Development of an atom mapping procedure for similarity searching in databases of three‐dimensional chemical structures

Abstract

A BSTRA CT This paper discusses a procedure for culculating the degree of similarity between an input target molecule and each of the molecules in a database qf 3 0 chemical structures. A mapping procedure is used to identijy pairs ofutoms, one from each of the molecules that are being compared, that lie at the centre of topographically similar volumes of 30 space. E.yperiments ure reported to validate the signijicance of the similarities that result from the use of the procedure and to determine the eflect on the calculated similarities of the structural representation that is used to characterise the molecules in a database.

Techniques for the calculation of three-dimensional structural similarity using inter-atomic distances

“…This has an associated probability of occurrence under the Sign test of 0.005, thus validating the use of the method for ranking databases of 3-D structures. A more detailed comparison of the rankings resulting from the Atom Mapping method and from random selection is presented elsewhere [19]. In this paper, we have discussed several techniques that allow the calculation of the degree of structural similarity between pairs of 3-D molecules using inter-atomic distance information.…”

“…Our initial studies have focussed on the first of these, since we have considerable previous experience of the use of inter-atomic distances for 3-D substructure searching (see, e.g., Ref. 18); the use of angular information for similarity searching is considered elsewhere [19].…”

Similarity recognition of molecular structures by optimal atomic matching and rotational superposition

“…So far, methods that match atoms between two molecules follow a heuristic approach that just provide an approximate solution to the atomic matching problem. [23][24][25] For proteins, secondary structure elements (SSE) have been matched heuristically by the stable-marriage algorithm. [26] The Hungarian algorithm [27][28][29] which finds an optimum matching between SSEs was also discussed in Ref.…”

Three-Dimensional Pattern Recognition from Molecular Distance Minimization

“…Molecular similarity procedures are needed for various purposes, as structural alignment of proteins, − chemical reactivity studies, − three-dimensional molecular template recognition, , 3D-structure−activity relationships, − pharmacophore identification, and drug design. − There are two types of molecular similarity: the shape similarity and the electronic similarity. Most shape similarity procedures are based upon interatomic distances. − Even when the two molecules have the same number of atoms, a rigorous treatment of the distances matrix leads to heavy combinatorial algorithms. Thus, efforts have been provided to find correspondences between subsets of atoms. − When the pairwise correspondence between two sets of same cardinality is known, the optimal superposition problem is analytically solvable with the Procuste algorithm. , All these geometric algorithms assumed that a molecule is a set of points.…”