Re: [Rdkit-discuss] Loading mols into postgresql using cartridge

[Greg L. <gre...@gm...>]

On Wed, Aug 22, 2012 at 5:20 AM, Greg Landrum <gre...@gm...> wrote:
>
> The current version of the cartridge does not support loading (or
> exporting) molecules as RDKit pickles. It should be pretty easy for me
> to add this capability though, so I will look into it.

The svn version of the cartridge is now capable of sending/receiving
molecules and bit-vector fingerprints as binary strings (bytea's in
postrgresql).

Here's a demonstration from Python.

For molecules:

In [7]: m = Chem.MolFromSmiles('c1ccccc1')
In [8]: mpkl = m.ToBinary()
In [9]: curs.execute("select mol_from_pkl(%s)",(buffer(mpkl),))
In [10]: curs.fetchone()
Out[10]: ('c1ccccc1',)
In [11]: curs.execute("select mol_to_pkl('c1ccccn1'::mol)")
In [12]: mpkl = curs.fetchone()[0]
In [13]: Chem.MolToSmiles(Chem.Mol(str(mpkl)))
Out[13]: 'c1ccncc1'

For bit vects:

In [14]: fp = AllChem.GetMorganFingerprintAsBitVect(Chem.MolFromSmiles('c1ccccc1'),2,512)
In [15]: pkl = DataStructs.BitVectToBinaryText(fp)
In [16]: curs.execute("select size(bfp_from_binary_text(%s))",(buffer(pkl),))
In [17]: curs.fetchone()
Out[17]: (512,)
In [18]: curs.execute("select
tanimoto_sml(bfp_from_binary_text(%s),morganbv_fp('c1ccccc1'))",(buffer(pkl),))
In [19]: curs.fetchone()
Out[19]: (1.0,)
In [20]: curs.execute("select bfp_to_binary_text(morganbv_fp('c1ccccc1'))")
In [21]: pkl = curs.fetchone()[0]
In [23]: nfp = DataStructs.CreateFromBinaryText(str(pkl))
In [24]: list(nfp.GetOnBits())
Out[24]: [64, 337, 389]
In [25]: list(fp.GetOnBits())
Out[25]: [64, 337, 389]


The corresponding C++ calls are:
 - to get a molecule pickle: MolPickler::pickleMol
 - to construct a molecule from a pickle: ROMol(pkl)
 - to get a fingerprint as binary text: BitVectToBinaryText
 - to fill an already constructed fingerprint from binary text (the
size of the fp should be 8* the length of the binary text):
UpdateBitVectFromBinaryText

Best,
-greg