Projected free energy
φ / ψSelect a map region. Arrow keys move the selection; the list below opens occupied bins.
Dark: more frequent. Light: less frequent. Gray: no observations; hatching: below the selected frame count.
FOLDIFF · MOLECULAR DYNAMICS
Which configurations occur more often — and is the sampling sufficient to trust the picture?
01 · INPUT DATA
CSV: trajectory,time_ps,phi_deg,psi_deg. At least 40 frames per trajectory, at most 250,000 total. All series must use the same time interval. JSON includes metadata and optional real 3D frames. A PDB alone is insufficient.
Example CSV · Example JSON project with 3DExample: three published 250 ns trajectories · 300 K.
02 · MAP & CONFIGURATIONS
Select a map region. Arrow keys move the selection; the list below opens occupied bins.
Dark: more frequent. Light: less frequent. Gray: no observations; hatching: below the selected frame count.
One saved frame from the selected bin, not an averaged structure. Rotate with mouse or touch.
03 · SAMPLING CHECKS
JSD: 0 means identical distributions, 1 means disjoint distributions. Values depend on grid and sample size. The 0.1 flag is exploratory, not a validated reliability threshold.
No autosave. Export your project to return to your data.
04 · METHOD & LIMITS
Fᵢ = RT ln(nₘₐₓ / nᵢ)
We divide the periodic φ / ψ plane into equal-area bins and count frames. In unbiased equilibrium sampling, frequency estimates probability. Its negative logarithm gives a projected free energy. The most frequent bin defines zero. Without adequate sampling, this remains an estimate from the observed data.
No pseudocounts or smoothing. An empty region means “no data”, not an infinite barrier. Four consecutive blocks per trajectory give a descriptive SD against one reference bin, not a confidence interval.
Metadynamics, umbrella sampling and other biased samples are not supported. We do not calculate transition rates, a full protein landscape or absolute mutant stability differences. Independently zeroed maps cannot be subtracted as ΔΔG. Domains and complexes first require suitable coordinates and a validated MD setup.
MDShare · ACEMD · AMBER ff-99SB-ILDN · TIP3P · 300 K.
The example retains every tenth frame: 75,000 of 750,000, at 10 ps intervals. Angles are recomputed from XTC. Time is reconstructed from frame index and the documented 1 ps interval: the XTC clock resets every 1,000 frames. Published coordinates omit solvent, although the simulation used water.
MDShare · source data · CC BY 4.0 · Coordinate validation · Formula validation · Full method and format