- Posts: 1
Cis peptide bond problem with backward mapping
- maxime.louet
-
Topic Author
- Offline
- Fresh Boarder
Less
More
5 years 9 months ago #7700
by maxime.louet
Cis peptide bond problem with backward mapping was created by maxime.louet
Hello everyone,
I'm trying to backmap a protein from CG to all-atom. My problem is: I observed quite often (about 2%) of the amino-acids being backmapped with cis peptide bonds (and more especially proline). I know cis/trans isomerizations happen in nature but not at this high rate (as far I know it's far less than 1%). I tried to add dihedral restraints to the all-atom itp file to maintain the omega angles at 180° with no luck. My guess is that this could be due to the position restraints which should be higher in magnitude. When I increase the force for the dihedral restraints the system explodes during the MD phase. I could remove these position restrains but I'd rather prefer to keep them as much as possible. Any idea why there is so much cis isomers of the peptide bond? Is that a common problem for backward mapping in all-atom? or a specific to some proteins? I note that the starting structure (I mean at the very beginning, before mapping to coarse grain and simulations), all peptide bonds were in trans conformation. I'm using the charmm36 force field for the backmapping routine. I would like to note as well that this cis-isomerization could affect different residues for different backmapping runs from the same CG conformation and seems to be quite random, the residue being trans for 1 run and cis for another run.
Any idea would be very much appreciated.
Maxime
I'm trying to backmap a protein from CG to all-atom. My problem is: I observed quite often (about 2%) of the amino-acids being backmapped with cis peptide bonds (and more especially proline). I know cis/trans isomerizations happen in nature but not at this high rate (as far I know it's far less than 1%). I tried to add dihedral restraints to the all-atom itp file to maintain the omega angles at 180° with no luck. My guess is that this could be due to the position restraints which should be higher in magnitude. When I increase the force for the dihedral restraints the system explodes during the MD phase. I could remove these position restrains but I'd rather prefer to keep them as much as possible. Any idea why there is so much cis isomers of the peptide bond? Is that a common problem for backward mapping in all-atom? or a specific to some proteins? I note that the starting structure (I mean at the very beginning, before mapping to coarse grain and simulations), all peptide bonds were in trans conformation. I'm using the charmm36 force field for the backmapping routine. I would like to note as well that this cis-isomerization could affect different residues for different backmapping runs from the same CG conformation and seems to be quite random, the residue being trans for 1 run and cis for another run.
Any idea would be very much appreciated.
Maxime
Please Log in or Create an account to join the conversation.
- cneale
-
- Offline
- Junior Boarder
Less
More
- Posts: 22
3 years 11 months ago #8533
by cneale
Replied by cneale on topic Cis peptide bond problem with backward mapping
Dear Maxime:
I observe this quite frequently as well. The best solution that I have is to add the following section to all amino acids in the .rtp file of the force field that you will use for backmapping (and then remove it later for production AA simulations) -- this is for charmm, but presumably the names are similar in other force fields:
[ dihedrals ]
-CA -C N CA 0.000 50.0 1
-CA -C N HN 180.000 50.0 1
-O -C N CA 180.000 50.0 1
-O -C N HN 0.000 50.0 1
Even with this fix, backmapping still leads to a lot of chirality issues in the protein and even more chirality and cis/trans issues in lipids. At least this might help with your particular issue.
Thank you,
Chris.
I observe this quite frequently as well. The best solution that I have is to add the following section to all amino acids in the .rtp file of the force field that you will use for backmapping (and then remove it later for production AA simulations) -- this is for charmm, but presumably the names are similar in other force fields:
[ dihedrals ]
-CA -C N CA 0.000 50.0 1
-CA -C N HN 180.000 50.0 1
-O -C N CA 180.000 50.0 1
-O -C N HN 0.000 50.0 1
Even with this fix, backmapping still leads to a lot of chirality issues in the protein and even more chirality and cis/trans issues in lipids. At least this might help with your particular issue.
Thank you,
Chris.
Please Log in or Create an account to join the conversation.
