These are the trajectories behind the branch C ala2 results. In the top panel, an unbiased L2 run hops between the C7eq/β and αR basins. In the bottom panel, well-tempered metadynamics builds up a bias until the walker reaches basins plain MD never visits. Water was simulated but isn't shown.
The backbone core is held fixed, so you see the φ and ψ rotations rather than the whole molecule tumbling. The dashed green line appears when the C7 hydrogen bond (ACE O to NME H) is shorter than 2.5 Å.
A Gaussian hill is added every picosecond where the walker is. The shading is the running free-energy estimate, −γ/(γ−1)·V, shifted so its minimum is zero. Hills shrink as the bias grows, which is the well-tempered part. Opens on the final surface; press Play to watch it build.
Basin boundaries come from the branch C analysis config. C7eq/β is φ < 0 with ψ ≥ 50° or ψ < −120°, plus φ ≥ 120°. αR is φ < 0 with −120° ≤ ψ < 50°. αL/C7ax is 0 ≤ φ < 120°.
Sources: runs/c/ala2/L2-md_histogram-r0-b3c7163251e5/rep_2 and runs/c/ala2/L3-wtmetad_reweight-r0-0cfdc6ec0a11/rep_0. Metadynamics settings: σ = 0.35 rad, initial height 1.2 kJ/mol, bias factor γ = 6, one hill per picosecond. The free-energy estimate is unreweighted and only for illustration. The branch C numbers come from the reweighted analysis.