When
Student: Woody March-Steinman, Program in Applied Mathematics
Title: Modeling and Machine Learning Approaches for Quantifying Cell Fate Decisions Across Time, Space, and Biological Scales
Advisor: Dr. Andrew Paek, Department of Molecular and Cellular Biology
Location: ENR2 N595 | Zoom (password: Allopatric)
Abstract: Live-cell imaging enables molecular biologists to investigate cellular processes by tracking protein activation and inhibition in individual cells in response to stimuli, providing insight into cellular behavior. Computational and mathematical methods are increasingly necessary to analyze the temporal and spatial data generated by these experiments. These methods can help unravel cell-fate decisions and mechanisms behind tissue organization in high-dimensional, multimodal data.
This work investigates spatiotemporal patterning in patient-derived colonic organoids. A deep learning approach is developed to identify stem-cell regions in organoid monolayers and quantify changes in tissue homeostasis following drug perturbations, including when ground-truth labels are uncertain. The tool supports broader analysis of this two-dimensional organoid system, which retains some physiological relevance while being straightforward to image over time.
A method is also developed to annotate cell death during automated tracking and segmentation. Drug screens and oxidative-stress experiments can cause substantial cell death, which must be annotated to generate accurate single-cell trajectories and censor observations appropriately in survival analyses. An interpretable hybrid approach, consisting of a single-frame classifier followed by a nonstationary hidden Markov model sequence corrector, is applied to annotate cell-death trajectories in single-cell time-series data. The tool can be integrated into existing cell-tracking pipelines and provides uncertainty estimates.
Together, these approaches provide tools to quantify tissue homeostasis and cell death under perturbations, support broader experimentation with colonic organoid monolayers, and enable hypothesis generation for drug discovery and translational research in cancer and colon health.