- tsjerk
-
- Offline
- Expert Boarder
Less
More
- Posts: 103
3 years 11 months ago - 3 years 11 months ago #8534
by tsjerk
Replied by tsjerk on topic Cis peptide bond problem with backward mapping
Hi Chris,
This should not happen at all with the version on Github. The problem was due to an 'enhancement', and onfortunately someone took that version to distribute with gmx5 compliance. The current version on Github yields correct backbone structures using the algorithm outlined in the paper. If you still see problems with that one, please send me the input structure so I can have a look.
As for lipids (and other molecules), chirality and double bonds indeed require quite a bit of care in the mapping definition. If you run into problems with that, let me know. I may be able to devise a scheme.
This should not happen at all with the version on Github. The problem was due to an 'enhancement', and onfortunately someone took that version to distribute with gmx5 compliance. The current version on Github yields correct backbone structures using the algorithm outlined in the paper. If you still see problems with that one, please send me the input structure so I can have a look.
As for lipids (and other molecules), chirality and double bonds indeed require quite a bit of care in the mapping definition. If you run into problems with that, let me know. I may be able to devise a scheme.
Last edit: 3 years 11 months ago by tsjerk. Reason: - added answer to second issue
Please Log in or Create an account to join the conversation.
- cneale
-
- Offline
- Junior Boarder
Less
More
- Posts: 22
3 years 11 months ago #8535
by cneale
Replied by cneale on topic Cis peptide bond problem with backward mapping
Dear Tsjerk:
I do get this quite frequently using the version from github (presuming I obtained it correctly). The version that I have has initram marked as VERSTAG=20150906-13-TAW.
Unfortunately I am not allowed to upload the input structures, which I realize is a shame.
Perhaps the difference relates to the fact that the CG structures that I am using may have deviated from the initial crystal structure too far. One thing I have recently come to realize is that Martini allows beta strands to flip over (and other less dramatic conformational changes that are not typically observed in AA simulations). Perhaps this is the source if the issues that I am observing.
Thank you,
Chris.
I do get this quite frequently using the version from github (presuming I obtained it correctly). The version that I have has initram marked as VERSTAG=20150906-13-TAW.
Unfortunately I am not allowed to upload the input structures, which I realize is a shame.
Perhaps the difference relates to the fact that the CG structures that I am using may have deviated from the initial crystal structure too far. One thing I have recently come to realize is that Martini allows beta strands to flip over (and other less dramatic conformational changes that are not typically observed in AA simulations). Perhaps this is the source if the issues that I am observing.
Thank you,
Chris.
Please Log in or Create an account to join the conversation.
- cneale
-
- Offline
- Junior Boarder
Less
More
- Posts: 22
3 years 11 months ago #8536
by cneale
Replied by cneale on topic Cis peptide bond problem with backward mapping
Sorry, forgot to mention that I am indeed using gmx 5.1.5. I am not sure why that would cause a problem, but if changing to a newer version of gromacs is expected to fix things then that would be great to know.
The lipid stereochemical issues have all been resolved on our end. I was just letting a previous poster know that they are not resolved out of the box.
The lipid stereochemical issues have all been resolved on our end. I was just letting a previous poster know that they are not resolved out of the box.
Please Log in or Create an account to join the conversation.
- tsjerk
-
- Offline
- Expert Boarder
Less
More
- Posts: 103
3 years 11 months ago #8537
by tsjerk
Replied by tsjerk on topic Cis peptide bond problem with backward mapping
Hi Chris,
Ah, yes, the Martini beta-strand problem. That is a very hard nut to crack, because the distortions are so severe and on top of that, the beta strands in Martini are linear configurations of BB beads, rather than zig-zags, so it is pretty much impossible to get sufficient geometric data for rebuilding the chains correctly. That is something that has to be solved in Martini. If you have a notoriously bad protein case for me, I'd be happy to chew on it and see what I can come up with.
Best,
Tsjerk
Ah, yes, the Martini beta-strand problem. That is a very hard nut to crack, because the distortions are so severe and on top of that, the beta strands in Martini are linear configurations of BB beads, rather than zig-zags, so it is pretty much impossible to get sufficient geometric data for rebuilding the chains correctly. That is something that has to be solved in Martini. If you have a notoriously bad protein case for me, I'd be happy to chew on it and see what I can come up with.
Best,
Tsjerk
Please Log in or Create an account to join the conversation.
Time to create page: 0.106 seconds