- Posts: 105
MARTINI simulation of protein-protein assosiation
- James Starlight
- Topic Author
- Visitor
separate ndx group and all of the rest of the protein (rest chains which should be centred in the trajectory before calculations of grid) as another group. Assuming that during make_ndx session to select chain A I just put Chain A (or
alternatively resid 1-100 for my case) how I can define selection
'Protein and not Chain A' or 'Protein and not resid 1-100' excluding
chain A or first 100 residues from the second selection?
Thanks for help again!
Please Log in or Create an account to join the conversation.
- Pim
- Offline
- Expert Boarder
So Protein and not residue 1 to 100 is "Protein &! r 1-100"
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
It's strange I have done all of those steps properly
1- MERGE ALL MY TRAJECTORIES
trjcat -f resp_complex_conf1.xtc resp_complex_conf2.xtc resp_complex_conf3.xtc resp_complex_conf4.xtc resp_complex_conf5.xtc resp_complex_conf6.xtc resp_complex_conf7.xtc -o resp_complex_merged.xtc -tu ps -cat
2- gmx densmap to calculate density plots of ligand defined in new2.ndx:
gmx densmap -f resp_complex_merged.xtc -s ref.pdb -n new2.ndx -od densmap.dat -w
and in that case obtained Segmentation fault error
Group 16 ( r_1-104) has 236 elements
Group 17 (Protein_&_!r_1-104) has 4117 elements
Select a group: 16
Selected 16: 'r_1-104'
Reading frame 1800 time 708000.000 Segmentation fault
Here ref.pdb consist of only protein atoms defined in groups 16 and 17 of new2.ndx
At the same time If I chose group of receptor (17 in that case) from the same inputs - it produces normal density map. Where here might be source of the error? What are the best option for the visualization of density maps? E.g I have not found any of PDB files produced. Is it possible to upload density map to VMD together with the ref.tpr for the its visualization ?
Thanks again!
J.
Please Log in or Create an account to join the conversation.
- Pim
- Offline
- Expert Boarder
- Posts: 105
If you want, you can make a PostScript image from the xpm output using xpm2ps -f file.xpm
Alternatively, you can take the .dat output and use any plotting program to make a 2D plot. I don't think you can load it into VMD.
About the first part: I have no idea. Check if perhaps one of your xtc files is broken or changes at time 708000? Unlikely, but if your xtc file is very big, maybe your computer is running out of memory and you can try to execute the operation on a trajectory without water atoms?
Please Log in or Create an account to join the conversation.
- flaviyan
- Offline
- Senior Boarder
- Posts: 54
look into the options gmx densmap options -aver this gives you the average density with respect to planes which is what you want. I am not sure about the segmentation fault.
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
By means of reducing number of the input trajectories has solved my problem.
Now will think about analysis in details.
But from the densmap.dat produced by
gmx densmap -f resp_complex_merged.xtc -s ref.tpr -n new2.ndx -od densmap.dat -aver z
it still unclear how to obtain some clues regarding possible interfaces of binging between both proteins - here it plots on X the actual distance on Z and on Y it's occupancy.
So some kind visualization of densities mapped on 3D structure might give me more insights I guess.
Might other analysis based on the merged trajectory be useful e.g something related to clusterization of binding interfaces?
Thanks again for the help!
J.
Please Log in or Create an account to join the conversation.
- flaviyan
- Offline
- Senior Boarder
- Posts: 54
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
Please Log in or Create an account to join the conversation.
- flaviyan
- Offline
- Senior Boarder
- Posts: 54
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
I also tried to found some solution based on the co-variance matrix method e.g to make PCA of the merged trajectory 1) selecting ligand atoms as the input group for the cov matrix and 2) project snapshots on the eigenvectors to see how the trajectory will be clusterized along first PCs but in my case it seems it might be useless because usually PCA captures only the internal molecular motions and the first six eigenvectors (translation and rotation which might be degrees of freedom by which sampling of the interfaces actually is achieved) are not included to the analysis.
MB someone know some server for the visualization of binding inter-phases?
J.
