Abstract: The revolution in high-throughput genome-wide and single-cell assays, coupled with powerful computational resources, gives us an opportunity to examine living systems in unprecedented quantitative detail. Yet in almost all of biology, quantitative reasoning has so far failed to achieve the "unreasonable effectiveness" that made mathematics so useful in physics. There are many reasons for this, including the complexity of molecular interaction networks and the fact that living processes are adaptive and nonstationary. But I would argue that a major impediment to making useful and insightful models is the challenge of correctly identifying collective coordinates that meaningfully describe the state of the system. In this talk, I'll discuss some of our work on "biologically-informed" machine learning -- combining existing biological knowledge with data-driven dimension reduction -- to extract models that are not only accurate, but interpretable. I will demonstrate how we can use these methods to characterize and predict adaptation dynamics in living systems (including seasonal adaptation and drug resistance), and pose some provocative questions for the future.
Noyce Conference Room
US Mountain Time
Our campus is closed to the public for this event.
Rosemary BraunAssociate Professor Molecular Biosciences, Applied Math, & Physics at Northwestern University