normal Analysis of the protein-protein interphaces established in CG simulations

  • James Starlight
  • James Starlight's Avatar Topic Author
  • Visitor
6 years 4 months ago - 6 years 4 months ago #7425 by James Starlight
Dear Martini users!

I am focusing on the simulation of GPCR oligomerization using MARTINI ELNEDYN model.

My system consists of 16 GPCRs embedded within big membrane and solvated with non-polarizable water and ions. The aim is to analyse per-residual contacts established between oligomers of GPCRs, that were observed during microsecond runs. Besides, I would like to clarify the binding interphases (e.g. particular alpha-helices of each monomer, which may form the stacking areas for the dimers).

I would be very grateful for the suggestion of some method for the clusterization of the binding interphases. For the first part I would like to avoid backmapping, working just with the CG models. The hypothetical workflow is to i) accept as the input the pdb file(s) consisted of 16 CG GPCRs in different oligomeric states and ii) to provide some statistics of the convergence between established contacts of the GPCRS observed in different MARTINI simulations (in case when I provide several pdbs as the input).

I thanks everyone very much for the help!

James
Last edit: 6 years 4 months ago by James Starlight.

Please Log in or Create an account to join the conversation.

More
6 years 4 months ago #7427 by peterkroon
Hmmm... Tough cookie.

Your analysis won't be fundamentally different from what you would do if it were an atomistic simulation. It'll still be hard to do proper though. You'll probably need to write your own; have a look at MDAnalysis (a Python module)

I would start by looking at the contact matrix between a GPCR and it's nearest neighbour (maybe 2 neigbours? Or all?) and see if that provides intelligble data. You can look at different things: the number of contacts throughout your simulation, or the average/maximum duration of a single contact, ...; both will tell you different things.
I would expect these matrices to be symmetric, so you can then reduce them to the diagonal, and get a normal plot with the number of contacts per residue, this will then give you binding sites.
Convergence you can study by the usual means, by doing a running average/standard deviation of your observables. You just end up with more dimensions in this case ;)

Please Log in or Create an account to join the conversation.

  • James Starlight
  • James Starlight's Avatar Topic Author
  • Visitor
6 years 4 months ago #7430 by James Starlight
Hello,

thank you so much for the info!

Indeed, the contact matrices between the martini beads of different protomers is what I would like to do in order to make the analysis of the CG models. Does the VMD or CHIMERA accept the martini models? As I know in Chimera there is a plug-in for the clusterisation of the pdbs, which could be useful for the clusterisation of the binding interfaces, couldn't it?

Some methodological questions: how I can select the individual GPCR monomer in the gro file consisted of whole system, assuming that it has been made via the replication of the single unit using genconf (so the protomers are not defined as individual chains)?

James

Please Log in or Create an account to join the conversation.

More
6 years 4 months ago #7432 by peterkroon

Indeed, the contact matrices between the martini beads of different protomers is what I would like to do in order to make the analysis of the CG models. Does the VMD or CHIMERA accept the martini models? As I know in Chimera there is a plug-in for the clusterisation of the pdbs, which could be useful for the clusterisation of the binding interfaces, couldn't it?

I'd expect so, but don't know for sure. Try it.

Some methodological questions: how I can select the individual GPCR monomer in the gro file consisted of whole system, assuming that it has been made via the replication of the single unit using genconf (so the protomers are not defined as individual chains)?

Make an index file using gmx make_ndx or gmx select.

Please Log in or Create an account to join the conversation.

  • James Starlight
  • James Starlight's Avatar Topic Author
  • Visitor
6 years 4 months ago #7435 by James Starlight
OK, thanks !
I have seen the publications where the people do the clusterization of the interfaces based on the Coarse-grained inputs (not using the back-mapping)..

