# Generating PDB file for custom DNA sequence

I asked this question in biology.stack exchange and was referred to this community. I am going to request migration for this question. https://biology.stackexchange.com/q/105024/67415

I want a PDB file of DNA sequence I am interested in (10 base pairs) for docking. I will appreciate if any of you can please clarify some of my doubts.

Do you think, I can generate PDB file by using this link http://www.scfbio-iitd.res.in/software/drugdesign/bdna.jsp

How can I get the pdb file of the DNA sequence I want from protein data bank, I did search it but did not find the one I want. Is it a way to extract the PDB file?

What are other methods I can use to generate PDB file of B-DNA sequence?

Please delete if it is against the group for posting again and redirect me to where I can request migration of this question.

I normally use PyMOL. In PyMOL you can generate a model of a sequence of DNA with the command fnab — a sister command to the better known fab, which makes a protein sequence of a given secondary structure. I've not encountered that site, but it says it uses experimental data, which is a better approach.

## Side note 1

OP is asking for DNA, but if anyone stumbles across this, do note that RNA is a totally different ballgame and requires more advanced solutions (example).

## Side note 2

I should say docking a known sequence is fine, but not docking to guess the sequence is not.

For docking, if you have a zinc finger, helix-loop-helix, leucine zipper, MADS box or similar domain which binds DNA in a consistent manner then it is straightforward. However, predicting the bound sequence is far from trivial or accurate. For example, in MEF2C the N-terminal amine there are three reader-head dictating specificity and one is a floppy terminus, which most forcefield-based methods would struggle to get right.

## Testing

So let's say we want to test PyMOL's fnab in the GUI. We could pick a protein, MEF2A example from above:

fetch 1EGW, type=pdb1
print cmd.get_fastastr('chain E') # to get the sequence
fnab AAAGCTATTATTAGCTT
align obj01 and chain A, 1EGW and chain E
show sticks, chain E+F
hide cartoon
remove solvent


The RMSD is 1.6 Å which is not great, but pretty decent considering that it is with a generated sequence that does not take into account backbone distortion from the bound protein.

• Thank you for your reply. I want to say that I know my sequence and the docking position. However, the big doubt in my mind is that if the pdb file generated through say pymol/avogadro/chimera is similar to the PDB file from the protein data bank? I want to learn how different it is if any. Oct 26 '21 at 15:19
• Added a quick (nearly) coding-free test. Oct 26 '21 at 15:48
• With probably the same results as PyMOL method there is x3dna webserver, creating an idealized DNA double-helix of customizable sequence: web.x3dna.org/fibermodel/regseq Oct 29 '21 at 16:57
• PDB files from the PDB or generated are essentially the same in term of organization, though you would have a more complicated header and different B-factors for each atoms when the file comes from the data bank. Be careful not to mistake mmCIF format with PDB one. As for a difference in the structures Matteo answer covers it. Oct 29 '21 at 17:18

There is a tool for making DNA segments of any curvature. It exports as a PDB, and the people behind the tool are quite committed and helpful.

https://www.lifeexplorer.info/

Here is a curved DNA segment I generated using their tool; they can make much more dramatic bends.

I regularly use the site you cited (scfbio...) to generate idealized straight DNA segments for illustration purposes. It is not sequence-dependent. For an actual solved structure you can extract specific chains from existing structures containing DNA (e.g. RNA polymerase II structures by the Cramer group). These structures will have their own sequence, not your specific one, and will be distorted by protein-DNA interactions. Other than that, there likely won't be a PDB structure of your specific sequence. And even if there was, the conditions likely would differ from the ones your structure was acquired at. I may be entirely off, but it sounds like you should be fitting individual bases, sugars and phosphates, not continuous poly-d-nucleotides.

Shown below is the pymol visualization of a 100bp dsDNA input for B-DNA from the resource you cited It does not take into account the DNA flexibility (persistence length etc)