Using chemistry data in Wikidata in AI

Virtual presentation at the #ACSSPRING2025 meeting. Abstract: “The #ACSFall2023 talk “Chemistry in Wikidata visualized with Scholia” gave an overview of the information about chemicals and metabolites in Wikidata, how Scholia was used to provide access to this knowledge, and how open source tools were used to support curation of the knowledge. This update will discuss updates in the chemistry in Wikidata (more external identifiers , more links to literature, more physicochemical data), how to use the increased linking between resources (e.g. how to use SPARQL to find lipids in natural product literature not listed in LIPID MAPS), and how provenance information can support your options to validate AI modelling approaches. This presentation will also introduce shape expressions for chemistry in Wikidata to support quality assurance and consistency. Furthermore, this work will also discuss how the open license (CCZero) and Application Programming Interfaces (APIs) of Wikidata allow data curation of external databases, reuse in custom datasets to enrich their content. Open source code example will document how these tasks are performed.“