problem with sdf->pdb conversion

Hello,

I'm trying to extract PDB/CIF files from a multi-structure SDF file.  At
the moment, I'm just writing out one at a time (based on molecule #
using -f and -l command-line options).  I have confirmed that the
extract is correct by writing to an intermediate MOL file; however, the
PDB file that is produced is quite incorrect, with improper
connectivity, etc., and the corresponding CIF file is also incorrect.

Here's the MOL snippet:

Structure36
 OpenBabel05300818252D

 11 11  0  0  0  0  0  0  0  0999 V2000
    1.2091    5.5646    0.0000 O   0  0  0  0  0
    2.7394    6.4432    0.0000 C   0  0  0  0  0
    1.2091    4.1920    0.0000 C   0  0  0  0  0
    2.4182    3.4862    0.0000 C   0  0  0  0  0
    0.0000    3.4862    0.0000 C   0  0  0  0  0
    3.4957    4.1061    0.0000 C   0  0  0  0  0
    2.4182    2.0725    0.0000 C   0  0  0  0  0
    0.0000    2.0725    0.0000 C   0  0  0  0  0
    4.6695    3.4251    0.0000 O   0  0  0  0  0
    1.2091    1.3667    0.0000 C   0  0  0  0  0
    1.2091    0.0000    0.0000 Br  0  0  0  0  0
  1  2  1  0  0  0
  1  3  1  0  0  0
  3  4  2  0  0  0
  3  5  1  0  0  0
  4  6  1  0  0  0
  4  7  1  0  0  0
  5  8  2  0  0  0
  6  9  1  0  0  0
  7 10  2  0  0  0
  8 10  1  0  0  0
 10 11  1  0  0  0
M  END

What I am doing wrong? No matter what I do, I can't seem to get a vali
PDB using OpenBabel, either directly from the SDF or the MOL file
itself.  However, I know the MOL output is valid, as I've used other
tools that can successfully produce a PDB/CIF pair (e.g. PRODRG).

Thanks in advance for any help.

Roni



-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

On dimanche 01 juin 2008, Roni Gordon wrote:
> I'm trying to extract PDB/CIF files from a multi-structure SDF file.  At
> the moment, I'm just writing out one at a time (based on molecule #
> using -f and -l command-line options).  I have confirmed that the
> extract is correct by writing to an intermediate MOL file; however, the
> PDB file that is produced is quite incorrect, with improper
> connectivity, etc., and the corresponding CIF file is also incorrect.

   Could you supply an example sdf file and the exact command used that
produces the invalid pdb and cif output ? And open a bug report on sf (see:
http://sourceforge.net/tracker/?group_id=40728).
   What version of OB are you using ?

--
Vincent Favre-Nicolin
Université Joseph Fourier
http://v.favrenicolin.free.fr
ObjCryst & Fox : http://objcryst.sourceforge.net

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

On May 30, 2008, at 6:28 PM, Roni Gordon wrote:

> extract is correct by writing to an intermediate MOL file; however,  
> the
> PDB file that is produced is quite incorrect, with improper
> connectivity, etc., and the corresponding CIF file is also incorrect.

I don't think translating to a CIF is a good idea here -- you're  
taking a MOL file with 2D coordinates only and expecting to have good  
fractional 3D coordinates. As far as PDB, I'd be curious to know what  
you mean by "quite incorrect." The PDB file I generate (attached) has  
connect records, HETATOM, etc. If you can be a bit more specific  
(e.g., version of Open Babel, etc. as mentioned previously) it would  
certainly help.

Cheers,
-Geoff




-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

Well, I'm starting from an SDF file, which I presume has correct 3D
coordinates; the reason I used MOL as an intermediate was to use PRODRG
to produce a "valid" PDB file that I know is correct (this was my
previous procedure before I discovered OpenBabel).

I've attached the two different PDB files from babel v2.1.1 vs prodrg
using the same MOL input (in my case, I've included Hs, but it doesn't
seem make a difference).  My modeling software (Coot) doesn't connect
the atoms in the same way, so I'm confused.

Roni

On Mon, 2008-06-02 at 12:55 -0400, Geoffrey Hutchison wrote:
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

        Hi,
[Please answer to the ML so that all can help]

On lundi 02 juin 2008, Roni Gordon wrote:
> Sorry -- using OB v2.1.1 with the following commands:
>
> openbabel -isdf library.sdf -omol output.mol -f 36 -l 36
>
> openbabel -isdf library.sdf -opdb output.pdb -f 36 -l 36
>
> Unfortunately, I can't disclose the contents of the SDF file, it has
> thousands of compounds.

  Could you extract a sub-file with just a few structures so that we can
reproduce the problem ?
  And as Geoff mentioned - what do you mean by "quite incorrect" - wasn't the
original data 2D ?
 
--
Vincent Favre-Nicolin
Université Joseph Fourier
http://v.favrenicolin.free.fr
ObjCryst & Fox : http://objcryst.sourceforge.net

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

On mardi 03 juin 2008, Roni Gordon wrote:
> Well, I'm starting from an SDF file, which I presume has correct 3D
> coordinates; the reason I used MOL as an intermediate was to use PRODRG
> to produce a "valid" PDB file that I know is correct (this was my
> previous procedure before I discovered OpenBabel).
>
> I've attached the two different PDB files from babel v2.1.1 vs prodrg
> using the same MOL input (in my case, I've included Hs, but it doesn't
> seem make a difference).  My modeling software (Coot) doesn't connect
> the atoms in the same way, so I'm confused.

