We apply data and biology-driven approaches to produce highly performing biosignatures specific for clinically significant disease end-points, in multiple diseases including cancer and diabetes. We employ innovative automated machine learning and text data mining bioinformatic tools to optimise and accelerate discovery. We then validate models in patient cohorts, focusing on liquid biopsies, to offer mature blood-based solutions readily available for minimally invasive diagnostics anticipated to disrupt clinical management. Novel data-driven drug targets are also identified.