Dynamic sensor selection for biomarker discovery.

Recent advances in biotechnologies enable monitoring of biological systems with unprecedented resolution, yet identifying and interpreting biological signals remains a major challenge in clinical and research settings. Classically, biomarkers are measurable indicators of the state of biological processes. Given the large number of molecules in modern datasets, a major challenge is identifying the best biomarkers for a particular setting. Here, we apply observability theory to establish a general methodology for biomarker selection. We demonstrate that observability identifies biologically meaningful sensors in a range of time series transcriptomics data. To address unique biological constraints, we introduce the method of dynamic sensor selection to maximize observability over time, thus enabling observability over regimes where system dynamics themselves are subject to change. Our observability-guided biomarker discovery framework extends to multiple data modalities, as demonstrated with the joint use of transcriptomics and chromosome conformation data. We demonstrate the generality of this approach by evaluating the observability of neural activity measured in movies and electroencephalograms. These applications highlight the broad utility of observability-guided biomarker selection, spanning agriculture, biomanufacturing, and neural systems.