News archive

Norman's paper on the Fundamental Redesign of the TIGER2hs Kernel to Address Severe Parameter Sensitivity is published

Figure 1. Schematic representation of different exchange kernels used in TIGER2 based simulations. The kernel for the originally published TIGER2 method is termed SE, while the ladder exchange later introduced in TIGER2A is termed LE. The new global pair exchange kernel developed in this study is termed PE and discards the sorting of replicas by potential energies. Note that these kernels are independent from the energy calculation methods applied, hybrid explicit/implicit solvent as in TIGER2h(s), or strong averaging as in TIGER2A. In principle, any kernel LE, SE, or PE can be combined with any TIGER-based energy calculation method TIGER2, TIGER2A, or TIGER2h(s). Also, the heating, sampling, and quenching phase cycles are preserved independent from exchange kernels or energy calculations. For all exchange kernels, the current baseline replica is highlighted in orange, and its state is written prior to the exchange attempt, except for LE, where multiple structures are registered during multiple exchange attempts in a consecutive manner. Diamonds represent the exchanging replica pairs.
Figure 5. Normalized secondary structure probability (y-axis) for all three tested peptides as a function of the average temperature change per exchange ⟨ΔT/χ⟩ (x-axis). The values are normalized against the corresponding REMD simulations, and a value of 1 on the y-axis denotes perfect agreement. Dashed lines denote a 10% interval from the REMD reference value. The former TIGER2hsSE algorithm (left) allows no reasonable prediction of replica number or Tmax to resemble the REMD results. The novel TIGER2hsPE kernel (right) shows strong clustering near 1 on the y-axis and below 30 on the x-axis. The quality of results can be estimated by adjusting ⟨ΔT/χ⟩ over changes in Tmax, regardless of the simulated peptide or replica number. For the Trp-cage, the sum of α- and 310-helix probability was considered.
Figure 7. (top) Cumulative α-helix probability along the conformational ensemble and (bottom) α-helix probability per residue for REMD, TIGER2APE, and TIGER2hsPE.

Norman's paper named Fundamental Redesign of the TIGER2hs Kernel to Address Severe Parameter Sensitivity published in the Journal of Chemical Information and Modeling. Congratulations Norman!


Code repository on Github


Zurück zu allen Meldungen

aktualisiert am 01.08.2023 durch
N. Geist