Creating an AI based data assistant to bridge genotype to metadata - linking primary clinical data to biobank samples.

Current methods to integrate data require the manual curation and conversion of data as current medical data discovery and data search systems rely on data being standardised. In the field of biobanking data standards are really poorly followed and biobanks tend to have their own terminology sets for both the characteristics of the donor (such as disease and treatments) but also the data surrounding the samples. It is known that in the UK there are millions of samples that remain unused. One of the main reasons for samples not being used is the lack of sufficient clinical annotation, as biobanks struggle to convert textual reports into a standard such as ICD or SNOMED. Projects, such as the UKCRC Tissue Directory and Coordination Centre then struggle to match the needs of the researchers with the availability of samples due to a lack of standardised data available. The result is that 75% of SMEs based in the UK source their samples from outside of the UK which is clearly not beneficial to support inward investment.
In the use case of the UKCRC Tissue Directory and Coordination Centre the purpose of the Directory is to allow researchers to find the biobank that is most likely to have the samples required to support their research. However, the biobanks registering in the Directory are required to do an exact conversion between data terms. For those searching it is important to find a short-list of biobanks that are worth contacting based on an in-depth search. Their preference would be greater depth of queries with the sacrifice of accuracy. To integrate and harmonize such data so as to make them amendable not only as tractable entities and, perhaps more crucially, as vital elements of research hypothesis formation. technical/syntactic interoperability and semantic interoperability needs to be orchestrated across all accessible datasets. The aim of this project is to utlise the UKCRC TDCC and research and define approaches to technical/syntactic interoperability and guidelines building upon those already in place. We have support from BC Platforms to establish an inter-connected set of tools that would allow a researcher to discover, enquire and procure the necessary samples. We aim to research semantic interoperability in terms of how biomedical data can be represented in different content models and coded with different terminology systems. We will utilise the emerging artificial intelligence technologies in the fields of text mining, semantic interoperability, as well as semantic AI so as to develop a health research information environment based on integrated structured and unstructured, clinical and research datasets and amenable to data
mining and analysis documents as input vectors for ML techniques allowing them to potentially discovering novel insights/associations by identifying important classification features of their own accord.