Wearables and -omics data combined with machine learning and mathematical modeling hold great promise for the development of novel methods to quantify circadian processes in humans.
Until now, much emphasis has been on the development of sophisticated algorithms without clearly stating the purpose of the biomarker. Little effort has been devoted to comparison of these biomarkers to gold standards, validating them in realistic protocols, or defining the required accuracy across use cases.
A variety of performance measures have been used in different studies, and reaching a consensus on the performance metrics of biomarkers will facilitate comparison across methods. It will be useful to evaluate these new methods within a framework that is based on concepts developed in circadian rhythm research together with concepts from the field of biomarker development.
