Case studies in chemical curation III: complexes and component mixtures

The issue of specified mixtures that include pharmacologically active chemical structures is complex and curatorially challenging, as described by this presentation on the topic.

This post will briefly cover our handing of salts. A substance in an approved drug preparation may not only be specified as a salt in national regulatory documentation but also in the physical  molecular form within the tablet itself.  A well known example is atorvastatin calcium (or more strictly, from the FDA insert label, hemi-calcium trihydrate, CID 656846). For many reasons (primarily our MMOA and activity-mapping focus) we have opted for the stringency of representing “parents” (i.e. not salts)  as primary ligands, as can be seen in our ligand entry.

Database metrics can be used to illustrate the complexities in drug data sets and also supports our approach to simplifying the relationship mappings.  PubChem (June 2014) contains 8072 CIDs with an INN and 5636 with a USAN (the caveat is these tags are from secondary sources, not the authorities themselves).  Of these, INN are 91% singleton “parents”  but only 58% for USANs.  While the pharmaceutical importance of salt forms is clear, to enhance the clarity of our ligand entries we usually align them to the parent INN structures. However, we may include links to pharmacologically-relevant salt forms. If you want to see all mixtures (salts and/or combinations) associated with a ligand you can click on the “Related substances, Mixture” tag in the PubChem CID (the caveat is that for well established drugs these can be large “overkill” lists from patent-only specified salts or mixtures).

One of the issues we face for in vitro activity mapping is that bioassay data is often split between database records of not only parents but also different salt forms specified in different extractions of experimental descriptions.  We make our judgments based on inspecting the original papers but where active compounds are extensively diluted into assay buffers, our inter-publication normalisation to parent is a justifiable simplification.  Those wishing to perform similar experiments would obviously need to check the reference, for example which salt a commercial reagent source may have specified (but sometimes not).

Contributed by Chris Southan and Helen Benson