- Posts: 14
Cutoff of deleting lipids around protein
- zhiyue
-
Topic Author
- Offline
- Fresh Boarder
Less
More
9 years 11 months ago - 9 years 11 months ago #3844
by zhiyue
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Cutoff of deleting lipids around protein was created by zhiyue
Hi all,
I'm trying to set up a protein-membrane complex. After I embeded the MARTINI protein into the MARTINI lipid bilayer, I tried to remove the overlapping lipids with protein. But I'm not sure the most appropriate cutoff value. Does anyone have any suggestion? Thanks in advance.
Best
I'm trying to set up a protein-membrane complex. After I embeded the MARTINI protein into the MARTINI lipid bilayer, I tried to remove the overlapping lipids with protein. But I'm not sure the most appropriate cutoff value. Does anyone have any suggestion? Thanks in advance.
Best
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Last edit: 9 years 11 months ago by zhiyue.
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
9 years 11 months ago #3845
by mnmelo
Replied by mnmelo on topic Cutoff of deleting lipids around protein
Hi zhiyue,
Martini is quite robust, and will probably tolerate a couple of overlaps if you let it minimize properly. In any case, keep in mind that equilibrium distances between Martini beads are around 0.5 nm, so that's a good minimum clearance to begin with. Increase it if you can't get a proper minimization afterwards.
If you have proteins with a ring-like topology (channels, for instance) you'll also have to take care of removing the lipids that fall within the hollow volume of the protein.
Martini is quite robust, and will probably tolerate a couple of overlaps if you let it minimize properly. In any case, keep in mind that equilibrium distances between Martini beads are around 0.5 nm, so that's a good minimum clearance to begin with. Increase it if you can't get a proper minimization afterwards.
If you have proteins with a ring-like topology (channels, for instance) you'll also have to take care of removing the lipids that fall within the hollow volume of the protein.
Please Log in or Create an account to join the conversation.
- zhiyue
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 14
9 years 11 months ago - 9 years 11 months ago #3846
by zhiyue
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Replied by zhiyue on topic Cutoff of deleting lipids around protein
Hi mnmelo,
Thanks for your reply! I appreicate it. Thanks for your further comment on proteins with ring-like topology. That's what I'm dealing with. I've been struggling to decide the right cutoff to compromise no overlapping with keeping enough lipid molecules in the hole. Yes there was no overlapping lipids, but the lipids in the hole was less than expected. Now according to your suggestion, I can reduce the cutoff further to keep more lipids and those overlaps can be removed by minimization. Given this, my question is, what do you mean by "minimize properly"? Is there any specila consideration? Thanks in advance.
Best
Thanks for your reply! I appreicate it. Thanks for your further comment on proteins with ring-like topology. That's what I'm dealing with. I've been struggling to decide the right cutoff to compromise no overlapping with keeping enough lipid molecules in the hole. Yes there was no overlapping lipids, but the lipids in the hole was less than expected. Now according to your suggestion, I can reduce the cutoff further to keep more lipids and those overlaps can be removed by minimization. Given this, my question is, what do you mean by "minimize properly"? Is there any specila consideration? Thanks in advance.
Best
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Last edit: 9 years 11 months ago by zhiyue.
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
9 years 11 months ago #3847
by mnmelo
Replied by mnmelo on topic Cutoff of deleting lipids around protein
By proper minimization I mean up to 5000 steps of steepest descents energy minimization, which should end up with a maximum interparticle force no greater than 10^3.
You can probably use a lower clearance than 0.5nm if you find you want to keep more lipid molecules. As I said, Martini is robust to overlaps and minimization should still run fine.
If your protein is not too bulky you can even give it a try and see whether you can energy-minimize the system without removing any of the lipid molecules. Simple transmembrane helices can sometimes be inserted just like this.
Manel
You can probably use a lower clearance than 0.5nm if you find you want to keep more lipid molecules. As I said, Martini is robust to overlaps and minimization should still run fine.
