Molecular dissimilarity

Subheading 2.5. provides a very brief discussion of molecular dissimilarity measures that are basically the complement of their corresponding molecular similarity measures. This section also presents reasons as to why similarity is preferred over dissimilarity, except in studies of diversity, as a measure of molecular resemblance. 

Bawden, D. Molecular dissimilarity in chemical information systems, hi Chemical Structures2. The InternationalLanguage ofchemistry, Ed. Warr, W.A., 1993, Heidelberg Springer-Verlag, pp. 383-388. 

A diversity metric is a function to aid the quantification of the diversity of a set of compounds in some predefined chemical space. Diversity metrics fall into three main classes (1) Distance-based methods, which express diversity as a function of the pairwise molecular dissimilarities defined through measurement. (2) Cell-based methods, which define diversity in terms of occupancy of a finite number of cells that represent disjoint regions of chemical space. (3) Variance-based methods, which quantify diversity based on the degree of correlation between a compound s important features. 

D. Bawden, in Chemical Structures 2 The International Language of Chemistry, W. A. Warr, Ed., Springer-Verlag, Berlin, 1993, pp. 383-388. Molecular Dissimilarity in Chemical Information Systems. 

The concept of molecular dissimilarity is introduced, and shown to be a powerful complement to the well-established notion of molecular similarity. It provides a quantitative assessment of structural variation and diversity. Apphcations within chemical information systems are discussed. These include ranking of search output, selection of representative sets of structures, file screening, data analysis, and creativity stimulation. 

In this presentation, I wish to introduce the related topic of molecular dissimilarity, and to suggest that it may prove to be of equal, or perhaps even greater, value. This concept is not so well-established as its converse. It has been implemented in the Pfizer in-house systems, and has also been investigated at Upjohn. Of the commercial systems, only ORAC/OSAC currently offers a specific dissimilarity option. 

I suggest that, on the contrary, molecular dissimilarity may actually be an even more powerful concept than similarity. This is because calculation of dissimilarity gives quantitative access to the twin concepts of structural variation and... 

Molecular dissimilarity may be defined in many different ways. For the purposes of this presentation, I shall regard it as simply the converse of molecular similarity. High dissimilarity imphes low similarity, and vice versa. 

Again, molecular dissimilarity may be numerically defined and calculated in a variety of ways. Taking the simple definition above, however, the calculation is precisely the same as for molecular similarity. The same numerical value is used to describe similarity and dissimilarity. In applying the concept of molecular similarity in chemicsd information systems at Pfizer Central Research, we use the Tanimoto coefficient as the quantitative similarity measure. This measure, first used for the purpose at Sheffield University, has been widely adopted as the standard measure of molecular similarity, its value varying between 1.0 (identity) and 0.0 (no similarity). We use it, without adaption, as a measure of dissimilarity also. High values implies low similarity, and vice versa. 

Molecular dissimilarity is a powerful concept with many practical applications. In providing access to information on structural variation and diversity, it gives scope for stimulation of creativity and innovation. Facilities for generating and using dissimilarity information should be included within all chemical information... 

---