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
proc contactFreq { {sel1} {sel2} {percent 0} {outFile stdout} {mol top} } {
if { $outFile != "stdout" } {
set outFile [open $outFile w]
}
puts $outFile "[clock format [clock scan now]] Search started."
set allAtoms {}
set allCount {}
set numberOfFrames [molinfo $mol get numframes]
for { set i 0 } { $i < $numberOfFrames } { incr i } {
molinfo $mol set frame $i
set frameCount {}
set frameAtoms [atomselect $mol "$sel1 and within 4 of ($sel2)"]
foreach {segid} [$frameAtoms get segid] {resname} [$frameAtoms get resname] {resid} [$frameAtoms get resid] {name} [$frameAtoms get name] {
set atom [list $resid $resname $segid]
if {[lsearch $frameCount $atom] != -1} continue
lappend frameCount $atom
set loc [lsearch $allAtoms $atom]
if { $loc == -1 } {
lappend allAtoms $atom
lappend allCount 1
} else {
lset allCount $loc [expr { [lindex $allCount $loc] + 1 } ]
}
}
$frameAtoms delete
}
puts $outFile "[clock format [clock scan now]] Search finished."
puts $outFile "Find interactions:"
puts $outFile "Residue \t\tfraction"
#print count after sorting
set outData {}
foreach { a } $allAtoms { c } $allCount {
lappend outData [concat $c $a]
}
foreach { data } [lsort -integer -index 1 $outData] {
set c [lindex $data 0]
set fraction [expr { 100*$c/($numberOfFrames+0.0) }]
if { $fraction >= $percent } {
puts $outFile [format "%s-%s-%s \t\t %.2f%%" [lindex $data 3] [lindex $data 2] [lindex $data 1] $fraction]
}
#set beta according to the fraction, this is optional
set atom [atomselect $mol "segid [lindex $data 3] and resname [lindex $data 2] and resid [lindex $data 1]"]
$atom set beta $fraction
$atom delete
}
if { $outFile != "stdout" } {
close $outFile
}
}
assuming that I uploaded the mergd trajectory to vmd
and
defined selection 1 as chain A (this is smaller protein)
and selection 2 as not chain A (this is bigger protein)
I tried to execute
contactFreq "chain A" "not chain A"
but no residues are found.
Is it possible to adapt that script for the martini trajectories?
What another contact map based solutions might be useful here?
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
Briefly I need to determine binding interface (particularly to find residues which are contribute to binding from the receptor) between 2 proteins, emerged during several independent md runs (in my case all trajectories are merged together):
www.youtube.com/watch?v=u3vKKBq4G6s
I have tried to use contact maps based vizualisation aproaches in VMD and Pymol but unfortunately it was a problem with the representation of the martini atoms in both of them (e,g in vmd the sequence of martini model have not been recognized at all).
Thanks for help!
J
Please Log in or Create an account to join the conversation.
- flaviyan
- Offline
- Senior Boarder
- Posts: 54
If you are good with python I would recommend using the MDAnalysis library of tools which already has tools to calculate contact maps.
Another suggestion would be to look into the TCL script for the contactmap plugin form VMD and modify to recognize the CG protein residues.
Gromacs also has a tools to get contacts gmx mindist gmx mdmat
Hope this helps.
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
A good suggestions - thank you !
Also I guess prody python package could do the same things, right ?
Gleb
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
trying to calculate contact maps between 2 separate groups defined in index ndx
Group 16 ( r_1-104) has 236 elements
Group 17 (Protein_&_!r_1-104) has 4117 elements
unfortunately it has also produced an empty contact map :(
mb the atom selection in ndx should be made in another fashion?
J.
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
in addition I am interesting in the clusterization of binding interfaces established during MD e.g based on the position of one protein regarding another.
For my case I have performed such clusterization from MD movie plugin of CHIMERA which recognize martini's models quite well (which was based on current selection which was the atoms of smaller protein). As the result I obtained alot of overlapped clusters. Question - Is it possible here to increase accuracy of the clusterization and to remove some degrees of freedom which are not important for my case ? E.g to pre-proces trajectory by means of PCA before clusterization (e.g to discard rotation-translation from the protein) might be good or not?
Please Log in or Create an account to join the conversation.
- flaviyan
- Offline
- Senior Boarder
- Posts: 54
Please Log in or Create an account to join the conversation.
- flaviyan
- Offline
- Senior Boarder
- Posts: 54
Please Log in or Create an account to join the conversation.
- James Starlight
- Topic Author
- Visitor
Btw besides contact maps I also interesting to track hypothesis whether the electric field made by the charged residues located on the water-exposed surface of receptor is the main attraction force responsible for the binding of smaller protein. How it could be validated in the Martini's models ? Will the analysis of electrostatic surface of MD trajectory be good for this reason? Are there any special solutions for the software (by now Chimera are better for the Martini's trajectories than the VMD)
J.
Please Log in or Create an account to join the conversation.
- flaviyan
- Offline
- Senior Boarder
- Posts: 54
Please Log in or Create an account to join the conversation.