If your protein is not too bulky you can even give it a try and see whether you can energy-minimize the system without removing any of the lipid molecules. Simple transmembrane helices can sometimes be inserted just like this.
Manel
Please Log in or Create an account to join the conversation.
- zhiyue
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 14
9 years 10 months ago - 9 years 10 months ago #3878
by zhiyue
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Replied by zhiyue on topic Cutoff of deleting lipids around protein
Hi Manel,
Sorry for the late response. Thanks for further suggestion. I've tried to reduce the cutoff to 0.16 nm and minimize the structure as you suggested. It went well. Then I see how Martini is robust to overlaps. But I'm still a little worried about the system I built. With minimum cutoff of 0.5 nm, then the region between chains of my protein is completely devoid of lipids, which is unphysical. But If I reduced the cutoff to, say 0.16 nm, then some lipids are within the equilibrium distance of protein and some lipid is "inserted" into the protein interior, which is unphysicall too. I can visualize the structure and delete those "inserted" ones. But I'm not quite sure about thoese are within 0.5 nm of protein. I'm afraid someone may aruge that it is not right to keep those lipids.
Additionally, even if I kept these "inserted" lipids the structure could still be minimized. I don't understand why this could happen.
Thanks
Best
Sorry for the late response. Thanks for further suggestion. I've tried to reduce the cutoff to 0.16 nm and minimize the structure as you suggested. It went well. Then I see how Martini is robust to overlaps. But I'm still a little worried about the system I built. With minimum cutoff of 0.5 nm, then the region between chains of my protein is completely devoid of lipids, which is unphysical. But If I reduced the cutoff to, say 0.16 nm, then some lipids are within the equilibrium distance of protein and some lipid is "inserted" into the protein interior, which is unphysicall too. I can visualize the structure and delete those "inserted" ones. But I'm not quite sure about thoese are within 0.5 nm of protein. I'm afraid someone may aruge that it is not right to keep those lipids.
Additionally, even if I kept these "inserted" lipids the structure could still be minimized. I don't understand why this could happen.
Thanks
Best
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Last edit: 9 years 10 months ago by zhiyue.
Please Log in or Create an account to join the conversation.
- zhiyue
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 14
9 years 10 months ago #3879
by zhiyue
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Replied by zhiyue on topic Cutoff of deleting lipids around protein
Manel,
Forgot to mention. During the early stage of minimization, mdrun gave warnings like
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Step 237, time 0.237 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 0.004551, max 0.199469 (between atoms 7544 and 7540)
bonds that rotated more than 30 degrees:
atom 1 atom 2 angle previous, current, constraint length
7544 7540 54.6 0.3097 0.3718 0.3100
7548 7544 54.2 0.3141 0.3157 0.3100
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
But as minimization went on, no warnings were reported. What do these warnings imply?
BTW, minimization generated some *.pdb files at some steps indicating changes in coordiantes:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
==> step171b_n1.pdb <==
TITLE initial coordinates
REMARK THIS IS A SIMULATION BOX
CRYST1 180.286 189.321 93.469 90.00 90.00 90.00 P 1 1
ATOM 9108 PO4 POP 1 9.081 58.561 28.358
ATOM 9237 NH3 POP 1 3.915 49.158 27.880
ATOM 9238 PO4 POP 1 8.747 48.300 28.924
ATOM 9575 NH3 POP 1 3.944 57.522 28.985
ATOM 9576 PO4 POP 1 1.119 53.507 28.199
ATOM 9109 GL1 POP 1 8.602 58.099 32.462
ATOM 9110 GL2 POP 1 7.229 55.230 32.862
==> step171c_n1.pdb <==
TITLE coordinates after constraining
REMARK THIS IS A SIMULATION BOX
CRYST1 180.286 189.321 93.469 90.00 90.00 90.00 P 1 1
ATOM 9108 PO4 POP 1 9.080 58.563 28.354
ATOM 9237 NH3 POP 1 3.915 49.159 27.883
ATOM 9238 PO4 POP 1 8.742 48.298 28.923
ATOM 9575 NH3 POP 1 3.945 57.524 28.990
ATOM 9576 PO4 POP 1 1.117 53.507 28.199
ATOM 9109 GL1 POP 1 8.603 58.103 32.459
ATOM 9110 GL2 POP 1 7.228 55.231 32.862
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I noticed the differences are tiny. What does it imply?
