Using 3D alignment for virtual screening

Hi,

as I have currently no Java development environment installed on my computer, I hope you can answer my question concerning the 3D alignment algorithm which comes with Marvin 5.2.3.

A possible application for this algorithm would be a ligand based virtual screening, where you use the known conformation of a binder as a rigid template and align multiple compound structures to that in a sequential manner.

My questions is about the coordinates of the aligned structures. Can I directly compare the structures with the original template, e.g. in the binding site of a protein? Or do I need to somehow superpose the coordinates in order to do that?

In other words, is the superposition algorithm keeping the atom coordinates of the template rigid, or does it only fix rotatable bonds and translate the molecule in 3D?

Cheers, Ingo

Hi Ingo,

thank you for your interest in 3D alignment. Your plan to use our alignment API for Virtual Screening is absolutely OK.

Under I call the ligand eg. obtained from the binding site of pdb protein as query and the candidate drug-like molecules in the DB as targets.

To summarize, what the alignment can do:

1. Compare one query and one target by aligning them. During the overlay these molecules are translated and rotated in 3D, and also ALL the rotatable dihedrals can be rotated. Of course you can "freeze" the dihedrals for only one or both of the molecules treating the input conformations as rigid.

Two simple scenarios can be handled:

1a. You have a known binding (bioactive) conformation of the quey so it is held rigid. You have target molecules either in 2D or 3D and they are treated flexible. I would emphasize that ChemAxon can do the 2D->3D conversion with the "Clean3D" module.

1b. If there is no known binding conformation or only a 2D query is accessible: After converting the query to 3D it can treated flexible also.

2. Align more than two molecules onto each other eighter rigid or flexible manner. This can come in handy for 3D QSAR or when creating pharmacopore hypothesis.

The core flexible alignment engine is ready and well tested however its scoring function and initial orientation generator continuously improving. These improvements will lead to nicer and nicer alignments in the versions of the near future. Currently the alignment maximizes something similar as the volumetric overlap of the molecules.

To get a first impression, how alignment works you can try in marvin sketch:

http://www.chemaxon.com/product/align.html

http://www.chemaxon.com/marvin/help/calculations/conformation.html#align

http://www.chemaxon.com/forum/viewpost23792.html#23792

I hope this was useful.

Regards,

Adrian