PROTEIN LOCAL MOTION
What it does
The protein local motion ready-made script has been designed to obtain some quick (approximate) conformational sampling of your protein system.
How it does
The control file performs ~100 steps using large ANM moves coupled to side chain optimization and a final constrained minimization. By default the backbone perturbation follows a direction randomly chosen for 6 Monte Carlo PELE steps (out of which it usually accepts ~30-40%). After these 6 steps, it randomly changes to another perturbation direction, which is based on 60% a given ANM mode plus 40% a random combination of the reminder of the modes.
The side chain prediction includes N side chains excited along the move (top_side N parameter, N=25 by default). The final minimization includes a 10 kcal/mol alpha carbon constraint to the structure derived from the initial perturbation (caconst 10).
We recommend playing around with different parameters that will adapt better the simulation to your specific system.
ANM modes type: it is possible to focus the sampling on one mode, with consecutive searches in the positive and negative direction. For this, use:
anm_altm_freq 10 & anm_altm_type 4 & lanmanm mode 4 &
Where we perform 10 positive, 10 negative, 10 positive… for mode number 4.
Here we can still keep the mixing of other modes by using:
lanmanm mix_modes 0.6 &
Where we use the 60/40% mix.
ANM mode length/frequency: by enlarging the size of the ANM move, how long we keep a mode, etc. we can obtain different sampled amplitudes. However, to a large extent (and when we perform an exhaustive search –thousands of points with a large number of processors), the level of sampling will depend mostly on the temperature, the ANM parameters affecting largely the acceptance/rejection ratio. The values used by default gave good results in our benchmark studies when comparing with microsecond molecular dynamics (see below). However, playing around with these parameters could provide valuable information. This is particularly true when performing quick explorations.
Omitting nodes: it is possible to eliminate nodes from the ANM matrix. Some protruding parts of the protein (such as C or N terminals, loops, etc) might take most of the motion from the ANM modes. While these parts surely move, its motion might not be important and can reduce the protein main moves. In these cases we delete them from the ANM network. Thus, we recommend comparing the results before/after omitting some of these regions. To omit a residue from the network add to the control file:
lanmanm omit_no A:204 A:210 &
Which will omit those 7 residues. Use A:204 A:204 to omit only one residue.
The outcome will include a series of pdb structures ready to visualize in any viewer (such as VMD). RMSD values to the initial structure, or to a native one, can be produced as well. Typically we use clustering techniques to further analyze the results. Keep in mind that when starting from a pdb structure, the initial steps will have a continuous energy decrease, reaching a plateau after several steps (equilibration process). As seen in our benchmark studies, these PELE trajectories constitute very good coordinates for further metadynamics free energy studies.
For quick analysis use the default script on ~8 processors. More extensive analysis would increase the number of steps (default is 100) and use ~16-24 processors.
Remember that the sampling is derived from an ANM model! Side chains and all-atom minimization might refine it but it is still a simple approximation to the dynamics.
Benchmark studies included comparison of the PELE simulations to 3 microscond MD trajectories on Ubiquitin. Comparison with metadynamics calculations on T4 Lysozyme was also performed.
- Cossins, B.P., Hosseini, A. & Guallar, V. Exploration of Protein Conformational Change with PELE and Meta-Dynamics [J. Chem. Theo. Comp. 8:959-965 (2012)].