Thanks.
Best
Forgot to mention. During the early stage of minimization, mdrun gave warnings like
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Step 237, time 0.237 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 0.004551, max 0.199469 (between atoms 7544 and 7540)
bonds that rotated more than 30 degrees:
atom 1 atom 2 angle previous, current, constraint length
7544 7540 54.6 0.3097 0.3718 0.3100
7548 7544 54.2 0.3141 0.3157 0.3100
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
But as minimization went on, no warnings were reported. What do these warnings imply?
BTW, minimization generated some *.pdb files at some steps indicating changes in coordiantes:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
==> step171b_n1.pdb <==
TITLE initial coordinates
REMARK THIS IS A SIMULATION BOX
CRYST1 180.286 189.321 93.469 90.00 90.00 90.00 P 1 1
ATOM 9108 PO4 POP 1 9.081 58.561 28.358
ATOM 9237 NH3 POP 1 3.915 49.158 27.880
ATOM 9238 PO4 POP 1 8.747 48.300 28.924
ATOM 9575 NH3 POP 1 3.944 57.522 28.985
ATOM 9576 PO4 POP 1 1.119 53.507 28.199
ATOM 9109 GL1 POP 1 8.602 58.099 32.462
ATOM 9110 GL2 POP 1 7.229 55.230 32.862
==> step171c_n1.pdb <==
TITLE coordinates after constraining
REMARK THIS IS A SIMULATION BOX
CRYST1 180.286 189.321 93.469 90.00 90.00 90.00 P 1 1
ATOM 9108 PO4 POP 1 9.080 58.563 28.354
ATOM 9237 NH3 POP 1 3.915 49.159 27.883
ATOM 9238 PO4 POP 1 8.742 48.298 28.923
ATOM 9575 NH3 POP 1 3.945 57.524 28.990
ATOM 9576 PO4 POP 1 1.117 53.507 28.199
ATOM 9109 GL1 POP 1 8.603 58.103 32.459
ATOM 9110 GL2 POP 1 7.228 55.231 32.862
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I noticed the differences are tiny. What does it imply?
Thanks.
Best
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
9 years 10 months ago #3880
by mnmelo
Replied by mnmelo on topic Cutoff of deleting lipids around protein
Hi again,
I'd say that now it is up to you to decide what's relevant. You mention you want to keep a certain number of lipids within the protein. This suggests you have some data on how many there should be, and therefore I don't think it's wrong to tweak clearances until you get that number right. If you then get lipids in places they shouldn't be, it is not wrong to manually remove them (again, you're based on information on how the system should be, and I think this is by far more acceptable than to have lipids sticking through beta sheets or inside alpha-helices).
As to the minimization, yes, Martini really is quite robust. And it was not designed to check the physical validity of your constructs. In any case, my bet is that such systems will quickly become unstable if simulated at CG time steps.
As to the warnings you're getting, this is GROMACS' way of letting you know some constraints rotated a lot in a single step. This is usually ok (and expected when pushing the boundaries of overlap tolerance), especially if they subside before the end of the minimization. If you get these during a production MD run, then I'd worry.
The pdbs you're getting are the coordinates GROMACS had before and after applying the constraint algorithm. You can inspect them to see how the atoms LINCS was complaining about rotate, and possibly identify the cause of such instability. Again, this is not so important for you now because overlaps cause this sort of things before getting sorted out. Worry only if you get these during a production MD run.
