Similarity Searching in Databases of 2D Structures

Having outlined the basic rationale for the development of similarity searching, we shall commence our review of the available approaches by discussing two of the characteristics that can be used to evaluate the merits of a particular similarity measure. They are introduced here with reference to measures of 2D similarity but are equally applicable when considering the 3D measures that are described later. 

A further characteristic of a measure that must be considered is its appropriateness, since, as was mentioned in the Introduction, the definition of similarity is application dependent. When very large databases need to be pro- 

---

In the earlier section on the Chemical Abstracts Service (CAS), we gave a brief overview of the range of similarity measures that are under investigation by Fisanick et al. and described the novel descriptors that these workers have developed for similarity searching in databases of 2D structures. Their studies have also involved both geometric and property based 3D similarity measures in all, a molecule is characterized by a maximum of no fewer than 3172 different features. 

Schuffenhauer A, Gillet VJ, Willett P. Similarity searching in files of 3D chemical structures analysis of the BIOSTER database using 2D fingerprints and molecular field descriptors. J Chem Inf Comput Sci 2000 40 295-307. 

Many different structural descriptors have been developed for similarity searching in chemical databases [4] including 2D fragment based descriptors, 3D descriptors, and descriptors that are based on the physical properties of molecules. More recently, attention has focused on diversity studies and many of the descriptors applied in similarity searching are now being applied in diversity studies. Structural descriptors are basically numerical representations of structures that allow pairwise (dis)similarities between structures to be measured through the use of similarity coefficients. Many diversity metrics have been devised that are based on calculating structural (dis)similarities, some of these are described below. 

D similarity searching The motivation to perform 2D similarity searches follows from the similarity property principle. All compounds having a similarity above a certain threshold are retrieved from the database by 2D similarity searching. In most cases the similarity is calculated by one of the aforementioned similarity measures based on molecular fingerprints. Therefore, in the first step the fingerprint for the query structure is calculated. 

The standard screening approach when several active molecules have been identified is pharmacophore mapping followed by 3D database searching. This approach assumes that the active molecules have a common mode of action and that features that are common to all of the molecules describe the pharmacophoric pattern responsible for the observed bioactivity. This is a powerful technique but one that may not be applicable to the structurally heterogeneous hits that characterize typical HTS experiments or sets of competitor compounds drawn from the public literature. In such cases, it is appropriate to consider approaches based on 2D similarity searching and we present here a comparison of approaches for combining the structural information that can be gleaned from a small set of reference structures. 

---