Protein homology modelling and molecular dynamics

Contents

Using Modeler

Using QwikMD

• Qwik Start
• VMD folder
• Considering solvation
• Atomselect
• Show hydrophobic molecules
• Graphics representations
• PDB columns
• The infamous patch grid error

Using Modeller

Modeller is used to model the protein structure of a target protein based on the protein structure of a template protein

1 Download & Installation (Key: MODELIRANJE)

2 Prepare the target

• Get protein sequence in fasta format from uniprot

3 Prepare the template

• BLASTp the target sequence against PDB to find a suitable template
• Get the pdb id of the template from the blastp results
• Search the protein data bank with the pdb id
• Download the pdb file & fasta sequence

4 Alignment

• Perform pairwise sequence alignment with Emboss Matcher
• Download the alignment file

5 Prepare modeller files

• Download the 2 files in the modeller folder (targetname.ali & targetname.py)
• In total, the .py file requires 4 changes and the .ali file requires 7 changes
• In both files replace all instances of "targetname" and "templatename"
• In the targetname.ali file, insert the aligned sequences (must be same length, truncate ends if necessary)
• NB: change nothing else in the .ali file, even the ":.:.:." have functions

6 Run modeller

• The working directory should now contain the pdb file and the edited .ali and .py files only

cd [working directory]
mod9.20 [targetname].py

• If you get an error referring to "mismatched sequences" or "unequal sequence length", just open the log file to find the mismatch, then edit the .ali file accordingly

7 Select best model

• You have instructed modeller to build 5 models ("a.ending_model = 5" in .py file)
• Modeller has created a pdf file for each model and a log file
• The most favourable model is the one with the lowest free energy
• Property Density Function is the measure of free energy with the units "molpdf"
• Scroll to the end of the log file where the 5 models
• The model with the lowest "molpdf" value is your best model
• To quick check the best model, load and align the model and template pdfs in Pymol

8 Model Evaluation

• Ramachandran plot with Rampage
• Analyze non-bonded interactions with Errat
• Determine the compatibility of an atomic model (3D) with its own amino acid sequence (1D) with Verify3D
• Multiple evaluation toolbox with SAVES

Using QwikMD

• Download VMD
• Install VMD

As a novice to molecular dynamics (MD), I have spent a few weeks following the tutorials and jumping in at the deep end. Take my advice, QwikMD is by far the best place to start. The image below is one of the QwikMD workflows on membrane proteins but also available are more extensive tutorials.

Qwik Start

• Extensions --> simulations --> QwikMD
• Easy Run --> load example.pdb
• Chain/Type selection : all
• Molecular dynamikcs: explicit, 0.15mol/L, NaCl
• Temp 27C
• Simulation Time: 10.0
• Simulation setup --> "save" --> select working directory + provide a file prefix --> "save"
• Prepare
• Copy and paste all 5 lines of Periodic Boundary Conditions (including header) from previous step into prefix/run/prefixqwikmd_production_1.conf
• Start equilibration
• View results:

mol delete all
mol load psf prefix_QwikMD.psf dcd qwikmd_equilibration_0.dcd

VMD folder

file	contents
instructions_cytosolic.txt	Commands to perform MD in VMD on cytosolic proteins
instructions_membranebound.txt	Commands to perform MD in VMD on membrane bound proteins
prepsystem.tcl	Example script to perform MD in VMD on membrane bound proteins
getBoxSize.tcl	Determines periodic boundary conditions for config file
template.config.namd	Sets all paramaters required for a single MD run in VMD

Considering solvation

Solvation tools tend to under solvate resulting in a system with too small density that causes the infamous patch grid error. Another more complex option is to first use DOWSER (to place buried waters), then Helmut Grubmuller's SOLVATE to generate a closely contoured solvent bubble around the solute. Then use VMD solvate to get the final cube. The main problem with VMD solvate is its inability to do a good job with matching the biomolecular surface, which is where DOWSER and Grubmuller SOLVATE excel. There is now a vmd interface for DOWSER.

Atomselect

Commands to use atomselect in TK console

set sel [atomselect top "water"] # selects all water molecules
$sel num # counts number of atoms in selection
$sel get name # gets names of all atoms in selection
$sel get {name backbone} # getting multiple attributes from each atom in selection
$sel get {x y z} # get coords of each atom in selection
$sel set {x y z} {{1.6 0 0}} # set coords of all atoms in selection
$sel set beta 0 # set beta of all atoms in selection --> now colour by beta

Show hydrophobic molecules

mol delete all
mol load psf example.psf pdb example.pdb
set all [atomselect top all]
$all set beta 2
set protein [atomselect top protein]
$protein set beta 1
set hydro [atomselect top hydrophobic]
$hydro set beta 0
#tk console --> graphics --> representations --> protein, beta, new cartoon 

Graphics representations

selected atoms	coloring method	drawing method
all not water	name	lines
protein	sec struct	newcartoon
(chain X and resid 102)	colorID 4	surf
water	name	points

PDB columns

column	3	4	5	11	12
content	name	resname	resid	segname (2nd last)	segid (last)

The infamous patch grid error

1 Cause:

The cell is shrinking too much. The system was started from a simulation box that is significantly larger than that of the equilibrated system. The reason why it shrinks so much initially, is that solvation tools under solvate resulting in a system with too small density. You get the error because NAMD only calculates the patch grid at the beginning of the simulation.

2 Solution:

Re-run but in protocol set temp to -213C and time to 0.5ns and leave overnight. The low temp just speeds up the first part of the simulation. The short time gives the tails a chance to reorient before the box starts moving. Stop after restart files are created. Re-start equilibration simulation again with the last configuration (restart files). Rinse and repeat until eventually you will get a box with relatively stable density, which will allow you to run long simulations without the error.

3 To re-start:

• Create a backup just in case something goes wrong
• sudo vmd
• rename qwikmd_equilibration_0.log so it won't be overwritten
• simulation setup --> load --> select prefix.qwikmd
• loading trajectories --> select qwikmd_equilibration_0 --> ok
• reset time to 10.0
• start equilibration simulation (may produce error, ignore, press button again)
• If reset error occurs again --> re-start with time at 0.5 until new restart file created
• IF no reset error --> re-start with temp at 27C

4 To know when the error has not occured:

After beginning the program there is always many lines stating "MINIMIZER RESTARTING CONJUGATE GRADIENT ALGORITHM DUE TO POOR PROGRESS" until finally, it will say "TCL: Setting parameter langevinpistontemp to 60, TCL: Running for 500 steps". The error then occurs within a few min on the next line "FATAL ERROR". If no error, it will write multiple lines beginning with "PRESSURE", "GPRESSURE", "PRESSAVG".