This issue was moved to a discussion.
You can continue the conversation there. Go to discussion →
【How to make a block copolymer with nonstandard PGLeu monomers】 #190
Comments
|
@kurokawaikki we are always happy to help with it. I try my best walking you through this. First you need to choose Martini2 or Martini3 as the force-field. Please note that the current PEG martini3 does not work well in combination with membranes, so using Martini2 might be better. However, the LEU is pretty bad in Martini2. We are currently working on improving PEG for Martini3. I'd expect and updated version to be released at most in 4-5 weeks from now. As soon as there is some working option for Martini3 PEG I can share it with you, if you like. The rest of the answer is written assuming you want to use Martini3. Please also note that you should verify the properties of PGLeu independently from the PEG either against atomistic simulation data or experimental data. But don't worry about that now first we obtain some parameters. I understand you already have an all-atom simulation of your polymer or at least a fragment of it? If that is the case, what you need to do is derive the bonded parameters for the missing monomer. In particular you need the bond and an angle between the main backbone and the side chain as well as bonds and angles between PEG and the modified PEG bead. The values you can obtain by mapping the trajectory essentially following this tutorial. It might also be useful for you to read through his issue, which deals with a similar question. To generate a preliminary itp file you need a .ff input file for the new monomer. Let's give it the residue name 'PGLeu'. To clarify I would treat the PG-LEU unit as one monomer i.e. 3 beads following the Martini3 mapping. The syntax of the .ff file is documented to some extent here. For you case I've written a template file as shown below. Note the places with '<some_text>' you need to replace with the parameters obtained from the mapping. To generate the itp file save this file as PGLeu.ff and the use it with polyply as follows: Now all you need to do is fill in the missing parameters based on the atomistic simulations. I would also recommend adjusting the bead-types of the PG fragment as well as the LEU fragment. However, once you have obtained the bonded parameters we can take that step. Simply reply to this issue. Please try of you can generate the itp and let me know if you have more questions. |
|
Thank you very much for your guidance. Thank you again for your great help. |
|
I have now progressed to the step of parameterization of PEGLeu monomer. My question is how can I map my molecule to CG model? In my PEGLeu case, if I choose 4-1 mapping, I will need three beads. However, I have no idea on how to choose the bead type from the building block. Most important of all, what is the relation between the CG model that I fit and the beads that you chose here? [ atoms ] I am completely new to CG model and gromacs and I am sorry for lots of questions. |
|
Hi, Let me try to answer your questions: The best way I would say is by similarity to what already exists. For PGLeu I've checked how many beads the Leu part of the amino-acid has and then simply added one bead for the backbone as that is most similar to PEG. We also have some more defined mapping rules in the SI of the Martini3 paper (see page 36) I would follow the following 3 guiding principles:
In principle, you should always validate your bead assignment by comparing to experimental and/or simulation data of either the polymer or the fragment. For example, if you have an atomistic simulation of your polymer, you can compute the radius of gyration from that. Then try different beads types for the backbone to see which matches the RG data best. I don't really understand what you mean by this questions. |
|
Thank you very much!! I now have some sense on how to find the appropriate bead type. For example, I cannot find the SN2r in the building block table. I now understand that SN2r -> PEG, [P2 and SC2] -> Leu Because I am now in the mapping stage in the parametrization, I try to fully understand the meaning that you write. Thank you very much for your great help! I can try to proceed to next step. After I finish this, if you needed, maybe I can help to write a small tutorial for beginners like me. |
|
@kurokawaikki any help with making tutorials for beginners is greatly appreciated. Hope the parametrization goes well! |
|
@fgrunewald Thank you very much! I encounter some problem on how to get the right parameter of my polymer. I followed the tutorial that you gave me. I choose to use Bartender xtb way to find the initial parameters, because I do not know how to guess the initial value on my own. The following figures are the bond and angle distributions. Red is the CG model that Bartender gave me and the blue line is the atomistic simulation data parameter with ATB. The bond distribution: The angle distribution: I have tried many times to change the value generated by Bartender to match the average bond length and angle that were calculated from atomistic simulation data. |
|
To be clearer, here is my CG itp file generated with Bartender. `; Topology by Bartender - www.github.com/rmera/bartender [bonds] I modified the Leu part. Originally, Leu was represented with 2 beads. I changed to 3 beads and take on of the beads contains amine to be positively charged. Could you give me some advice on how to derive correct parameters especially on the force constant? Thank you very much. |
|
@kurokawaikki I have too look at the data and need to ask some questions about bartender to some experts in our group. I'm not very familiar with the software. Then I get back to you. Overall it, however, doesn't look bad. The discrepancy in bond length can be related to the fact that bartender doesn't sample the atomistic dihedral conformations. |
|
@kurokawaikki let's start with a small checklist:
Secondly regarding the LEU, you have to be very careful remapping these residues. Typically it is better to change the bead-type when there is a charge present. Like the C/N terminals of proteins have the same number of beads, but just a different bead type. |
-> Yes I have selected the H atom when I building the model in the CG builder.
-> Yes, I mapped it as CH2-O-CH2. But the last one at the tail is CH2-OH
-> There is no log output and no warning. But I noticed that some of the xtb data and fitted function (generated by Bartender) did not matched in some of the output figures. For example, the first one is good fitted function and the second one is not.
Because I did not add any information on dihedrals (my polymer contains single bond in back bone so I think it is not necessary), will this be the problem?
