FAIR-IFICATION OF STRUCTURED CLINICAL DATA [article]

Varsha Gouthamchand, Andre Dekker, Leonard Wee, Johan van Soest
2021 medRxiv   pre-print
One of the common concerns in clinical research is improving the infrastructure to facilitate the reuse of clinical data and deal with interoperability issues. FAIR (Findable, Accessible, Interoperable and Reusable) Data Principles enables reuse of data by providing us with descriptive metadata, explaining what the data represents and where the data can be found. In addition to aiding scholars, FAIR guidelines also help in enhancing the machine-readability of data, making it easier for machine
more » ... lgorithms to find and utilize the data. Hence, the feasibility of accurate interpretation of data is higher and this helps in obtaining maximum results from research work. FAIR-ification is done by embedding knowledge on data. This could be achieved by annotating the data using terminologies and concepts from Web Ontology Language (OWL). By attaching a terminological value, we add semantics to a specific data element, increasing the interoperability and reuse. However, this FAIR-ification of data can be a complicated and a time-consuming process. Our main objective is to disentangle the process of making data FAIR by using both domain and technical expertise. We apply this process in a workflow which involves FAIR-ification of four independent public HNSCC datasets from The Cancer Imaging Archive (TCIA). This approach converts the data from the four datasets into Linked Data using RDF triples, and finally annotates these datasets using standardized terminologies. By annotating them, we link all the four datasets together using their semantics and thus a single query would get the intended information from all the datasets.
doi:10.1101/2021.07.23.21261032 fatcat:p4qdenjjz5er7hc552lkqubhmi