‘Overcoming the Bottleneck’: Knowledge Architectures for Genomic Data Interpretation in Oncology

Abstract In recent years, oncology transitioned from its traditional, organ-based approach to ‘precision oncology’ centered on molecular alterations. As a result, it has become to a significant extent a ‘data-centric’ domain. Its practices increasingly rely on a sophisticated techno-scientific infrastructure that generates massive amounts of data in need of consistent, appropriate interpretations. Attempts to overcome the interpretation bottleneck have led to the establishment of a complex landscape of interrelated resources that, while displaying distinct characteristics and design choices, also entertain horizontal and vertical relations. Although there is no denying that the data-centric nature of contemporary oncology raises a number of key issues related to the production and circulation of data, we suggest that the focus on data use and re-use should be complemented by a focus on interpretation. Oncology practitioners refer to data interpretation resources as ‘knowledgebases’, an actor’s category designed to differentiate them from generic, multi-purpose databases. Their major purpose is the definition and identification of clinically actionable alterations. A heavy investment in human curation, of a clinical rather than exclusively scientific nature is needed to make them valuable, but each knowledgebase appears to have its own peculiar way of connecting clinical and scientific statements. In spite of their common goal, knowledgebases thus adopt very different approaches partly captured by the tension between trust and traceability.