Protein Conformational Transitions with GOdMD

What this workflow does

This workflow computes the transition pathway between two known states of the same protein. The example is adenylate kinase. The origin state is the closed form (PDB: 1AKE). The target state is the open form (PDB: 4AKE).

The workflow fetches both structures from the PDB. It extracts chain A from each one. It removes the AP5 inhibitor ligand from the origin structure. It computes a residue mapping between the two states with an EMBOSS sequence alignment. GOdMD then runs the discrete molecular dynamics transition. A final stage converts the trajectory to XTC format for visualization.

GOdMD uses a coarse-grained Maxwell-Demon method. It finds a physically plausible path between the two end states in minutes. An all-atom simulation of the same transition needs microseconds of sampling.

The compute problem

A transition study is a chain of small stages and one simulation stage. The hardware needs differ at each point.

The fetch, extract, and remove stages touch the network and a few kilobytes of text. They need a CPU and a few seconds each.

The godmd_prep stage runs an EMBOSS water alignment. It needs a CPU and one core.

The godmd_run stage is the only stage with a real compute cost. It samples many independent trajectories. More trajectories give a better path. A workstation core is enough for a small run. A large run belongs on a cluster node.

Most users run the whole chain on one laptop. They then limit the trajectory count to keep the run short. The result is a thin sampling of the pathway.

How Horus solves it

Horus assigns an executor to each stage. The GOdMD stage and the preparation stages are separate declarations. You change the executor: field on godmd_run alone to move it to a cluster node. The runtime.command string does not change.

This split matters here because the sampling stage scales with the trajectory count and nothing else does. You raise the trajectory count and move that one stage to a bigger machine. The fetch stages stay local and keep their network access.

biobb_godmd has a native osx-arm64 conda package. The whole workflow runs on Apple Silicon in the conda executor. You do not need Docker for this workflow.

Horus builds the conda environment from conda_env.yaml on the first run. It provisions GOdMD, EMBOSS, and the biobb stack. Later runs reuse the cached environment.

Pipeline

fetch_pdb_origin          Download 1AKE (closed state) from the PDB
   │
extract_chain_origin       Extract chain A
   │
remove_molecules_origin     Remove AP5 inhibitor ligand
   │                                    fetch_pdb_target          Download 4AKE (open state) from the PDB
   │                                       │
   │                                    extract_chain_target       Extract chain A
   │                                       │
   └───────────────────► godmd_prep ◄──────┘   Compute residue mapping (EMBOSS water alignment)
                              │
                         godmd_run              Run GOdMD conformational transition (1AKE → 4AKE)
                              │
                         cpptraj_convert         Convert trajectory mdcrd → XTC

Inputs and outputs

Inputs

The workflow has no file inputs. It fetches the two structures from the RCSB. configs/fetch_pdb_origin.yaml sets the origin PDB code. configs/fetch_pdb_target.yaml sets the target PDB code.

Outputs land in workflow_results/results/:

• 1ake.pdb, 4ake.pdb: the downloaded origin and target structures.
• 1ake.chains.pdb, 4ake.chains.pdb: the chain-A-only structures.
• 1ake.chains.nolig.pdb: the origin structure without the AP5 ligand.
• 1ake.aln, 4ake.aln: the residue-mapping alignments.
• 1ake-4ake.godmd.mdcrd: the raw GOdMD transition trajectory.
• 1ake-4ake.godmd.pdb: the GOdMD topology and reference structure.
• 1ake-4ake.godmd.xtc: the transition trajectory in XTC format.

Run the workflow

Install the horus-runtime one time:

uv sync

If you do not have uv, install it first:

curl -LsSf https://astral.sh/uv/install.sh | sh

You can also install the packages with pip:

pip install horus-runtime horus-environments

Then run the workflow:

uv run horus run workflow.yaml

The first run builds the conda environment. This takes a few minutes.

References

• BioExcel Building Blocks (biobb)
• biobb_godmd
• GOdMD web server
• Sfriso et al., Bioinformatics
• EMBOSS