Skip to content

Analysis: c(RGDfV)

One notebook, over the unscaled state of an explicit-solvent REST2 ladder that is already registered:

notebook what it shows
clustering.ipynb the symmetry enumerated (four operations on the torsions, sixteen once the carboxylate and the guanidinium are included), the columns and their weights, t_hdbscan with the 1 % floor, every frame written in the labelling nearest the density peak, the backbone table, representative structures, how the choice of torsions changes the answer, and the resonance-equivalent atoms of the Asp and Arg side chains

The explanation is in Methods: symmetry-aware clustering and the torsion representation; the notebook shows results only.

The data

Registered dataset hpREST2/RGDfV-explicit-HMR-REST2 ($MD_DATA/2026/hpREST2/RGDfV-explicit-HMR-REST2): c(RGDfV), 79 solute atoms, explicit TIP3P, an 8-state ladder with tau up to 0.5, 500 ns per replica, three independent runs. State 0 is the unscaled Hamiltonian, so solute_state0_prod1.nc of each run is the physical ensemble; the other seven states are hot and are not used. The notebook reads every 75th frame of each run (375 ps, 1334 frames per run, 4002 in all).

The dataset was run with md-tools 0.5.4 and registered afterwards; this analysis ran on 0.6.5 with the three methods that are not released yet (THDBSCAN.density_peak, THDBSCAN.align_all, SymmetricTorsionDescriptor.align_to_reference, and the resonance option), and the notebook says so at the top.

To run it yourself:

micromamba activate analysis-env          # see Installing, step 5
export MD_DATA=/path/to/your/data-root
jupyter lab docs/tutorial/rgdfv/analysis/

What is not here

The notebook does not run a simulation and does not compare the REST2 ladder with a plain run, and the two clusters it finds are density basins, not metastable states: nothing in it looks at time. The 1 % floor, the torsion set and allow_single_cluster each change the answer, and the last section of the notebook shows by how much.