Can 3DNA DSSR handle Left-handed DNA?

[GengshiWu]

Dear Xiang-Jun,

Our lab is currently studying left-handed DNA (L-DNA). I’d like to clarify that by L-DNA we mean the mirror image of standard B-DNA, rather than the L-DNA fiber model included in the 3DNA DSSR library.

Specifically, our structure (b40-rb10.5_out_minimized_aligned2Z_i_x.pdb) was generated by applying a mirror transformation to a standard B-DNA model (b40-rb10.5_out_minimized_aligned2Z.pdb) that we constructed using the DSSR fiber B-DNA model but with modified parameters for 10.5 bp/turn.

When we analyzed this mirror-image L-DNA with x3dna-dssr analyze, the program successfully identified the helix start and end points (as shown in the attached image), but the helix form was not recognized as L-DNA.

My questions are:

1. Is my understanding correct that DSSR can identify the geometry but does not recognize the form of our L-DNA?

2. Is there any way to make DSSR recognize the form of our L-DNA?

Thank you for your time and and for updating this excellent tool!
I also attached the two files mentioned above

Gengshi Wu

[xiangjun]

Hi Gengshi,

Thank you for sharing details about your L-DNA structure (mirror image of standard B-DNA). It seems to represent a novel form, and I plan to investigate it further. If feasible, I will update DSSR to accommodate this new form.

Best regards,

Xiang-Jun

[xiangjun]

Hi Gengshi,

I've checked the two PDB files you uploaded: a standard B-DNA model (ending with aligned2Z.pdb), and its mirror-image (L-DNA), with x-coordinates negated (ending with aligned2Z_i_x.pdb). For clarity, I've extracted the first 6 atoms from each file, so readers easily see the differences.

# a standard B-DNA model          x
ATOM      1  P    DA A   1       0.621   9.421  -1.028  1.00  0.00           P
ATOM      2  O1P  DA A   1       0.606  10.783  -1.583  1.00  0.00           O
ATOM      3  O2P  DA A   1       1.439   9.186   0.166  1.00  0.00           O
ATOM      4  O5'  DA A   1      -0.863   8.945  -0.676  1.00  0.00           O
ATOM      5  C5'  DA A   1      -1.693   8.405  -1.689  1.00  0.00           C
ATOM      6  C4'  DA A   1      -2.528   7.255  -1.154  1.00  0.00           C

# a mirror-image, with x-coordinates negated
ATOM      1  P    DA A   1      -0.621   9.421  -1.028  1.00  0.00           P
ATOM      2  O1P  DA A   1      -0.606  10.783  -1.583  1.00  0.00           O
ATOM      3  O2P  DA A   1      -1.439   9.186   0.166  1.00  0.00           O
ATOM      4  O5'  DA A   1       0.863   8.945  -0.676  1.00  0.00           O
ATOM      5  C5'  DA A   1       1.693   8.405  -1.689  1.00  0.00           C
ATOM      6  C4'  DA A   1       2.528   7.255  -1.154  1.00  0.00           C

Mathematically, this is a simple operation, but you have created a hypothetical DNA type that is unknown to DSSR. Nevertheless, DSSR is functioning as intended:

• For the standard B-DNA model, DSSR classifies the dinucleotide step as 'B' correctly.
• For the L-DNA, DSSR cannot classify the dinucleotide steps. and designates them as '.' (see the section below). Also as noted in the header section, DSSR only recognizes the common A-, B-, and Z-form helices.

  Note: a helix is defined by base-stacking interactions, regardless of bp
        type and backbone connectivity, and may contain more than one stem.
      helix#number[stems-contained] bps=number-of-base-pairs in the helix
      bp-type: '|' for a canonical WC/wobble pair, '.' otherwise
      helix-form: classification of a dinucleotide step comprising the bp
        above the given designation and the bp that follows it. Types
        include 'A', 'B' or 'Z' for the common A-, B- and Z-form helices,
        '.' for an unclassified step, and 'x' for a step without a
        continuous backbone.
      --------------------------------------------------------------------
  helix#1[1] bps=40
      strand-1 5'-AAAAAAAAAATTTTTTTTTTCCCCCCCCCCGGGGGGGGGG-3'
       bp-type    ||||||||||||||||||||||||||||||||||||||||
      strand-2 3'-TTTTTTTTTTAAAAAAAAAAGGGGGGGGGGCCCCCCCCCC-5'
      helix-form  .......................................

1. Is my understanding correct that DSSR can identify the geometry but does not recognize the form of our L-DNA?

Yes, that is correct. DSSR can see this is a left-handed helix, but it (currently) does not recognize your L-DNA form.

Along the line, one could imagine another form where y-coordinates are flipped. What would it be called? How would DSSR suppose to classify it?

2. Is there any way to make DSSR recognize the form of our L-DNA?

In principle, yes. One just needs to incorporate the domain-specific knowledge into the DSSR code. I need to know more about your L-DNA before considering how to do that.

