I never used Charmm. It might be due to PBC problem.

However, try trjconv command to remove "pbc" issues 

use flag -pbc nojump

-pbc mol

or -pbc cluster

Hi,

I suggest you use the AMBER force field instead of CHARMM, using Gromacs itself. Apart from that, I feel this information is little for me to provide a reason for the distortion of your double helix. But, maybe the articles I've mentioned below will help you in some way. Your question has made me curious too. If possible, do let us know when you solve the problem.

http://www.wesleyan.edu/molbiophys/graduatestudents/BiophysJ732313A5NanosecMolDynamTraj.pdf

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4139815/

As Navneet told use pbc then visualise the trajectory. Still if you are finding this type of problem, then there most be some problem with the simulation step. As Dheeraj mentioned Amber ff is good for DNA, it's somewhat true. You have to use proper restrain during equlibration also. 

Follow this paper for your system preparation. I hope it'll be useful.

http://pubs.acs.org/doi/abs/10.1021/acs.jpcb.5b06229?journalCode=jpcbfk

Hi, 

Fluctuation of terminal DNA bases is, to some extent, normal. How many DNA bases are covered with protein? 12 bases are extremely necessary to use, you can add 2-3 bases extra (from 12-mer to 15-mer) and during analysis, ignore those extra bases. It will keep your DNA somewhat intact.

In one article, the authors restrained the terminal DNA bases to mimic the whole DNA (or longer DNA), you can do this as well (sorry I can not recall the paper).

Regards,

Thanks All.

Navneet: It's not a pbc problem. 

Dheeraj: I have done some other only protein simulations with Charmm36 and so I wanted to stick with charmm36. I will go through the articles you recommended. I am also thinking of using Amber for a check. What other information do you need to understand my problem?

Krishnendu: I will go through that paper.

Md. Ayaj: I have thought of using 2 residues longer sequence at both the terminal. However, I am worried about the outcome of that. Please share with me any such papers if you know any. I will try restraining the terminal residues.

Hi,

I have used Charmm36 for protein simulations too. I haven't dealt with DNA simulations much. But, AMBER-ff is believed to be very suitable for DNA. I am not sure what I am missing or what more information I need as of now. Perhaps, you can try simulating your complex with AMBER once and let us know if everything works fine. Also, I agree with Muhammed. You can try adding extra bases at the termini and allow them to fluctuate as much as they want (as long as the necessary bases are intact). You may reduce the simulation time (ns) to save time on this trial and error process.

A little late to the party, but my $0.02: poly(A) and poly(AT) sequences are very hard to model. You have a very complex case in which a DNA sequence that will sample B' conformations is now bound to a protein; DNA-binding proteins intrinsically destabilize their cognate DNA sequences to some extent. A proper control to determine suitability of the force field is a simple simulation of the DNA alone, and compare it to available experimental evidence like SAXS. I know data exist for d(AT)10 in solution. AMBER does not fare better than CHARMM for these unique sequences.

Hi, Justin,

I have run control setups for DNA with both Charmm36 and Amber99bsc1, with parmbsc1 correction. I found that Amber99bsc1 gives significantly good results (rmsd, roll, twist etc.). However, Amber99bsc1 gives significantly bad results for protein control systems compared to Charmm36. I have also tried Amber99SB, which give good results for protein, comparable to Charmm36, but bad for DNA. I am thinking of using forcefield parameters from Amber99bsc1 for DNA and forcefield parameters of Amber99SB for protein. Will it be reasonable? If yes, can you suggest me any reference on how to do it for GROMACS. 

Don't use .gro files for the reference (-s) passed to trjconv. Visualize the system and make sure PBC effects are properly accounted for. Your order of operations is also not right, see http://www.gromacs.org/Documentation/Terminology/Periodic_Boundary_Conditions#Suggested_trjconv_workflow

parmbsc1 is an improved DNA parameters but not for protein. You can either use amber14sb or amber99sb. But I cannot see a good reason why these versions of amber will perform bad/differently to charmm36. Terminal base pairs are always flexible. I would suggest extending the length of the DNA by 3-5 base pairs to mitigate there effect on core DNA behavior. Can you post a vmd snapshot of the final structure?

before starting with trjconv, first make sure that your reference structure (must be tpr as Justin has suggested, tpr is required for -pbc mol) is also not broken (pbc fixed). I recommend using tpr used for energy minimization. Dump the gro/pdb to ensure that the protein-dna complex is whole. A possible work flow...

gmx trjcat prun.xtc -n index.ndx -o protein-dna.xtc (removing water first assuming water is not of interest)

gmx trjconv -s protein-dna.xtc -pbc mol -o nojump-protein-dna.xtc -s em-protein-dna.tpr

or

gmx trjconv -s protein-dna.xtc -pbc nojump -o nojump-protein-dna.xtc -s em-protein-dna.tpr

---that can fix most issues, if not then go with this

step1: gmx trjconv -s protein-dna.xtc -pbc whole -o whole-protein-dna.xtc -s em-protein-dna.tpr

step2: gmx trjconv -s whole-protein-dna.xtc -pbc cluster -o cluster-protein-dna.xtc -s em-protein-dna.tpr

output of step2 should fix even the most crazy pbc issues....

##### RMSD calculation for dna-peptide

better calculate the rmsd for protein and dna separately. Dump an index with protein backbone and dna-backbone and then use gmx rms