FlexAppeal molecular dynamics, prepared and analysed

Which contacts carried the pose, and for how much of the scored window. Typed from heavy-atom geometry, so these are polar, ionic and hydrophobic contacts rather than a full interaction fingerprint.

When each contact was present, busiest first. Gaps are the contact breaking and reforming.

Per-residue backbone fluctuation, with the binding-site residues marked. A mobile pocket makes every pose measurement noisier.

Every frame against every other. Blocks along the diagonal are distinct states; a uniform field is a run that stayed in one.

Which parts of the receptor move together. Blue is correlated, red is anticorrelated.

Which parts of the molecule move. Peaks are usually loops and termini.

Per-residue assignment at every frame. Vertical stripes are transient melting; horizontal bands are stable elements.

How often each residue pair is in contact across the whole trajectory.

Each measure on its own axis. Energies, temperature and density differ by orders of magnitude, so sharing a plot between any two of them would make their alignment arbitrary and any apparent correlation fictional.

The whole protein as one number, against the experimental active and inactive structures of this same protein: +1 is the active references, -1 the inactive ones. The thin line is per frame and the bold one is smoothed. The shaded bands start at s = ±0.2, which is where a label is decided; inside them the protein reads as intermediate.

Each structural motif scored separately, in the colours section 6.1 fixes for them. The whole-protein score above is these weighted by residue count, which is why a small motif moving a long way can leave it almost unchanged. Motifs of unequal discriminating power carry equal residue weight, so read this panel rather than the headline number.

The last frame, motif by motif, with the distances the score is computed from and the nearest reference structure in each ensemble.

At 0.2 ns, the last frame analysed. A label is decided beyond s = ±0.2; inside that the motif reads as intermediate, which is a measurement and not a missing answer. The two half means say whether a motif was still moving: one number at the end cannot distinguish a motif that settled from one that was on its way somewhere. The last column is left blank where too few residues distinguish the reference sets to read a state from their secondary structure.

Secondary structure per residue per frame, with each motif marked down the left edge and the reference ensembles' own consensus in the two columns on the right. A motif whose secondary structure matches the inactive references and not the active ones is the clearest evidence this analysis produces, and it is visible along a row.

The reference structures on the plane their published criterion is drawn on, coloured by what the catalogue calls them, with this run's path across it in dark grey. This is the panel where a threshold that does not separate the clusters is obvious rather than argued about.

The ensemble this run was scored against, and whether our spatial label agrees with the annotation each structure came with. Counted, that agreement is the positive control for the whole method: if the labels did not reproduce the catalogue on structures whose state is known, nothing this page says about a trajectory would mean anything.

21 structures, resolved at run time from the catalogue rather than listed in the code. P3: the threshold on the TM3-TM6 distance, fitted on this set under leave-one-out, reproduces GPCRdb's own state call on 0 of 0 structures. No structure helped decide the line it is judged by. A structure reads "not used" when it has no usable state label or the alignment dropped it; the core panel says how many that was.