Different results between web and command line

Describe the bug

Relevant files attached: 3sgj_smr.tar.gz

I'm trying to obtain the PLIP interactions for file 3sgj_smr.pdb (derived from PDB 3SGJ, see attachment above). When I run the same file on the command line, I get very different results than when I run it on the command line

I submitted a request to the web interface by uploading this file: 68269632-0eee-493b-993d-b2a5362840e0. In the SMALLMOLECULE section, under NAG, I can see several hydrogen bonds between NAG-_-1 and several residues in chain A: LYS.24, ASP.43, ASN.75, ARG.79.

I downloaded the XML report from this web report (3sgj_smr_plipweb_report.xml) and I can clearly see those interactions in there.

When I run PLIP on the command line with

plip -f 3sgj_smr.pdb -x

I can only see NAG-_-1 making hydrogen bonds with LYS.24. I can see no interactions with ASP.43, ASN.75 or ARG.79. See the file 3sgj_smr_plipcmd_report.xml in attachment.

This XML report was generated with PLIP 2.1.3, Python 3.6.6 and OpenBabel 3.1.1. However I get similar results with our older build PLIP 1.4.2/Python 2.7.11/OpenBabel 2.4.1 (all on Centos7 systems), as well as the current code on the master branch (which I guess is equivalent to v.2.1.4).

To Reproduce Web:

Go to https://projects.biotec.tu-dresden.de/plip-web/plip/index
Select PDB file and Browse for file 3sgj_smr.pdb
Run analysis
Wait a bit
Open SMALLMOLECULE > NAG, > NAG-_-1 and look at Hydrogen bonds table.
Download XML with "... in XML format" link (giving 3sgj_smr_plipweb_report.xml)

Command line:

Run command plip -f 3sgj_smr.pdb -x
Look at report.xml file (equivalent to 3sgj_smr_plipcmd_report.xml).

Expected behavior I was expecting to see the same interactions reported from the command line run and on the web.

What could explain these differences?

Specifically here are the 2 interactions I can see in 3sgj_smr_plipcmd_report.xml


