3D Alignment

I am using MarvinSketch 5.5.0.1 in Mac OS X (10.6.7). When using the 3D alignment feature, I get different results depending upon several factors. For example, the initial "autopilot" result with two particular compounds using 2 user-defined atoms and the first option (align by extended atom types) gives a Tanimoto coefficient of 0.318. However, by visual inspection, the molecules appear to be quite similar in shape. If I right-click in the MarvinView window that shows the alignment and select the option to transfer the structures back to MarvinSketch, the overlapping structures then appear in MarvinSketch. If I then repeat the 3D alignment, the Tanimoto coefficient improves dramatically to 0.704. Is this interative procedure legitimate? Is there a way to get optimal alignment without using an interative procedure such as this?

By the way, if i use another program (OpenEye vROCS) to do the alignment automatically without interventions such as user-assigned atoms or user interations, I get a "shape Tanimoto" of 0.739 and a "color Tanimoto" of 0.130, for a combined Tanimoto of 0.869. Although the absolute Tanimotos from different programs using different algorithms cannot be directly compared, it is perhaps of interest that both procedures give coefficients >0.70. However, the visual alignment appears better with MarvinSketch; that is, the molecules are aligned in an intuitive way with similarly shaped parts of the molecules overlapping. Yasara-Structure also produces an intuitive alignment without user intervention, but it does not calculate a Tanimoto coefficient (it will, however, do an RMSD based on the number of matched atom pairs).

Hi,

"Is there a way to get optimal alignment without using an interative procedure such as this?"

The alignment algorithm is parametrized for the fastest operation in Msketch menu. To go for more accurate alignment please increae the "Initial conformation count " in the 3D alignment window from the default 2 to ~4 or ~7 .

The alignment is also designed to work in batch mode with minimal user interaction for lots of molecules. This functionality is accessible in jchem / screen3d.

I would suggest to compare 3D Tanimotos of different molecules calculated with the same software. Different algorithms use different equations and methodology to calculate 3d similarity.

Kind regards,

Adrian