MDFF with explicit solvent

In the last section you learned how to set up a simple MDFF simulation in vacuo. Now you will learn how to set up a similar simulation in explicit solvent.

Preparing the initial structure

We will start with the structure that was already prepared in the previous section, i.e., files 1ake-initial_autopsf.psf and 1ake-initial_autopsf-docked.pdb.

1

Start a new VMD session.

2

Load the initial structure you prepared for MDFF in vacuo in the previous section:

mol new 1ake-initial_autopsf.psf
mol addfile 1ake-initial_autopsf-docked.pdb

3

Embed this structure into a water box using the solvate plugin. In the VMD Main Window, choose Extensions $\rightarrow$ Modeling $\rightarrow$ Add Solvation Box (Fig. 8). Set the box padding to 20Å for maximum y, 5Å for minimum y, and 10Å for the remainder dimensions, as shown in the figure and click on Solvate. It is important that the target map falls completely within the water box, which is why we chose a larger padding in one of the dimensions. VMD will generate the files solvate.psf and solvate.pdb, which will be automatically loaded upon completion of this step. Load the target map (from the previous section) to visually ensure that the density falls completely within the water box (Fig. 9).

Figure 8: The VMD Solvate window.

Figure 9: By loading the target density map you can visually ensure that the target density falls within the boundaries of the water box. If part of the target density is outside the water box, you need to adjust the padding accordingly and regenerate the water box.

4

For MD simulations in explicit solvent, it is usually desirable to have a neutrally charged system. This can be achieved by adding neutralizing counterions to the simulation system. One can also add additional ions to mimic in vivo or in vitro conditions. In this example, we will simply neutralize the system by adding either Na$^+$ or Cl$^-$ ions using the autoionize plugin. In the VMD Main Window, choos Extensions $\rightarrow$ Modeling $\rightarrow$ Add Ions (Fig. 10). Uncheck the button defining the ion concentration, leaving only neutralization active, as shown in the figure. Click on Autoionize. VMD will generate the files ionized.psf and ionized.pdb.

Figure 10: The VMD Autoionize window.

5

Generate a PDB file containing the per-atom scaling factors $w_j$ in Equation 1, as in the previous section. By default, water molecules and ions are not coupled to the target map, i.e., they can equilibrate freely according to the MD force field and don't experience additional forces from MDFF:

mdff gridpdb -psf ionized.psf -pdb ionized.pdb
-o ionized-grid.pdb

6

Generate secondary structure restraints as in the previous section:

package require ssrestraints
ssrestraints -psf ionized.psf -pdb ionized.pdb
-o ionized-extrabonds.txt -hbonds

7

Generate restraints to prevent cis/trans peptide transitions and chirality errors:

mol new ionized.psf
mol addfile ionized.pdb
cispeptide restrain -o ionized-extrabonds-cispeptide.txt
chirality restrain -o ionized-extrabonds-chirality.txt

Preparing the density map

As you can see in Fig. 9, we already ensured that the target density map (blue) is completely within the water box, which is a requirement for MDFF in explicit solvent. However, the current implementation of NAMD's gridforces feature also requires that the entire target map be within the water box, which is clearly not the case (Fig. 9). To address this limitation, we will trim the target map so that it lies completely within the water box.

1

Trim the resampled map in all dimensions by a few of Angstroms to ensure it will be within the water box during the simulation. The volutil plugin provides some features for manipulating volumetric maps. To trim the target map by 7Å in all dimensions, run:

package require volutil
volutil 4ake-target_autopsf-grid.dx -trim 7
-o 4ake-target_autopsf-grid-trimmed.dx

NOTE: If the target map and the box are rotated with respect to each other, it may be necessary to resample the target map into a cell that fits completely inside the water box, which is not yet supported by VMD.

2

Visualize in VMD the water box and the newly trimmed map to verify that it falls completely within the water box. Also ensure that the target macromolecular volume is contained in the new map. Load the trimmed map with the command:

mol new 4ake-target_autopsf-grid-trimmed.dx

Running the MDFF simulation

Generate NAMD configuration files similarly to the previous section.

1

For MDFF simulations in solvent we need to define periodic boundary conditions. We also use a different method to calculate electrostatic interactions that is more appropriate for this kind of simulation. All of this is taken care of by providing the extra option -pbc to mdff setup:

mdff setup -pbc -o adk-solvent -psf ionized.psf
-pdb ionized.pdb
-griddx 4ake-target_autopsf-grid-trimmed.dx
-gridpdb ionized-grid.pdb
-extrab {ionized-extrabonds.txt ionized-extrabonds-cispeptide.txt
ionized-extrabonds-chirality.txt} -gscale 0.3 -numsteps 100000

Note that we requested a simulation twice as long as in the previous section, since explicit-solvent MDFF simulations typically take longer to converge.

2

Once again, generate a second NAMD configuration file in which only energy minimization will be performed with a much higher scaling factor $\xi$:

mdff setup -pbc -o adk-solvent -psf ionized.psf
-pdb ionized.pdb
-griddx 4ake-target_autopsf-grid-trimmed.dx
-gridpdb ionized-grid.pdb
-extrab ionized-extrabonds.txt ionized-extrabonds-cispeptide.txt
ionized-extrabonds-chirality.txt} -gscale 10
-minsteps 2000 -numsteps 0 -step 2

3

Quit VMD.

4

Run NAMD using the configuration files generated by VMD, i.e., run the following commands in a terminal (or submit them to a cluster):

namd2 adk-solvent-step1.namd > adk-solvent-step1.log
namd2 adk-solvent-step2.namd > adk-solvent-step2.log

This step should take about 40 minutes on a cluster with 48 processors. If you don't want to wait, you can proceed to the next step and use the provided trajectory files, as explained in the next section.

Analyzing the results

The resulting trajectories will be saved to files adk-solvent-step1.dcd and adk-solvent-step2.dcd. If you want to continue working through the tutorial before the simulations are complete, you can use the provided trajectory files adk-step1-result.dcd and adk-step2-result.dcd instead. Once again, please note that due to the stochastic nature of molecular dynamics simulations it is expected that the trajectories obtained will differ from the ones provided. As in the previous section, load the resulting trajectory, as well as the target structure, and repeat the analysis of the RMSD ad CCC. Did the use of explicit solvent improve the MDFF results in this particular case?