  <bindingsite id="10" has_interactions="True">
    <identifiers>
      <longname>NAG</longname>
      <ligtype>SMALLMOLECULE</ligtype>
      <hetid>NAG</hetid>
      <chain>_</chain>
      <position>1</position>
      <composite>False</composite>
      <members>
        <member id="1">NAG:_:1</member>
      </members>
      <smiles>OC[C@H]1O[CH][C@@H]([C@H]([C@@H]1O)O)NC(=O)C</smiles>
      <inchikey>IKCYXSONMXYELC-LXGUWJNJSA-N
</inchikey>
    </identifiers>
    <lig_properties>
      <num_heavy_atoms>14</num_heavy_atoms>
      <num_hbd>4</num_hbd>
      <num_unpaired_hbd>2</num_unpaired_hbd>
      <num_hba>6</num_hba>
      <num_unpaired_hba>4</num_unpaired_hba>
      <num_hal>0</num_hal>
      <num_unpaired_hal>0</num_unpaired_hal>
      <num_aromatic_rings>0</num_aromatic_rings>
      <num_rotatable_bonds>3</num_rotatable_bonds>
      <molweight>204.20046</molweight>
      <logp>-1.8433000000000002</logp>
    </lig_properties>
    <interacting_chains>
      <interacting_chain id="1">A</interacting_chain>
    </interacting_chains>
    <bs_residues>
      <bs_residue id="1" contact="False" min_dist="7.3" aa="LEU">20A</bs_residue>
      <bs_residue id="2" contact="False" min_dist="7.2" aa="VAL">37A</bs_residue>
      <bs_residue id="3" contact="False" min_dist="6.2" aa="PRO">23A</bs_residue>
      <bs_residue id="4" contact="False" min_dist="3.8" aa="PHE">21A</bs_residue>
      <bs_residue id="5" contact="False" min_dist="6.3" aa="GLU">36A</bs_residue>
      <bs_residue id="6" contact="False" min_dist="5.8" aa="VAL">81A</bs_residue>
      <bs_residue id="7" contact="False" min_dist="7.3" aa="PHE">19A</bs_residue>
      <bs_residue id="8" contact="False" min_dist="4.3" aa="ARG">79A</bs_residue>
      <bs_residue id="9" contact="False" min_dist="4.5" aa="VAL">40A</bs_residue>
      <bs_residue id="10" contact="False" min_dist="3.8" aa="THR">38A</bs_residue>
      <bs_residue id="11" contact="False" min_dist="6.8" aa="CYS">39A</bs_residue>
      <bs_residue id="12" contact="False" min_dist="4.5" aa="PRO">22A</bs_residue>
      <bs_residue id="13" contact="True" min_dist="2.6" aa="LYS">24A</bs_residue>
    </bs_residues>
    <interactions>
      <hydrophobic_interactions/>
      <hydrogen_bonds>
        <hydrogen_bond id="1">
          <resnr>24</resnr>
          <restype>LYS</restype>
          <reschain>A</reschain>
          <resnr_lig>1</resnr_lig>
          <restype_lig>NAG</restype_lig>
          <reschain_lig>_</reschain_lig>
          <sidechain>True</sidechain>
          <dist_h-a>2.52</dist_h-a>
          <dist_d-a>3.45</dist_d-a>
          <don_angle>162.86</don_angle>
          <protisdon>False</protisdon>
          <donoridx>4918</donoridx>
          <donortype>O3</donortype>
          <acceptoridx>151</acceptoridx>
          <acceptortype>N3</acceptortype>
          <ligcoo>
            <x>-23.495</x>
            <y>-0.723</y>
            <z>16.244</z>
          </ligcoo>
          <protcoo>
            <x>-21.042</x>
            <y>1.101</y>
            <z>14.634</z>
          </protcoo>
        </hydrogen_bond>
        <hydrogen_bond id="2">
          <resnr>24</resnr>
          <restype>LYS</restype>
          <reschain>A</reschain>
          <resnr_lig>1</resnr_lig>
          <restype_lig>NAG</restype_lig>
          <reschain_lig>_</reschain_lig>
          <sidechain>True</sidechain>
          <dist_h-a>1.72</dist_h-a>
          <dist_d-a>2.58</dist_d-a>
          <don_angle>139.53</don_angle>
          <protisdon>True</protisdon>
          <donoridx>151</donoridx>
          <donortype>N3</donortype>
          <acceptoridx>4917</acceptoridx>
          <acceptortype>O3</acceptortype>
          <ligcoo>
            <x>-21.147</x>
            <y>-1.467</y>
            <z>14.833</z>
          </ligcoo>
          <protcoo>
            <x>-21.042</x>
            <y>1.101</y>
            <z>14.634</z>
          </protcoo>
        </hydrogen_bond>
      </hydrogen_bonds>
      <water_bridges/>
      <salt_bridges/>
      <pi_stacks/>
      <pi_cation_interactions/>
      <halogen_bonds/>
      <metal_complexes/>
    </interactions>
    <mappings>
      <smiles_to_pdb>5:4908,6:4909,7:4910,8:4911,3:4912,2:4913,12:4914,14:4915,11:4916,10:4917,9:4918,4:4919,1:4920,13:4921</smiles_to_pdb>
    </mappings>
  </bindingsite>

I can see no hydrophobic interactions:

<hydrophobic_interactions/>

In contrast I can see 7 hydrogen bonds and 1 hydrophobic interactions on the web page:

(and also in the XML file downloaded there)

Dear xrobin,

thanks for reporting this inconsistency. First of all, you should prefer to use the v2.1.4 command line tool in your case as the web server is (unfortunately) running an outdated version of PLIP. So theoretically it suffers from all the issues that we resolved recently, e.g. inconsistent hydrogen bond annotation, etc.

