Question

The CDK used to have a library to validate chemical graphs against common problems, including unknown atom types. Additionally, the file reader for the MDL SD format has a STRICT mode, which makes it fail if the file is not conform the format. However, the CDK currently does not have a library of command line utilities that make functionality available.

What Open Source tools are available to inspect the file and the content of a MDL SD file, and report errors?

Answer 1

Hello,

I would suggest to use a simple program based on Indigo toolkit:

#include "molecule/molecule.h"
#include "molecule/sdf_loader.h"
#include "molecule/molfile_loader.h"
#include "base_cpp/scanner.h"

int main (int argc, char *argv[])
{
   bool calc_hydro = true;
   bool allow_query = false;
   const char *idfield = 0;

   if (argc <= 1)
   {
      fprintf(stderr,
              "Usage: sdfcheck file.sdf [-id idfield] [-no-hydro-check] [-allow-query]\n");
      return -1;
   }

   const char *filename = argv[1];

   for (int i = 2; i < argc; i++)
   {
      if (strcmp(argv[i], "-id") == 0)
      {
         if (++i == argc)
         {
            fprintf(stderr, "expecting an identifier after '-id'\n");
            return -1;
         }
         idfield = argv[i];
      }
      else if (strcmp(argv[i], "-no-hydro-check") == 0)
         calc_hydro = false;
      else if (strcmp(argv[i], "-allow-query") == 0)
         allow_query = true;
      else
      {
         fprintf(stderr, "unknown parameter: %s\n", argv[i]);
         return -1;
      }
   }

   try
   {
      FileScanner scanner(filename);
      SdfLoader loader(scanner);
      int cnt = 0;

      if (idfield != 0)
         loader.initProperties(idfield);

      while (!loader.isEOF())
      {
         loader.readNext();

         const char *id = 0;

         if (idfield != 0 && loader.properties.at2(idfield) != 0)
            id = loader.properties.at(idfield).ptr();

         cnt++;

         if (idfield == 0)
            printf("reading molecule #%d... ", cnt);
         else if (id == 0)
            printf("reading molecule #%d, no %s ... ", cnt, idfield);
         else
            printf("reading molecule #%d, %s=%s ... ", cnt, idfield, id);

         fflush(stdout);

         Molecule mol;
         try
         {
            BufferScanner molscan(loader.data);
            MolfileLoader molload(molscan);

            molload.loadMolecule(mol, allow_query);
            if (calc_hydro)
               mol.calcImplicitHydrogens();
            printf("\n");
         }
         catch (Exception &e)
         {
            printf("error: %s\n", e.message());
            continue;
         }
      }
   }
   catch (Exception &e)
   {
      printf("error: %s\n", e.message());
   }

   return 0;
}

The steps to compile the program on Linux are:

1. Download any Indigo source package containing 'graph' and 'molecule' projects (actually, every our package has them, so you can download any of them, for example Bingo)
2. Unpack bingo-src-XXXX.zip
3. Go to bingo-src-XXXX folder
4. Paste the above code into sdfcheck.cpp file
5. cd graph; make CONF=Release32; cd ..
6. cd molecule; make CONF=Release32; cd ..
7. gcc sdfcheck.cpp -o sdfcheck -O3 -m32 -I. -Icommon molecule/dist/Release32/GNU-Linux-x86/libmolecule.a graph/dist/Release32/GNU-Linux-x86/libgraph.a -lpthread -lstdc++

(In order to get a 64-bit executable, replace 32 with 64 in steps 5, 6, and 7.)

Then you can run the program like this:

./sdfcheck ~/sdf/database.sdf | grep error

or

./sdfcheck ~/sdf/database.sdf -id molregno -no-hydro-check | grep error

or

./sdfcheck ~/sdf/database.sdf -id molregno -no-hydro-check -allow-queries | grep error

With best regards,

Dmitry

Answer 2

OpenBabel has some checking code, as part of the parsing. It isn't designed for giving reports and just sends the messages to the log.

For what it's worth, my experience has been that most people don't care much for format validation. Who would use this?

Answer 3

Being curious in nature, and thus hacker, I worked out a solution for the CDK, with the option for line numbers of positions. The patch is not entirely finished yet, but I blogged about it with some examples:

http://chem-bla-ics.blogspot.com/2010/01/validating-mdl-sd-files-and-symyx.html