Cheers,
Manel
I'd say that now it is up to you to decide what's relevant. You mention you want to keep a certain number of lipids within the protein. This suggests you have some data on how many there should be, and therefore I don't think it's wrong to tweak clearances until you get that number right. If you then get lipids in places they shouldn't be, it is not wrong to manually remove them (again, you're based on information on how the system should be, and I think this is by far more acceptable than to have lipids sticking through beta sheets or inside alpha-helices).
As to the minimization, yes, Martini really is quite robust. And it was not designed to check the physical validity of your constructs. In any case, my bet is that such systems will quickly become unstable if simulated at CG time steps.
As to the warnings you're getting, this is GROMACS' way of letting you know some constraints rotated a lot in a single step. This is usually ok (and expected when pushing the boundaries of overlap tolerance), especially if they subside before the end of the minimization. If you get these during a production MD run, then I'd worry.
The pdbs you're getting are the coordinates GROMACS had before and after applying the constraint algorithm. You can inspect them to see how the atoms LINCS was complaining about rotate, and possibly identify the cause of such instability. Again, this is not so important for you now because overlaps cause this sort of things before getting sorted out. Worry only if you get these during a production MD run.
Cheers,
Manel
Please Log in or Create an account to join the conversation.
- zhiyue
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 14
9 years 10 months ago #3881
by zhiyue
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Replied by zhiyue on topic Cutoff of deleting lipids around protein
Hi Manel,
Thanks for detailed explanation. I'll run a production run to see whether it will work. Thanks again.
Best
Thanks for detailed explanation. I'll run a production run to see whether it will work. Thanks again.
Best
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Please Log in or Create an account to join the conversation.
- zhiyue
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 14
9 years 10 months ago - 9 years 10 months ago #3891
by zhiyue
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Replied by zhiyue on topic Cutoff of deleting lipids around protein
Hi Manel,
I have another pertinent question. If normal cutoff (0.5 nm) is used, then the gap between lipid bilayer and the transmembrane domain of the protein shall be generated. Then water molecules shall occupy these vacuum when adding water layers. This is obvisouly unphysical. So I want to know how I can avoid it if I don't want to reduce the cutoff to keep more lipids (i.e., reduce the space). Thanks in advance.
Best
I have another pertinent question. If normal cutoff (0.5 nm) is used, then the gap between lipid bilayer and the transmembrane domain of the protein shall be generated. Then water molecules shall occupy these vacuum when adding water layers. This is obvisouly unphysical. So I want to know how I can avoid it if I don't want to reduce the cutoff to keep more lipids (i.e., reduce the space). Thanks in advance.
Best
Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Last edit: 9 years 10 months ago by zhiyue.
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
9 years 10 months ago #3896
by mnmelo
Replied by mnmelo on topic Cutoff of deleting lipids around protein
As you say yourself, that is unphysical, so simply remove waters that are embedded in the membrane any deeper than the glycerol region.
It seems to me you are afraid of overmanipulating the setup of your system. That is a legitimate concern, but, as I said earlier, if you have a solid basis to justify your adjustments, that is what you should do. In your example, having a layer of water between transmembranar segments and lipid tails is, indeed, argument enough for removing waters.
Of course, as with any system, proper equilibration prior to analysis is important. In this case you can wait until the area of the bilayer equilibrates, which indicates that at least any gaps have been filled.
Manel
It seems to me you are afraid of overmanipulating the setup of your system. That is a legitimate concern, but, as I said earlier, if you have a solid basis to justify your adjustments, that is what you should do. In your example, having a layer of water between transmembranar segments and lipid tails is, indeed, argument enough for removing waters.
Of course, as with any system, proper equilibration prior to analysis is important. In this case you can wait until the area of the bilayer equilibrates, which indicates that at least any gaps have been filled.
Manel
Please Log in or Create an account to join the conversation.
Time to create page: 0.104 seconds