We discuss the concept of recasting the data-rich scientific journal article into two components, a narrative and separate data components, each of which is assigned a persistent digital object identifier. Doing so allows each of these components to exist in an environment optimised for purpose. We make use of a poorly-known feature of the handle system for assigning persistent identifiers which allows an individual datafile from a larger fileset to be retrieved according to its file-name or its MIME type. The data objects allow facile visualisation and retrieval for re-use of the data, and facilitates other operations such as data mining. Examples from five recently published articles illustrate these concepts.