I change the Leu to three beads... So I have to change back to two beads and change the bead type of beads containing C/N (positively charged). Which bead type do you suggest? Additionally, could you tell me in general way how do you derive the force constant without Bartender? In the tutorial, it only stated that you should try and error to fit the parameter to atomistic simulation results. I understand that I can fit the bond angel and bond length and even dihedrals, but I cannot find any tutorial on how to derive force constant. Could you also give me some advice on that? Thank you very much! |
|
@kurokawaikki I understand. My opinion is that for a flexible fragment like polymer bartender's sampling of the degrees of freedom is not good enough. That's why you see a difference in bond length and angles. On top of that come the strange fits that you show. Probably using a classical approach of running an all-atom simulation, mapping it and than matching the distributions is better. At least for the PEG part you can take the parameters from the library. Having said that I think your biggest problem is to guess force-constants. The way it is usually done is the following:
As a first guess: Bond force constants typically are between 2000-7000 anything about 9000 should probably be a constraint. Angles typically range from 10-100 in terms of force-constant. There are other tools like pycgmap or fast_forward that can do the mapping and extract distributions. In principle SwarmCG can automatically derive bond and angle force-constants using a machine learning approach. However, SwarmCG can be resource intensive and is perhaps not the most easy to use. But it could work in your case. |
|
@ricalessandri this issue I mentioned |
|
Dear Dr. @fgrunewald,
I have some question in the following:
I think after I got the right parameter, I can use it to build my polymer with polyply. Finally, as for the tutorial for the beginners on Martini CG, I can write my experience on this topic and explain along the steps. Thank you very much again. |
|
Hi @kurokawaikki, I think the SwarmCG distributions look good. Very good job on that! To answer your questions:
When you're happy with the parameters you can use polyply to make the itp files and coordinates for any length of polymer. In case you have any questions on how to write the .ff input files let me know in this issue. If you want we can add the parameters to the library down the road as well. I also would be super happy about a tutorial. Perhaps you can open a discussion thread about the tutorial here. Then we keep it separated from this particular case. You can use insane or TS2CG to make the membrane. I would use TS2CG in case you want to make a vesicle. If you have questions about making membranes/vesicles I'm happy to help. However, I would ask you to raise a new issue here and tag me in it. Then either I or someone else will help you. Let me know if it all works out in the end either way. I'm curious now. Cheers, |
|
Dear Dr. @fgrunewald , I have done a parameterization on both model. I increase the model a little bit to include PG > PG >> PG parameters. The parameter of first one is: The goodness of fitting is shown in the figure: I have tried different combination of force constant, length, angle, but I can not get further improvement. The parameter for second model is: The goodness of fitting is shown here: I think the first one is a little bit better. Additionally, do we have the list for all bead types corresponding to chemical structures? Am I missing something in the papers, because I did not find the whole list in martini 3 paper and SI? Finally, here is the ff file that I wrote. It seems working... I read the documentation for FF. However, the syntax is different from the sample that you provided. Is there some place I can find more documentation? Thank you very much for your great help. The final initial.itp is: |
|
Hi @kurokawaikki, Your SwarmCG results look good to me for either mapping and I agree the first one is a little better. We have some more detailed tables internally related to currently unpublished molecules mostly. It is anyways best to verify bead-type assignments against some form of experimental data. I think only specifying the PG PG BB angle is fine. The FF syntax has a very large range with many specialized short-cuts. The documentation provided at the moment gives a conservative description that will always work. However, with the upcoming documentation of martinize2/vermouth more information on the FF format will become available. If you have any specific questions feel free to ask here. But I take your model is working for now. A last remark perhaps. Be aware that the Membrane Affinity of the current PEG model is overestimated. We have developed a preliminary update to the Martini interaction table that is currently validated. For more details on that you can write me an email. Unfortunately I cannot disclose this publicly yet. Cheers, |
|
Dear Dr. @fgrunewald, Thank you very much. I thought my model is working. So, I will go to the next step on making a membrane system and the polymeric micelles to test the endosomal escape properties. If I have problem on that, can I ask you on the different github issues? In addition, I will start to writing the tutorial. Thank you very much again! Hopefully, it will work at the end~ |
|
@kurokawaikki Happy to help with this! I hope it works for you. If you have any problems feel free to raise another issue. I'm closing this one for now. Also thanks for putting in the tutorial draft. If you want to expand it, I think the outline is quite complete already. |
This issue was moved to a discussion.
You can continue the conversation there. Go to discussion →
Dear developers,
I am new to the CG model simulation. I was using all atom MD simulation with Desmond. Currently, I would like to study the process of endosome escape, but all atom simulation is too slow and fail to serve the purpose.
![]()
https://i.ibb.co/nkQRSNL/Untitled.png
My model basically contains 2 parts. The first one is the endosome lipid bilayer (I obtained from this reference: A molecular view on the escape of lipoplexed DNA from the endosome). The second part is my polymers. The polymer contains 2 parts: one is PEG with leucine in side chain and the other one is PEG tail (It is shown in the following figure).
I would like to know how to parametrize it and generate a topology file that can be used in the martini force field.
Because polyply is powerful in making the CG model, I would like to build my polymer model with it. However, my block copolymer contains nonstandard monomer. How could I obtain and parameterize the monomer of PEG with Leu in side chain?
Finally, is there
a way to import the nonstandard monomer model into polyply and use it and PEG monomer in the polyply to make a PEG-PGLeu copolymer with polyply?
Could you give me some guides and hints on how to walk through the process?
Thank you very much for your great help and your time.
The text was updated successfully, but these errors were encountered: