Re: [Rdkit-discuss] pharmacophore align

[Taka S. <ser...@gm...>]

Dear Greg, and rdkit-discuss members.
Thank you for your advice.
That code is very useful for me !

And I have one question.
Can I get selected pharmacophore feature positions from AtomIds?
Or how can I get selected pharmacophore features from AtomIds?
>From feature object, it is easy to do that by useing "feature.GetPos()" method.
I want to do same thing by using AtomIds.

Sincerely.
Taka

2012/7/20 Greg Landrum <gre...@gm...>:
> Dear Taka,
>
> On Thu, Jul 19, 2012 at 9:44 AM, Taka Seri <ser...@gm...> wrote:
>> Dear all.
>> I have some questions about Pharmacophore.
>> I want to align molecules that from SDF files,  by Pharmacophore.
>> At first, I set Pharmacophore by using "EmbedPharmacophore()" method.
>> And checked molecules by "EmbedPharmacophore()" method.
>> Then,  generated 3D structures by "EmbedPharmacophore()" method.
>> Finally, I want to align these conformers to the Pharmacophore.
>> Could anyone give me a small example?
>
> There is unfortunately no really good sample code available for this.
> There is some very old code from a GUI application that was built with
> the RDKit but that is no longer supported (or part of the current SVN
> app) here:
> http://rdkit.svn.sourceforge.net/viewvc/rdkit/trunk/Python/qtGui/Search3D/SearchUtils.py?revision=2&view=markup&pathrev=5
> in the function AlignMatchToReference()
>
>
> Here's an attempt to distill that information down:
>
> First you need the alignment package:
>
> from rdkit.Numerics  import rdAlignment
>
> which contains the function GetAlignmentTransform:
>
> In [5]: rdAlignment.GetAlignmentTransform?
> Type:       function
> Base Class: <type 'builtin_function_or_method'>
> String Form:<Boost.Python.function object at 0x102fe0dc0>
> Namespace:  Interactive
> Docstring:
> GetAlignmentTransform( (object)refPoints, (object)probePoints [,
> (object)weights=[] [, (bool)reflect=False [, (int)maxIterations=50]]])
> -> object :
>     Compute the optimal alignment (minimum RMSD) between two set of points
>
>
>      ARGUMENTS:
>
>         - refPoints : reference points sepcified as a N by 3 Numeric array or
>                       sequence of 3-sequences or sequence of Point3Ds
>         - probePoints : probe points to align to reference points - same format
>                       restrictions as reference points apply here
>         - weights : optional numeric vector or list of weights to
> associate to each pair of points
>         - reflect : reflect the probe points before attempting alignment
>         - maxIteration : maximum number of iterations to try to minimize RMSD
>
>      RETURNS:
>
>         a 2-tuple:
>           - SSD value for the alignment
>           - the 4x4 transform matrix, as a Numeric array
>
>
> This will give the transformation required to align one set of points
> (the ph4 points from your embedded molecule) to another set of points
> (the ph4 points from your reference ph4).
>
> To do this you need the positions of each of the features:
> probePts = [list(x.GetPos()) for x in probeFeats]
> refPts = [list(x.GetPos()) for x in refFeats]
>
> And then you align them:
> ssd,tform = Aligner.GetAlignmentTransform(refArr,probeArr,weights=weights)
>
> If your molecule has no chiral centers, it's a good idea to try
> reflecting the alignment:
> ssd2,tform2 = Aligner.GetAlignmentTransform(refArr,probeArr,weights=weights,reflect=True)
> if ssd2<ssd:
>         tform = tform2
>         ssd = ssd2
>
>
> the two return values are the sum of squared deviations of the atomic
> positions (easily convertible to the RMSD) and the transformation
> matrix. You can apply the transformation matrix to your embedded probe
> molecule (to align it to the ph4) with the function
> AllChem.TransformMol:
> AllChem.TransformMol(probeMol,tform)
>
> I hope this helps.
>
> -greg