Three-Dimensional Similarity Based on Geometric Properties

Three-Dimensional Similarity Based on Geometric Properties 

If molecules are thought of as wire models, then one would select a geometric definition of similarity that depended on the location of the atomic nuclei. Similarity might be based on the distances between atoms of a specified type or between all atoms in the molecule. In medicinal chemistry one would typically measure the distances between the pharmacophoric atoms or their projected binding points. Such a description is rather straightforward and obvious for small molecules however, it is also a very powerful descriptor of protein structure. For example, one can easily recognize elements of secondary struaure such as helices or sheets from a shaded diagram of the distance between every Co of a protein.  

Particularly in molecular graphics, a commonly used steric variation of this concept is the practice of superimposing molecules by minimizing the root mean square (rms) deviation between corresponding atoms, either all or selected atoms, in two molecules. Alternatively, one may superimpose or compare molecules on the basis of positions of proposed macromolecular binding 

It is frequently important to consider not just the input conformation of a molecule, but several of its possible conformations. If the dataset contains even a few flexible molecules, the number of comparisons quickly becomes very large. To address this problem, Sheridan et al. used distance geometry 

---