AF3 unable to predict covalent bonds between DNA and protein

[alchemistcai]

Hi there!

I read the input doc and it says Defining covalent bonds between or within polymer entities is not currently supported.

I'm not sure what polymer exactly means in the context.So I use 1K4T, a protein-DNA covalent complex and run_alphafold.py gives an info:

I1201 13:58:33.149596 128177626838144 structure_cleaning.py:253] Reducing number of bonds for top1_dna_topotecan_hg from 1 to 0, of which 1 are polymer-polymer bonds and 0 are bad bonds.

And finally in pymol that bond doesn't show up.

So,does polymer in #Bond section mean proteins,DNA, and RNA exactly?

Does it mean that covalent bond definition for protein/DNA/RNA is not supported now,while ligand-ligand,ligand-protein,ligand-DNA and ligand-RNA are supported?

If I DO want to model a protein-DNA covalent complex,is it reasonable to define a single DT as a ligand and connect to the protein and other DNA units?

[alchemistcai]

Unfortunately, it seems that changing json only is not enough to generate reasonable covalent bonds between polymer-polymer.

In my test,these two kinds of definition both result in bad bonds :

    "bondedAtomPairs": [  // definition kind 1
      [["H", 1, "C3'"], ["D", 550, "O3P"]],   // connect origina DNA(chain A)  DT10 and protein's modifcation PTR(parent is TYR)
      [["H", 1, "O3P"], ["A", 9, "C3'"]]  // connect O3P of original chain A DT10 and C3' of DT9
    ], 

    "bondedAtomPairs": [ // definition kind 2
      [["H", 1, "C3'"], ["D", 550, "O3P"]], 
      [["H", 1, "P"], ["A", 9, "O3'"]]  // connect P of original chain A DT10 and O3' of DT9
    ]

Both lead to an additional atom O.The second kind also leads to wrong bond order for atom P.

The first definition only generate the bond between DT10 and protein.In cif output file,only one covalent bond occurs.The absent of the bond between DT10 and DT9 has no info nor warning.Also ,the bond length is too long.