Date of Award:

8-2021

Document Type:

Thesis

Degree Name:

Master of Science (MS)

Department:

Computer Science

Committee Chair(s)

Nicholas S. Flann

Committee

Nicholas S. Flann

Committee

Vladimir Kulyukin

Committee

Chad Mano

Abstract

Biological systems that contain multiple living cells exhibit complex self-organization during development of an embryo as individual cells coordinate their behaviors to form intricate patterns. Understanding the mechanisms that underlie this emergent behavior is within reach because of advances in cell imaging that can now track hundreds of cells’ states and positions in real time. However, computational methods are needed that can fit physical scientific models to these observations in space and time. This work introduces a new method to solve this problem that applies automatic differentiation and gradient descent, techniques that underlie deep-learning advances. The method fits a biomechanical model, represented as coupled ordinary differential equations, which describe forces among cells as they self-organize. A preliminary study using synthetic data generated by the model demonstrates the approach’s effectiveness in “reverse engineering” the model from observations of the cells moving. Even in the presence of noise, the method is able to determine the model parameters accurately. Two kinds of emergent behaviors were studied, where differential adhesion causes two distinct cell types to sort into clusters or to form regular mosaic patterns.

Checksum

0a30ccfb8c961f4c50e9217e3693ac2a

Available for download on Saturday, August 01, 2026

Share

COinS