InfoChems Reaction Classification

This reaction classification allows users to browse through hits of reaction searches, thus enabling them to focus on the types of reaction in which they are interested. 

Whereas a model-driven method imposes a rigid classification scheme onto a set of reactions, the data-driven methods try to derive a classification from the data presented. 

Gelemter and Rose [25] used machine learning techniques Chapter IX, Section 1.1 of the Handbook) to analyze the reaction center. Based on the functionalities attached to the reaction center, the method of conceptual clustering derived the features a reaction needed to possess for it to be assigned to a certain reaction type. A drawback of this approach was that it only used topological features, the functional groups at the reaction center, and its immediate environment, and did not consider the physicochemical effects which are so important for determining a reaction mechanism and thus a reaction type. 

HORACE used alternating phases of classification (which topological or physicochemical features are required for a reaction type) and generalization (which features are allowed and can be eliminated) to produce a hierarchical classification of a set of reaction instances. 

The method that was developed builds on computed values of physicochemical effects and uses neural networks for classification. Therefore, for a deeper understanding of this form of reaction classification, later chapters should be consulted on topics such as methods for the calculation of physicochemical effects (Section 7.1) and artificial neural networks (Section 9.4). 

---

Program CLASSIFY InfoChem reaction classification, InfoChem GmbH, Munich, http // www.infochem.de/classify.htm H. Satoh, 0. 

The main software tools provided by InfoChem for the handling of databases of several million structures/reactions comprise the InfoChem Fast Search Engine (ICFSE), the InfoChem Chemistry Cartridge for Oracle (ICCARTR1DGE), the reaction center identification program (TCMAP), and the widely used reaction classification algorithm CLASSIFY . 

Since many of these 2.5 million reactions of the InfoChem Reaction Database are variations of the same type of reaction, InfoChem has developed a sophisticated selection concept based on the identification of all the different individual reaction types included in this file. This so-called Classification Algorithm is also used by MDL Information Systems, San Leandro, CA. Reactions with identical reaction centers and neighboring atoms are considered to be of one reaction type. An analysis of the 2.5 million reactions led to the identification of over 390 000 different reaction types. In the ChemReact database each reaction type is represented by only one selected example reaction. ChemReact was further refined to create ChemSynth, ChemReactlOO, and ChemReact41. 

In order to create subsets of this large reaction file, which can be searched on-line via STN International or with commercial in-house retrieval systems such as REACCS/ISIS from MDL Information Systems, Inc. (USA), a sophisticated algorithm has been developed by InfoChem which identifies the different reaction types in this large database. This InfoChem Classification Algorithm is currently the only available concept for structuring large reaction databases and has, therefore, a high commercial value. 

---