Speaker 1: Zainab Almutawa

Title

Change in Synchronization in Small and Larger Clusters of Pancreatic Heterogenous Beta Cells

Abstract

Beta cells are unparalleled cells in the pancreas that play a significant role in regulating the blood sugar levels by secreting insulin which stimulates the absorption of sugar from peripheral tissues. Beta cells interactions within their cluster of an islet represents an influential role in response to the glucose and in insulin secreting with pulsatile insulin reflecting healthy beta cells. Beta cells are coupled with the closest neighbors in the cluster by coupling with gap junctions.  Gap junctions have a critical role of enhancing the glucose-stimulated insulin secretion, as coupling dysfunction is associated with diabetes.

Our endeavor is to understand how changes in interconnections between linked heterogenous beta cells can lead to changes in cellular synchronization. We initially study a small cluster of three different heterogeneous beta cells, and how a single cell can control the behavior of the network.  We investigate the synchronization of the system through both computational and theoretical analysis and apply our model in the input equivalent oscillators in the triangle configuration. By varying the coupling strength and the heterogeneity of one cell, we demonstrate how ablating or silencing an individual cell has the influence of the behavior of the small network.  We observe in a specific larger network subnetwork de-/synchronization under certain conditions. Ultimately, we aim to find a sufficient condition under which a single cell can disrupt islet synchronization.

Speaker 2: Fred Azizi

Title

An Investigation into the Data Utility and Privacy Trade-off with Distribution-Invariant Privatization

Abstract

Over the past decade, differential privacy has emerged as the standard approach for protecting the privacy of publicly shared data, becoming increasingly essential in our data-driven world. However, applying differential privacy often alters the original data distribution, which can negatively impact the accuracy and reliability of subsequent analyses. This challenge highlights the well-documented trade-off between privacy and utility. In our investigation, we explored the distribution-invariant privatization (DIP) method as a promising solution to effectively address this trade-off. Through simulations and analyses of real-world data, we evaluated the DIP method's ability to maintain robust privacy protection while achieving high levels of statistical accuracy. Additionally, we compared its performance to the Laplace mechanism as a baseline for reference.