Report:MARPAT on STN/Chemical Structure Tools/Generic Structure Building

Chemical Structure Drawing in MARPAT: Generic Structure Building and Translation Attributes

Building a structure search query in MARPAT is very similar to building a structure query for use in the REGISTRY file on STN, with a few small caveats.

Generic structures are constructed in the same way as specific structures, but introduce variability into the types of molecules and Markush structures that can be included in answer sets. In MARPAT, users most often conduct a substructure search (SSS). When constructing this type of query, leaving open valencies in the structure will effectively put a free bonding site on that atom with open variability*, such that anything could be bonded to that atom in the user's answer set (such as additional branches or even larger fused ring systems). The STN structure drawing tool also has built in variables that can be used to define a variable node within generic structure queries.

*For readers not familiar with chemistry or the concept of valency, a simple example can illustrate this concept. In chemistry, a carbon atom usually has four available bonding sites. In a chemical structure where a carbon atom is shown with two bonds, this usually implies that the other two positions are occupied by hydrogen atoms, which are usually structurally insignificant. However, if a carbon is drawn with only two bonds in STN and the structure is searched via an open substructure search in MARPAT, the system will allow other substitutions at the other two positions, instead of assuming they are occupied by hydrogen atoms.

Leaving an open valency in MARPAT will allow any substitution at that site. However, if the user wishes to allow only certain types of substitution on those sites, variable groups and/or G-groups may be used to limit the structure to certain chemical substitutions. These options are discussed below with emphasis on Markush structure searching. But first, there are a few features that may be used in REGISTRY, but cannot be applied in MARPAT; these features are discussed below.

REGISTRY search features not supported in MARPAT: Ring/Chain Attributes and Screens/Filters

Ring/chain attributes can be applied for both atoms and bonds in the structure searching query. Briefly, applying a ring/chain attribute would allow that portion of the structure query to be part of either a ring or a chain sub-structure as it appears in the answer structure. It is important to note that MARPAT can handle ring/chain attributes as they are applied to individual atoms (nodes) in the query structure, but it cannot handle ring/chain bonds. The reason is that ring/chain bond attributes put too much computational stress on the system. An attempt to load a query having ring/chain bond attributes in MARPAT will result in an error message.

In addition to this limitation, MARPAT will also not support structure filters (also called screens). Structure queries containing these filters may only be uploaded into MARPAT if the filters option is turned off.

The following sections present the options for introducing variability into chemical structure search queries in MARPAT, with an emphasis on how these features should be applied in a generic structure database like MARPAT (as opposed to a specific structure database, like REGISTRY). Match level translation attributes and other generic search tools specific to MARPAT searching are also discussed.