Revisiting the dataflow principle for chemical information processing

Abstract

Dataflow systems, such as Pipeline Pilot or KNIME have become important mainstream tools for data processing in chemistry. These established systems are all implemented relying on a data model emphasizing a strict row/column-centric data table view which does not facilitate interaction with individual chemistry objects, or non-uniform data contents. Resuming our pioneering work which resulted in the implementation of the first dataflow system for chemistry [1], we present in this contribution a different, object-centric approach for the design of re-usable chemical information processing sequences. Our system is based on the metaphor of a factory floor, instead of opaque pipelines. Individual machining stations perform configurable processing steps on objects such as structures, reactions, datasets or tables. Objects are transported between these - or temporarily set on the factory floor for storage or inspection. The combination of this general concept with the extensive scripting functionality of the Cactvs Chemoinformatics toolkit results in a system with capabilities notably different and more flexible than standard pipelining systems.