   In the attached test.mol, only 2D coordinates are included ? This can't be
the intermediate mol file used as input ? Unless prodrg does the 2D->3d
transform ?

--
Vincent Favre-Nicolin
Université Joseph Fourier
http://v.favrenicolin.free.fr
ObjCryst & Fox : http://objcryst.sourceforge.net

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

Dear all,

> Well, I'm starting from an SDF file, which I presume has correct 3D
> coordinates; the reason I used MOL as an intermediate was to use PRODRG
> to produce a "valid" PDB file that I know is correct (this was my
> previous procedure before I discovered OpenBabel).
>
> I've attached the two different PDB files from babel v2.1.1 vs prodrg
> using the same MOL input (in my case, I've included Hs, but it doesn't
> seem make a difference).  My modeling software (Coot) doesn't connect
> the atoms in the same way, so I'm confused.

The only difference I see is atoms order. To answer to Vincent, I
think test.mol is already a 3D conformation, but the molecule is
mostly planar. Prodrg doesn't transform from 2D to 3D.

In my tests with openbabel 2.1.1 :
- Sybyl doesn't read well test-babel.pdb (only 4 atoms are loaded)
- Pymol is OK with test-babel.pdb, but all bonds are single (well, OK,
it's normal)
- Cactvs Browser is OK with test-babel.pdb, the cyclic bonds are
correctly double/single
- Mview (Chemaxon) reads well  test-babel.pdb, but with "~" (any bond)
connection

Starting from a mol2 ligand, converted in pdb by openbabel or
molconvert (Chemaxon):
- Sybyl doesn't read well the ligand.pdb
- Pymol (all single bond) and Cactvs browser (double bonds where needed) are OK
- Mview is unable to load the ligand.pdb...

Again, between ligand.pdb"openbabel" and ligand.pdb converted with
sybyl (starting from mol2), I don't see any difference, apart from
atom ordering.

Hope these small comments could help.

I wish Coot could read mol2, it would help a lot...

By the way, when converting a ligand mol2 in cif, it lost the bond
type information (openbabel 2.1.1). Is there a way to keep bond type
in .cif file when converting with openbabel ? Because mol2->pdb->cif
with eg pdb2cif program is not very useful...

Regards,
Pascal

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

On mercredi 4 juin 2008, Pascal Muller wrote:
> The only difference I see is atoms order. To answer to Vincent, I
> think test.mol is already a 3D conformation, but the molecule is
> mostly planar. Prodrg doesn't transform from 2D to 3D.

  In the test.mol and babel.pdb file, all non-H atoms have 0 as z-coordinate,
but not in the prodrg: the -CH2-OH ligand has its OH group out of the ring
plane, and the O-CH3 ligand has its methyl group out-of-plane.
 ... So there has been a 2D->3D transform, even if it does not really break
the molecule, just reduces steric problems.

  btw rasmol reads both pdb's fine. And both molecules look fine too except
for some steric problems in the original 2D structure.

> By the way, when converting a ligand mol2 in cif, it lost the bond
> type information (openbabel 2.1.1). Is there a way to keep bond type
> in .cif file when converting with openbabel ? Because mol2->pdb->cif
> with eg pdb2cif program is not very useful...

  Well, there is a _chemical_conn_bond_type CIF keyword which we could use for
that. Of course it would require the software that reads the CIF to
understand that keyword.
  Would that be OK ?

(There is also a _geom_bond_multiplicity but that does seem to correspond to a
number of bonds rather than a bond type.)

  See the definitions @
http://www.iucr.org/iucr-top/cif/cifdic_html/1/cif_core.dic/index.html
 
        Vincent
--
Vincent Favre-Nicolin                   http://vincefn.net
Université Joseph Fourier       http://www.ujf-grenoble.fr
CEA/ Institut Nanosciences & Cryogénie  http://inac.cea.fr
ObjCryst & Fox             http://objcryst.sourceforge.net

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

Hi Vincent,

That does appear to be the case... I went back and checked the SDF to
find all Z=0.  PRODRG has the option for energy minimization, which
gives a reasonable 3D structure with appropriate CIF constraints; but
it's web-based.  Are there any equivalent command line tools, or other
ideas for producing 3D coordinates?

Thanks again,

Roni



-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
OpenBabel-discuss mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openbabel-discuss

Dear friends,

I have converted ZINC database SDF file to pdb with openbabel. Then I docked the molecule and checked the interaction with LigPlot, I found the bond order of oxygen were wrong. (Not like those of original ZINC molecule.) Similarly, when I visualize it with Marvin-sketch the bond order of oxygen were not correct. Please find the required file attached.LigPlot.pngZINC12958901.sdfZINC12958901.pdb

Regards
Jyoti

The article you have shared here very awesome
potty racers

The problem with converting sdf-> pdb you share quite well. I learned and understood many things from this problem. I hope you will have many articles to share with us.
vex 3