Hydrogen addition problems with non-experimental input structures

[dargen3]

Hello,
I would like to report a strange behavior of the software tool pdb2pqr30.

I am using pdb2pqr30 to protonate a structure with a command:
pdb2pqr30 --log-level DEBUG --with-ph 7.2 AF-P0DSE4-F1-model_v2.pdb AF-P0DSE4-F1-model_v2_protonated.pqr
The version of pdb2pqr30 is: 3.4.1
OS is Ubuntu 21.10

But the structure is protonated probably wrong.

Please help me if I am doing something wrong. If it is a software error, can be the error fixed, please? Alternatively, is there any estimate as to when the error might be fixed? PDB file can be downloaded from https://alphafold.ebi.ac.uk/entry/P0DSE4

Thank you. Regards, Schindler

[sobolevnrm]

Can you please tell us what specific aspect of protonation state is incorrect?

[intendo]

Is it the 3 bonds on the lower left atom that is red or the atoms that are not connected in the lower middle of the image?

[dargen3]

Wow, this is really fast reaction! :) It should be standard serine. So, carbon (grey color) should have 2 hydrogens instead of 3. Moreover distance between carbon and hydrogen is too small (0.65A)

[sobolevnrm]

This seems like it could more likely be a visualization issue rather than an issue with PDB2PQR. The PQR files that PDB2PQR produces have no information about bonding. However, the visualization above is drawing odd bonds between atoms--this sometimes happens in programs such as VMD when the bonding information is inferred from the radii in the PQR file (rather than the built-in radii used for PDB files).

When I visualize the results of the calculation above with PyMOL, I get the following images which do not show the bonding issues in the issue above.

[dargen3]

OK, missing bonds are probably error of Avogadro. But unresolved problem is, that carbon of serine has 3 hydrogens instead of 2. You can see it even in your picture from pymol too.

[dargen3]

Any progress here, please?

[sobolevnrm]

No, I haven't had time to work on this. Sorry.

[dargen3]

And do you have any idea if you'll ever have time to do that? We are planning to use pdb2pqr for a large Alphafold2 related project and I don't know whether to wait for the bug fix or find another tool please?

[sobolevnrm]

I will try to work on it this weekend. Sorry.

[sobolevnrm]

But unresolved problem is, that carbon of serine has 3 hydrogens instead of 2. You can see it even in your picture from pymol too.

This is not an extra atom -- its the alpha carbon hydrogen rotated the wrong way. There's a problem with the input structure that is affecting hydrogen optimization; e.g., see the error messages generated by PDB2PQR:

2022-04-02 08:03:53,582 DEBUG:debump.py:200:find_residue_conflicts:SER A 12 HA is too close to SER A 12 CB
2022-04-02 08:03:53,583 DEBUG:debump.py:161:debump_biomolecule:Starting to debump SER A 12...
2022-04-02 08:03:53,584 DEBUG:debump.py:162:debump_biomolecule:Debumping cutoffs: 2.0 for heavy-heavy, 1.5 for hydrogen-heavy, and 1.0 for hydrogen-hydrogen.
2022-04-02 08:03:53,584 WARNING:debump.py:172:debump_biomolecule:WARNING: Unable to debump SER A 12

I've never seen an debumping issue like this before with experimentally derived structures which is why I suspect it is a problem with the input file. This will take a while to debug.

[dargen3]

This is not an extra atom -- its the alpha carbon hydrogen rotated the wrong way. There's a problem with the input structure that is affecting hydrogen optimization; e.g., see the error messages generated by PDB2PQR:

OK, that is my mistake. You are right. Thank you for error messages.

I've never seen an debumping issue like this before with experimentally derived structures which is why I suspect it is a problem with the input file. This will take a while to debug.

Do you want more problematic structures for debugging? Please let me know if you find a bug in the structure so I can inform the Alphafold developers.

[sobolevnrm]

If you can share some additional structures here, that would be helpful. Thank you.

[HankewieDanke]

Just for documentation purposes, I have encountered this bug with other nn-predictors similar to Alphafold as well (ABodyBuilder2 -- specifically the nanobody predictor).
Sadly I can not share the structures generated.