In general, the sooner most disorders are identified and treated the better the clinical and functional outcomes. However, early warning signs and symptoms of an emerging disorder may be non-specific and difficult to detect. Blood-based biomarkers may improve early recognition of an emerging disorder and pave the way for the development of early (and possibly preventative) interventions. While no single blood analyte has emerged as diagnostic or predictive of common disorders such as diabetes and cardiovascular disease multianalyte panels of analytes have shown utility. The UK Biobank population cohort provides an opportunity to identify blood-based biomarkers that precede and predict the diagnosis of numerous medical disorders. The cohort includes approximately 500,000 individuals aged 40-69 recruited between 2006-2010. Plasma samples from a subset of 54,306 participants (46,673 randomly selected, 6385 disease-enriched1, 1268 COVID-19 study participants) were analyzed with the Olink proteomics platform providing relative concentrations of 1474 analytes. We propose to utilize a greedy machine learning algorithm, Coarse Approximation Linear Function (CALF) to identify subsets of Olink analytes that predict development of various medical disorders. We propose to utilize a greedy machine learning algorithm, Coarse Approximation Linear Function (CALF), to identify subsets of Olink analytes that predict later diagnosis with one or more medical disorders.
