Reaction databases

• There is a very large amount of known reaction data: over 3 million reactions among perhaps 5 million molecular structures. This is hardly surprising, since a big part of chemistry is figuring out how to turn one thing into another then publishing the method.

• Reaction databases live in a universe very close to that of chemical databases. Much of the difference is due to natural content semantics, i.e., chemical databases store information about molecular entities whereas reaction databases store relationships among these entities. Historical orientation is also a large factor, e.g., reaction databases are often designed for purposes such as synthesis planning. An important design aspect of reaction databases is storing and searching reaction atom mapping.

• Some of the larger reaction databases are:
  • Spresi (spresi98rxn, ~2,500,000 reactions)
  • CAS (casreact, ~1,500,000 reactions)
  • Beilstein (~1,000,000 reactions)
  • Current Chemical Reactions (~300,000 archetypical reactions)
  • Teilheimer, Cheminform, etc. (100,000-200,000 reactions)

• The easiest method for integrating between chemical and reaction databases is to cross-index reaction components (reactants, agents and products). Such crossindexing is often provided in the form of paired databases (e.g., Spresi98rxn and Spresi, CCR and Index Chemicus).

Next