are a study of their own.

What it wrote

1BRS/
├── build/     1BRS.cif  build-top.config  built.xml  built.pdb  built.solute.pdb  built.log
│              assembly.json      the expanded chains, their author chains and operators
├── input/  min/
└── cMD-run1/  cMD.out  cMD.log  cMD.xml  solute_prod1.nc  mdout.csv  eq/  ...

Next