Mining the Human Metabolome for Precision Oncology Research

Access to clinical data is critical for advancing translational research; but regulatory constraints and policies surrounding the use of clinical data often challenge data access and sharing. Mixed medical datasets (structured and unstructured) are increasingly dominating the clinical information space, hence, demanding AI-driven techniques such as Natural Language Processing-to reorganize them for effective usage. This paper excavates the HMDB (Human Metabolome Database), for efficient knowledge mining, supported by diversely certified oncology physicians and pharmacists' contributions. We propose a novel taxonomy for knowledge representation and establish a universe of discourse for disease clustering and prediction. Excavated data include metabolites and their respective concentration values, age, gender, as well as gene and protein sequences, of normal and abnormal patients. These data were then merged to form an AI-ready 'Omic' technology datasets. Preliminary results reveal that the proposed AI-ready datasets would aid precision oncology research by adding quality analysis to the present HMDB, and for explaining the variations in concentration values of cancer patients.