The Challenges with Substance Databases and Structure Search Engines

Abstract

Structure connection tables, which are commonly used for the representation of chemical structures in electronic databases, are valuable for substances where specific valence bond structures are known or can be drawn. How-ever there are many classes of substances (for example alloys, catenanes, polymers, or salts) which cannot be fully represented by valence bond structures. There are also issues of definition (such as when a substance is a co-ordination compound, or hydrate, or salt), and of bonding types (resonance, donor complexes, π-complexes). Producers of chemical substance databases may address these issues in different ways and generally need to introduce concepts (for example multicomponent substances) with which chemical scientists may not be familiar. In addition to these aspects of database content, the searcher needs to understand the algorithms behind the structure search engines. For example, the SciFinder search engine has considerable in-built chemical intelligence at the initial search level and then has many tools to mine the data once obtained; the CrossFire search engine also employs several algorithms by default and allows further options to vary them.