How to Analysis the Conformation Parameters of Unnatural base in MD

[shengxiehuang]

Dear Mr Lu,

Recently, we performed 200 ns of molecular dynamics (MD) simulations on some DNA sequences containing unnatural nucleobase pairs using Amber 14, a total of 100000 frames. Now, we try to illustrate the dynamicvariation of DNA structure in the whole process by analyzing DNA structural parameters (such as slide, roll, twist,etc) of all frames.

Amber has a built-command analyzing some of unnatural bases such as 8OG, <nastruct naout nastruct.dat resrange 274-305 resmap 8OG:G>, but it comes to error “N0 N1” when I use this command.

Now my question is that if 3DNA can analyze DNA parameters of all frames by setting some parameters. If possible, please recommend somereading material. Thank you very much!

Best wishes!
Xiehuang Sheng

[xiangjun]

Hi Dr. Sheng,

Thanks for your interest in using 3DNA to analyze the trajectories of your MD simulations and posting your question on the Forum. In principle, 3DNA should be able to handle modified bases, and to process MD simulation results. To make our discussion concrete, however, please attach a sample PDB data file to illustrate the problem you have. It could be a frame of your Amber simulations, or some variants of it if you are sensitive to the structure.

Best regards,

Xiang-Jun
 

[shengxiehuang]

Dear Mr Lu
       Thank you for your kindly reply.
       The unnatural base-pair  (named DX, DY) are not typical purine or pyrimidine.  The structure of DX， DY and a frame of DNA structure are in the attachment (have stripped out water and Na+).   How can I analysis the conformation parameters with 3DNA in a such case. thank you very much!

BEST wishes!
Sheng Xiehuang

[xiangjun]

Hi Dr. Sheng,

Thanks for your follow-up. Indeed, the three attached PDB files helped to identify the problem. As shown below, bases DX and DY are unusual, especially DY which has C1 instead N1 for the glycosidic bond with C1'.

-- DX

-- DY

3DNA and DSSR have no problem in identifying these two bases, matching them to 'g' and 'c' (for modified G and C) respectively. However, DY poses a problem to 3DNA up to v2.2 for the missing N1 atom (DSSR works as expected). When I tried 3DNA out with your file DNA.pdb, 'find_pair' crashed  with the following message:

Code: [Select]

handling file <DNA.pdb>
Match ' DX' to 'g' for residue  DX    9  on chain   [#9]
    check it & consider to add line ' DX     g' to file <baselist.dat>
Match ' DY' to 'c' for residue  DY   22  on chain   [#22]
    check it & consider to add line ' DY     c' to file <baselist.dat>
missing ' C5 ' atom : residue name ' DX', chain  , number [   9 ]
missing ' N7 ' atom : residue name ' DX', chain  , number [   9 ]
missing ' N1 ' atom : residue name ' DY', chain  , number [  22 ]
Segmentation fault (core dumped)
Your DNA.pdb file has helped me to fix the bug in 3DNA, and further refined DSSR. Please download 3DNA v2.3-2015oct19 and DSSR v1.4.2-2015oct19. The output file DNA.out and DNA-dssr.out are attached for your reference. The running commands are:

Code: [Select]

find_pair DNA.pdb | analyze  # DNA.out
x3dna-dssr -i=DNA.pdb -o=DNA-dssr.out
Have a try and report back if you have further problems.

Xiang-Jun

[shengxiehuang]

Dear Mr Lu
     Thank you for your elaborate explanatin. I will try it in following days. By the way, whether should I study both sortwares, 3DNA and DSSr. thank you very much!
      Best wishes!

Sheng Xiehuang

[xiangjun]

By the way, whether should I study both sortwares, 3DNA and DSSr.

It depends. You can use either or both. 3DNA v2.x has a Ruby script x3dna_ensemble to analyze MODEL/ENDMDL delineated ensembles, or you could try do_x3dna (http://rjdkmr.github.io/do_x3dna/). DSSR is a new program tailored for RNA structures (see Section DSSR-NAR paper on the Forum), even though it works just fine for DNA. Moreover, DSSR provides options --nmr and --json for the analysis of trajectories of MD simulations, as long as the you can convent the output to the standard MODEL/ENDMDL PDB format.

DSSR is run as x3dna-dssr, using the same underlying algorithms for base-pair parameters. DSSR (along with SNAP) has a completely new codebase, as part of would be 3DNA version 3. At that time, the key components in the widely used 3DNA v2.x (including analyze, rebuild, fiber etc) will be distilled into v3 as new programs. 3DNA v2.x will still be maintained and supported, for bug fixed etc, but no more new features.

I am not a MD practitioner, but I am keen to see how DSSR can be effectively used in this increasingly important area. Whatever your choice would be, please keep us informed of your progress.

Xiang-Jun

[shengxiehuang]

Dear Mr Lu
       I really appreciate your help in modifying the software script for my case. I can do it by myself, Now.  The procedure as following,
    a, releasing all snapshots from the trajectories of Amber MD simulations
    b, building pbfile.dat with pair_find
    c, using x3dna_ensemble to analyze all the snapshots.        x3dna_ensemble analyze -b pbfile.dat -p 'XX/XX_*.pdb' -o ensemble.out
    d, extracting all the parametes from the result.       x3dna_ensemble extract -f ensemble.out -a
    By the way, I will study the protein-DNA interaction in the next project. Which software should I use, thank you.

Best Wishes!
Xiehuang Sheng

[xiangjun]

Thanks for posting the detailed procedure of your MD analysis. As for the software for analyzing DNA-protein interactions, again, it depends. The current method may work there as well.

Good luck with you research!

Xiang-Jun

[shengxiehuang]

Dear Mr Lu,
   
     We can use x3dna_ensemble to analyze the MD trajectories. But we are faced with a new problem.  The MD analysis is time-consuming for large system which contains 100,000 snapshot. it will cost our over 24 hours. I wonder to know wheher x3dna_ensemble can work in parallel, such as x3dna_ensemble.MPI. Thank you very much!

Best wishes!

Xiehuang Sheng

[xiangjun]

No, the Ruby script 'x3dna_ensemble' does not work in parallel per se. You can have a look of the source code to see what's going on.

For your purpose, would it be possible that you first select a reasonable subset of the 100,000 snapshots to make sure the results make sense? If things work as expected, you can thn analyze the whole set. The overall speed would of course depend on your hardware and the molecular system. Is 24 or even 48 hours of computer time really that significant?

Alternatively, with sufficient programing skills, you could divide your snapshots into smaller chunks (say 100 pieces). You can then run 'x3dna_ensemble analyze' on each in parallel. In the end, you need to combine the result from each run together. This is just a general idea for you to think about and play with.

HTH,

Xiang-Jun

[shengxiehuang]

Alternatively, with sufficient programing skills, you could divide your snapshots into smaller chunks (say 100 pieces). You can then run 'x3dna_ensemble analyze' on each in parallel. In the end, you need to combine the result from each run together. This is just a general idea for you to think about and play with.

Dear Mr Lu,

It's a good idea. thank you very much!

Best wishes!

Xiehuang Sheng