Bases for 3D Chemical Structures

From the Chemical Abstracts compound registry with lO substances, in order to convert a 2D data base into a 3D one, one needs a program that iterates neither energy minimizations, nor molecular mechanics calculations. Instead, the program CONCORD of the University of Texas at Austin and TRIPOS Associates avoids bad intramolecular contacts in terms of van der Waals radii. This program was also able to convert 2D into 3D structures in about 50% of the cases among all atoms heavier than chlorine, only Br and I can be handled, and molecules with flexible bonds are sometimes obtained in conformations that differ considerably from real structures in crystals, as shown by comparison with the Cambridge Structural Database, which contains about 10 substances. Despite the fact that crystal structures will not necessarily coincide with molecular structures in solution. X-ray data furnish precious information. 

The following hierarchy of molecular features is probable in ascertaining similarity in the context of biological activity  

Disposition of hydrogen-bond donor and acceptor groups 

Under disposition one looks for angles and distances between specified gronps or atoms of given types. 

If one wishes to obtain 3D similarity descriptors ignoring atom types, one starts from distorted tetrahedra (sets of four atoms) with BNB or NBN bonding patterns these (together with distances between pairs of atoms) give the most effective similarity searches, validated by the Kendall test statistic, on using the Cambridge Structural Database (CSD) for carbohydrates. 

---