Tutorial 2 — Multidomain proteins & per-domain contact scaling

Goal: simulate a multidomain protein and control the contact-energy nscale within each domain and across domain interfaces using a domain.yaml file. We use adenylate kinase (1AKE chain A, 214 residues), a textbook multidomain enzyme.

Time: ~4 seconds on a CPU.

Prerequisite: do Tutorial 1 first.

1AKE 300 K trajectory

The run — Langevin dynamics at 300 K; per-domain and interface nscale keep both domains folded.

Regenerate after your run with python ../_viz/render_cg.py --psf traj/traj.psf --dcd traj/traj.dcd --out img --hero 0 --stride 2.

Files in this folder

File

Role

1AKE_A.pdb

Input structure (adenylate kinase, chain A).

domain.yaml

Domain definition — the new ingredient in this tutorial.

md.ini

Simulation config; note the added domain_def = domain.yaml line.

run_simulation.py

The runner script.

Why domains matter

In a structure-based model, every native contact gets an attractive well. By default all wells have the same relative nscale. But a multidomain protein often needs different stabilities for each domain and for the interface between them (e.g. to reproduce experimental melting temperatures, or to study domain-wise unfolding). domain.yaml lets you set a scale factor on the sidechain–sidechain contact energy:

  • within each domain (intra_domains[...].nscale), and

  • between domains (inter_domains).

The key idea: discontiguous domains

Adenylate kinase’s CORE domain is not a single stretch of sequence. It is made of two segments — residues 1–117 and 166–214 — that fold together around the inserted NMP-binding domain (118–165). TOPO handles this by letting one domain list multiple residue ranges.

Open domain.yaml:

n_residues: 214
intra_domains:
  A:
    residues: [1-117, 166-214]   # ONE domain, TWO sequence segments
    nscale: 1.1556
  B:
    residues: [118-165]
    nscale: 1.6871
inter_domains:
  A-B: 1.8611

Reading it out:

  • Domain A = residues 1–117 and 166–214; all A–A native contacts (including contacts between the two segments) scaled by 1.1556.

  • Domain B = residues 118–165; B–B contacts scaled by 1.6871.

  • A–B interface contacts scaled by 1.8611.

Note: an unspecified interface defaults to scale 1.0 (native contacts kept, unscaled) — the scaling only modulates contacts that already exist in the structure. To intentionally decouple two domains, set their pair to 0.0 explicitly. Full rules: Domain definition file.

Step-by-step

1. Note the one change in md.ini

Compared with Tutorial 1, the only addition is:

domain_def = domain.yaml

That single line switches on domain-aware contact scaling.

2. Run it

python run_simulation.py -f md.ini

In the build log you’ll see Domain definition file: domain.yaml, a contact matrices: (214, 214) line and a native contacts: … | non-native pairs: … summary, confirming the 214×214 contact matrices were built with your scaling applied.

3. Outputs

Same set as Tutorial 1, in the traj/ run folder: traj/traj.log, traj/traj.dcd, traj/traj.chk, traj/traj.psf, traj/traj_final.pdb, traj/traj_runinfo.log (plus the cached STRIDE *_stride.dat next to the input PDB).

Build your own domain.yaml

  1. Decide how many domains and which residues belong to each (a discontiguous domain just gets several ranges in its residues list).

  2. Set n_residues to the true residue count (must cover all residues; any residue you forget is auto-assigned to a fallback domain X with nscale 1.0).

  3. Give every domain a numeric nscale (a blank value errors).

  4. Add an inter_domains entry only for interfaces whose nscale you want to change from the default of 1.0 (set 0.0 to remove an interface; omitted pairs stay at 1.0).

The reference doc has six worked scenarios (single domain, contiguous, discontiguous, decoupled, 3+ domains, partial assignment): Domain definition file.

Try next

  • Make domain B much weaker (e.g. nscale: 0.5), run, and watch B unfold first at elevated temperature while A stays folded.

  • Set inter_domains: A-B: 0.0 to decouple the domains and observe them moving independently.

  • Continue to Tutorial 3 to learn restarting and the output files in depth.