One question - will gmx make_ndx work for the gro file consisted of the replicated system, which has the topology and the order of the atoms like

[ molecules ]
; the 1st protomer
Protein 1
POPC 118
CHOL 29
POPC 120
CHOL 30
W 6243
NA+ 62
CL- 76
; the 2st protomer
Protein 1
POPC 118
CHOL 29
POPC 120
CHOL 30
W 6243
NA+ 62
CL- 76
; the 3st protomer
Protein 1
POPC 118
CHOL 29
POPC 120
CHOL 30
W 6243
NA+ 62
CL- 76

?

Thanks !!

Please Log in or Create an account to join the conversation.

More
6 years 4 months ago #7439 by riccardo
I would expect gmx make_ndx to work, but trying is the best way to find out for sure.

Please Log in or Create an account to join the conversation.

  • James Starlight
  • James Starlight's Avatar Topic Author
  • Visitor
6 years 3 months ago #7450 by James Starlight
Yes, as I expected, the make_ndx recognizes correctly all the monomers (as protein group) as a single entity. Based from it, what should be trivial method to qualify statistically the binding events e.g based on the distance measurements?

I have tried to do it via UCSF Chimera - via find clushes/contacts. Unfortunately it has recognized the inter-monomer contacts as well, thus supposing that Chimera did not recognized MARTINI models correctly.

Please Log in or Create an account to join the conversation.

More
6 years 3 months ago #7452 by bart
Not sure how everything works you used (Chimera), but if you use an ndx file which has all protein chains in on selection, I wouldn't expect it differently. I am not sure if you can easily select the interactions between two (or more) index groups. If that is possible it would be easy to just make a selection of the individual chains of your protein (or to generate it yourself, an ndx file is very simple, just open it with vi and take a look). Another thing you could consider, but might hurt a bit, is that you could have a brief simulations of all components far from each other. Calculate the contact as you did and substract that number of contacst you get from the contact simulations. This should roughly be right, for the intramolecular contacts should not vary (much) over time.

Please Log in or Create an account to join the conversation.

More
6 years 3 months ago #7453 by bart
Not sure how everything works you used (Chimera), but if you use an ndx file which has all protein chains in on selection, I wouldn't expect it differently. I am not sure if you can easily select the interactions between two (or more) index groups. If that is possible it would be easy to just make a selection of the individual chains of your protein (or to generate it yourself, an ndx file is very simple, just open it with vi and take a look). Another thing you could consider, but might hurt a bit, is that you could have a brief simulations of all components far from each other. Calculate the contact as you did and substract that number of contacst you get from the contact simulations. This should roughly be right, for the intramolecular contacts should not vary (much) over time.

Please Log in or Create an account to join the conversation.

More
6 years 3 months ago #7454 by bart
Not sure how everything works you used (Chimera), but if you use an ndx file which has all protein chains in on selection, I wouldn't expect it differently. I am not sure if you can easily select the interactions between two (or more) index groups. If that is possible it would be easy to just make a selection of the individual chains of your protein (or to generate it yourself, an ndx file is very simple, just open it with vi and take a look). Another thing you could consider, but might hurt a bit, is that you could have a brief simulations of all components far from each other. Calculate the contact as you did and substract that number of contacst you get from the contact simulations. This should roughly be right, for the intramolecular contacts should not vary (much) over time.

Please Log in or Create an account to join the conversation.

  • James Starlight
  • James Starlight's Avatar Topic Author
  • Visitor
6 years 3 months ago #7455 by James Starlight
Hi Bart and thank you for the suggestions!

Actually, ndx file consist of one "protein" entity which is shared between all the protomers of my system. I think it would be difficult to manually edit each time an ndx file (assuming that I will be focused on the analysis of different MARTINI systems e.g where the same number of protomers were simulated in different membranes), wouldn't it?

Alternatively, what I can do via Chimera is to assign for each of the protomers the unique chain name which will be than recognized by VMD or CHIMERA (along with the loaded trajectory). Does someone know some analysis tool in VMD which might be useful to quantify the binding? E.g. what I have already tried is to calculate a radial distribution function for each of the chain separately (e.g between chain A and "not chain A" etc) to find the statistic distribution of the distances...

Please Log in or Create an account to join the conversation.

Time to create page: 0.121 seconds