Evolving BioAssay Ontology (BAO): modularization, integration and applications

The lack of established standards to describe and annotate biological assays and screening outcomes in the domain of drug and chemical probe discovery is a severe limitation to utilize public and proprietary drug screening data to their maximum potential. We have created the BioAssay Ontology (BAO) project (http://bioassayontology.org) to develop common reference metadata terms and definitions required for describing relevant information of low-and high-throughput drug and probe screening assays and results. The main objectives of BAO are to enable effective integration, aggregation, retrieval, and analyses of drug screening data. Since we first released BAO on the BioPortal in 2010 we have considerably expanded and enhanced BAO and we have applied the ontology in several internal and external collaborative projects, for example the BioAssay Research Database (BARD). We describe the evolution of BAO with a design that enables modeling complex assays including profile and panel assays such as those in the Library of Integrated Network-based Cellular Signatures (LINCS). One of the critical questions in evolving BAO is the following: how can we provide a way to efficiently reuse and share among various research projects specific parts of our ontologies without violating the integrity of the ontology and without creating redundancies. This paper provides a comprehensive answer to this question with a description of a methodology for ontology modularization using a layered architecture. Our modularization approach defines several distinct BAO components and separates internal from external modules and domain-level from structural components. This approach facilitates the generation/extraction of derived ontologies (or perspectives) that can suit particular use cases or software applications. We describe the evolution of BAO related to its formal structures, engineering approaches, and content to enable modeling of complex assays and integration with other ontologies and datasets.