Best regards,

Xiang-Jun

[GengshiWu]

Dear Xiang-Jun,

One thing to clarify: My previous naming of L-DNA and B-DNA was confusing. Although L-DNA forms a left-handed B-type helix, the “L” in L-DNA refers to the sugar chirality (L-2-deoxyribose), not helix handedness. By contrast, the DSSR fiber B-DNA uses D-2-deoxyribose. So, it is also called D-DNA.

With this being said, L-DNA is just the enantiomer (non-superimposable mirror image) of D-DNA. The mirror reflection changed the three chiral carbons of deoxyribose (C1′, C3′, C4′) from R–S–R (D-DNA) to S–R–S  (L-DNA).

Thank you for your time and dedication to DSSR!
Best,
Gengshi Wu

[xiangjun]

Hi Gengshi,

Thanks for your clarification regarding the naming of L-DNA, referring to the L-2-deoxyribose. It is a nice coincidence it is also left-handed.

Your L-DNA actually has prompted me to think more about this new (hypothetical) form of DNA, which is left-handed but with a B-type helix, by simply reversing x-coordinates of the classic B-DNA model. I can now have a clear mental image of this L-DNA: left-handed, with base-pairs flipped along the long axis. These are the two key features of Z-DNA, which characterizes a zig-zag backbone of CG repeats. The L-DNA, however, has a smooth backbone, as in B-DNA.

With this knowledge, it is not hard to classify L-DNA automatically with DSSR. I may consider to add the --L-DNA option specifically for this purpose, not to complicate the default DSSR output. It is after all not a common form, as A-, B-, and Z-form DNA. I would to see more examples of such L-DNA.

It is questions like yours that make DSSR more relevant and useful. I always appreciate users' feedback and encourage them to ask questions, sharing their experiences (both good and bad), freely and openly on the 3DNA Forum. Even in the age of AI, I still believe that there is no replacement for the human touch. I watch the 3DNA Forum closely and am quick to respond to users’ questions and concerns.

Best regards,

Xiang-Jun

[GengshiWu]

Xiang-Jun,

Thank you for considering an adding the L-DNA option in DSSR. And I think your summary captures exactly what I mean by L-DNA: a left-handed helix with B-like geometry and a smooth backbone, distinct from Z-DNA. At this moment, I don’t have additional L-DNA examples. But I’ll share more on the forum with the community as I'll need to generate and study them.

Best,
Gengshi Wu

[Di_Liu]

Hi Xiang-Jun,

L-DNA is totally synthetic from the mirror-image sugar, but now is very widely used. Here is a brief intro from Glen: https://www.glenresearch.com/reports/gr33-21
Let me further clarify: reversing the x-coordinates is actually performing the reflection about the yz plane, which is a special case. More generally, L-DNA can be generated by performing the reflection about any plane.
Thanks for considering to handle L-DNA with DSSR! L-sugar could be readily determined from the chiral carbon atoms on the sugar.

Di

[GengshiWu]

Thank you for the clarification!

[Di_Liu]

By the way, here is a pdb structure containing both L-DNA and L-RNA: https://www.rcsb.org/structure/4WB3

[xiangjun]

Hi Di and Gengshi,

Thanks for the additional information. They are exactly the kind of information I was looking for.

• Glen Report 33-21: L-DNA: Applications and the Recognition of Mirror Image Nucleic Acids
• Structural basis for the targeting of complement anaphylatoxin C5a using a mixed L-RNA/L-DNA aptamer (PDB entry 4WB3)

I'll study them carefully, and try to implement an option in DSSR to handle L-DNA atuomatically. I'll keep you updated.

Best regards,

Xiang-Jun

[xiangjun]

Hi Di and Gengshi,

Thanks for bringing L-DNA to my attention. I've updated DSSR to v2.7.0-2025dec09, which automatically recognizes L-DNA/RNA steps. Using your model PDB (`b40-rb10.5_out_minimized_aligned2Z_i_x.pdb`) as an example, the new DSSR output is as follows:

Code: [Select]

  helix#1[1] bps=40
      strand-1 5'-AAAAAAAAAATTTTTTTTTTCCCCCCCCCCGGGGGGGGGG-3'
       bp-type    ||||||||||||||||||||||||||||||||||||||||
      strand-2 3'-TTTTTTTTTTAAAAAAAAAAGGGGGGGGGGCCCCCCCCCC-5'
      helix-form  LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
I’ve also tested it on PDB entry 4wb2, using biological unit 1 in the file 4wb2.pdb1, and the following output was obtained:

Code: [Select]

  helix#1[3] bps=9
      strand-1 5'-gcgugugug-3'
       bp-type    |||||||||
      strand-2 3'-cgcacgcac-5'
      helix-form  LLxL.xLx

Since L-DNA isn't a common form and DSSR may require further improvements to handle it better, I haven’t updated the documentation or the note in the DSSR output yet. Please let me know if the new DSSR output makes sense to you and share any questions or suggestions. Future updates to this feature would be much faster than this initial implementation.

Best regards,

Xiang-Jun