Dynamic sensor selection for biomarker discovery

Area of Science:

• Systems Biology
• Biotechnology
• Data Science

Background:

• Biotechnologies allow high-resolution biological system monitoring.
• Identifying relevant biomarkers from large datasets is challenging.
• Classical biomarker selection methods struggle with complex biological data.

Purpose of the Study:

• To develop a general methodology for biomarker selection using observability theory.
• To identify biologically meaningful sensors in time-series data.
• To extend biomarker discovery to multiple data modalities and dynamic systems.

Main Methods:

• Application of observability theory for biomarker selection.
• Introduction of dynamic sensor selection to adapt to changing system dynamics.
• Integration of transcriptomics and chromosome conformation data.
• Evaluation using neural activity data (movies, EEG).

Main Results:

• Observability successfully identified biologically meaningful sensors in transcriptomics data.
• Dynamic sensor selection enhanced observability in changing biological regimes.
• The framework demonstrated broad applicability across diverse data types and systems.
• Successful application to agricultural, biomanufacturing, and neural system data.

Conclusions:

• Observability theory provides a robust framework for biomarker discovery.
• The dynamic sensor selection method addresses biological system variability.
• This approach offers a versatile tool for biomarker identification in various scientific fields.