determining helical pitch of curved DNA

[ashaytan]

Dear Xiang-Jun,

I would like to ask your advice on I couple of questions.

1) I'm using 3DNA to analyze DNA parameters in nucleosome, both in crystal structures and after MD simulations.
It has been speculated for quite some time, that the DNA in nucleosomes is over twisted (approx. 10 bp per turn instead of 10.5), so I want to show this using 3DNA analysis. An intuitive way to get the value of average helical pitch would be summing the values of twist returned by 3DNA along the DNA and dividing by the number of base pairs. However, I'm not quite sure if this approach is rigorous, and whether one should sum up the values of helical twist of dimer step twist. What do you think?

2) During MD simulations some base pairs might become distorted and they are not found by find_pairs. I can force the calculations on them by manually editing .inp file, however, I assume that the Twist, Roll, Tilt values reported for the base pair steps would be arbitrary.
Will the summation of all twist angles along the DNA helix be meaningful in this case in order to obtain overall twist of the DNA helix?


Thank you in advance,
Alexey

[xiangjun]

Hi Alexey,

Thanks for using 3DNA and posting your questions on the Forum.

It has been speculated for quite some time, that the DNA in nucleosomes is over twisted (approx. 10 bp per turn instead of 10.5), so I want to show this using 3DNA analysis. An intuitive way to get the value of average helical pitch would be summing the values of twist returned by 3DNA along the DNA and dividing by the number of base pairs. However, I'm not quite sure if this approach is rigorous, and whether one should sum up the values of helical twist of dimer step twist. What do you think?

I am aware of the paper "Helical repeat of DNA in solution" by Wang (PNAS 1979) where it was found that "Analysis of the set ... gives the DNA helix repeat in solution as 10.4 base pairs per turn under physiological conditions, with an estimated probable error of +/- 0.1." The "twist" measurement here is an average of covalently closed DNAs. In my understanding, it is related to linker number and writhe, but differs from the dinucleotide-step twist angles from 3DNA. See "Two perspectives on the twist of DNA" by Tobias et al. However, why not pursue your "intuitive way" further to see what you can get. It won't take much time, and the result (whatever it could be) would be informative.

2) During MD simulations some base pairs might become distorted and they are not found by find_pairs. I can force the calculations on them by manually editing .inp file, however, I assume that the Twist, Roll, Tilt values reported for the base pair steps would be arbitrary.
Will the summation of all twist angles along the DNA helix be meaningful in this case in order to obtain overall twist of the DNA helix?

Two points here. Regarding the issues of 'local' vs 'overall' twist, see above. Distorted base-pairs do not have "arbitrary" step parameters: they are derived based on the geometry given in a PDB/mmCIF input file, and can be used to 'rigorously' rebuild the original base pairs. More generally, the default cutoffs in 3DNA are a good starting point. However, there are always boundary cases where the choice is sort of arbitrary. For example, the vertical separation (stagger) is 2.5 Å by default. So a dinucleotide step with a value of 2.49 Å would satisfy this criterion, whereas 2.51 Å won't.

HTH,

Xiang-Jun
 

[ashaytan]

Hi, Xiang-Jun,

thanks very much for the reply and for the helpful hints.
The paper "Two perspectives on the twist of DNA" by Tobias et al. was very useful in understanding the exact definition of twist, writhe and linking number.
In this paper the authors introduced a new property called the "twist of supercoiling", which is different from base-pair step twist, but is connected to the topological invariants, the linking number and the writhing number.
So far I have found only one paper ( http://www.ncbi.nlm.nih.gov/pubmed/21482100 Olson & Zhurkin "Working the kinks out of nucleosomal DNA", Figure 3) which shows this property calculated for real structures, although the authors do not provide and reference to a software package that was used.

I wonder if there are plans to implement such a property for calculation in 3DNA?

Best wishes,
Alexey

[xiangjun]

I wonder if there are plans to implement such a property for calculation in 3DNA?

In principle, I'd be interested in implementing such a property in 3DNA. To that end, I need direct access to details: either a worked example, or a working program with sample input and output files. Pure text descriptions or mathematical formula as often seen in scientific publications help for general understanding, but they are certainly not enough from a programatic point of view.

Xiang-Jun