Theoretical conformational
sampling was achieved using full-atomistic equilibrium molecular dynamics.
Data were collected from five individual trajectory replicas of 1
μs length each. The trajectories were calculated using GROMACS
5.1.5 and GROMACS 2018.3,33 (link),34 employing CHARMM36m
force field parameters.35 (link) Modeling of the
initial systems was attained with CHARMM-GUI36 (link)−38 (link) using the X-ray
structures from PDB-ID4OOX (SAGA) or PDB-ID 2OBT (VA387), TIP3P water,39 (link) and 0.15 M NaCl ionization in a cubic box. The systems
were minimized with the steepest descent method and briefly equilibrated
for at least 0.1 ns in the NVT ensemble. For subsequent NPT production
sampling at 303.15 K, a Nosé–Hoover thermostat40 (link) and Parrinello–Rahman coupling41 (link) were employed. The simulation time step was
0.002 ps, and conformations were saved every 20 ps. For each protein,
5 trajectories of 1 μs length each were simulated. We note that
the VA387 simulations were performed later than the SAGA simulations,
which gave us access to much faster GPU nodes. To take full advantage
of the GPUs, we moved to a newer GROMACS version, with the consequence
that a few updates were made to the simulation protocol and input
parameters (see theSI ).
Data analysis
and visualization were carried out with VMD 1.9.3,42 (link) GROMACS tools, and the Python packages NumPy,43 (link) MDTraj,44 (link) and MatPlotLib.45 (link) Here, the side chain torsion angles of the Asn
residues, as well as the distances from the Cγ atoms of Asn
to the backbone nitrogen atoms of the subsequent amino acids, were
monitored. φ is defined as torsion angle between Ci-1-Ni-CAi-Ci, ψ between Ni-CAi-Ci-Ni+1, Χ1 between N-CA-CB-CG, and Χ2 between CA-CB-CG-OD1. The free energy maps were constructed
from the 2D probability densities as estimated by binning the data
to 100 × 100 bins of 2π/100 widths. The relative free energies
were computed as the negative natural logarithm of the probability
density. Clustering of the 4D torsional angle space was achieved with
the HDBSCAN46 (link) method using an extended
angle representation z(α) = [cos α, sin
α]. More details are given in theSI . One of the 5 MD trajectories of the SAGA P-dimer has been used
to generate conformers for ensemble docking in an earlier study.47 (link)
sampling was achieved using full-atomistic equilibrium molecular dynamics.
Data were collected from five individual trajectory replicas of 1
μs length each. The trajectories were calculated using GROMACS
5.1.5 and GROMACS 2018.3,33 (link),34 employing CHARMM36m
force field parameters.35 (link) Modeling of the
initial systems was attained with CHARMM-GUI36 (link)−38 (link) using the X-ray
structures from PDB-ID
were minimized with the steepest descent method and briefly equilibrated
for at least 0.1 ns in the NVT ensemble. For subsequent NPT production
sampling at 303.15 K, a Nosé–Hoover thermostat40 (link) and Parrinello–Rahman coupling41 (link) were employed. The simulation time step was
0.002 ps, and conformations were saved every 20 ps. For each protein,
5 trajectories of 1 μs length each were simulated. We note that
the VA387 simulations were performed later than the SAGA simulations,
which gave us access to much faster GPU nodes. To take full advantage
of the GPUs, we moved to a newer GROMACS version, with the consequence
that a few updates were made to the simulation protocol and input
parameters (see the
Data analysis
and visualization were carried out with VMD 1.9.3,42 (link) GROMACS tools, and the Python packages NumPy,43 (link) MDTraj,44 (link) and MatPlotLib.45 (link) Here, the side chain torsion angles of the Asn
residues, as well as the distances from the Cγ atoms of Asn
to the backbone nitrogen atoms of the subsequent amino acids, were
monitored. φ is defined as torsion angle between Ci-1-Ni-CAi-Ci, ψ between Ni-CAi-Ci-Ni+1, Χ1 between N-CA-CB-CG, and Χ2 between CA-CB-CG-OD1. The free energy maps were constructed
from the 2D probability densities as estimated by binning the data
to 100 × 100 bins of 2π/100 widths. The relative free energies
were computed as the negative natural logarithm of the probability
density. Clustering of the 4D torsional angle space was achieved with
the HDBSCAN46 (link) method using an extended
angle representation z(α) = [cos α, sin
α]. More details are given in the
to generate conformers for ensemble docking in an earlier study.47 (link)
Free full text: Click here