For your particular case it could be related to the custom input structure you are providing. This structure contains a carbohydrate polymer, which is missing the link entries present in the original PDB file required for PLIP to recognize this:

...
LINK         O4  NAG D   2                 C1  BMA D   3     1555   1555  1.41  
LINK         O3  BMA D   3                 C1  MAN D   4     1555   1555  1.43  
LINK         O6  BMA D   3                 C1  MAN D   6     1555   1555  1.42  
LINK         O2  MAN D   4                 C1  NAG D   5     1555   1555  1.43  
...

For now I can only speculate that the old version of PLIP (or maybe even the older OpenBabel version) handles the small molecules differently if no link entries are present.

If I use the structure from PDB 3sgj, the results are perfectly consistent between the web server and the command line (shown for ligand 3SGJ-NAG-D-1 at the end of this post).

The web server is maintained by our academic collaborators and afaik they are planning a relaunch currently. So, in a nutshell, I would kindly ask you to rely on the latest command line implementation for now and to provide link entries in your custom PDB files whenever you want to handle a polymer ligand. I have added a comment to the front page README that makes users aware of this issue.

Command Line

3SGJ_NAG_D_1_cmd

Web Server

3SGJ_NAG_D_1_web

Hi Florian,

Thanks a lot for your quick answer and the useful pointer. I've been investigating the issue in a bit more depth and I think I might have found where the issue actually comes from.

Indeed with the LINK annotations PLIP recognizes the sugar as a single polymer molecule and the output gets cleaner. However this seems to have little impact on the interactions that are detected (or at least only to the extent to change which bonds are reported; we still seem to be missing many of them).
It turns out what seems to bug PLIP the most are the insertion codes. For various internal/historical reasons we've been using a representation where the residues a polymer ligand have the same residue number and are distinguished with insertion codes. So for instance we have several NAG residues in position 1: 1A, 1B, 1E and 1G, as well as two MAN: 1D and 1F. If I renumber residues 1A-1G to 1-8 (similar to what's in the original PDB file) I can see that now all the bonds are reported by PLIP. Looking at the <members> section of the XML report, it looks like only a single residue with the same name/residue number pair is considered.
```
  <members>
    <member id="1">NAG:_:1</member>
    <member id="2">NAG:_:2</member>
    <member id="3">BMA:_:3</member>
    <member id="4">MAN:_:4</member>
    <member id="5">NAG:_:5</member>
    <member id="6">MAN:_:6</member>
    <member id="7">NAG:_:7</member>
    <member id="8">FUC:_:8</member>
  </members>
```
Compared with the results with insertion codes:
```
  <members>
    <member id="1">FUC:_:1</member>
    <member id="2">NAG:_:1</member>
    <member id="3">BMA:_:1</member>
    <member id="4">MAN:_:1</member>
  </members>
```
So it looks like some of the NAG 1 residues are ignored...

I'm attaching the "fixed" files with the LINK lines, both with insertion codes (1A-1G) and without (1-8) in case you're interested to check them out.: 3sgj_smr_link.tar.gz.

I will investigate how we can avoid insertion codes in our application.

Dear Xavier,

thanks for checking that so deeply. Indeed, PLIP gets puzzled by insertion codes and all alternative locations are either:

removed (default)
or handled as separate ligand with --altlocation flag.

The intended behavior that alternative locations describe monomers of a polymer is not supported at the moment. Is it fine for you to get away from insertion codes? I think implementing such a "feature" would be tedious.

Best Florian

Unfortunately (and quite unsurprisingly) the altlocation flag has no effect on insertion codes.

That being said, I can understand that you don't want to support a legacy feature that's no longer used for residue numbering in mmCIF files.

Maybe only one thing: could PLIP display a warning when it finds an insertion code, to make it clear that it's not supported? As there were results, with bonds that actually made sense, it took me way too long to even realize that something was off.

You're right, the alternative location flag does not affect insertion codes. I will convert this to a milestone for the next release.

pharmai / plip

Different results between web and command line #73

Command Line

Web Server