MIT

Mon 13 Jul 2020, 9:00am
Probability Seminar
Online (see Angel, Murugan or Perkins for link)

Simplicity and Complexity in Belief Propagation I

Online (see Angel, Murugan or Perkins for link)
Mon 13 Jul 2020, 9:00am10:00am
Abstract
There is a very simple algorithm for the inference of posteriors for probability Markov models on trees. Asymptotic properties of this algorithm were first studied in statistical physics and have later played a role in coding theory, in machine learning, and in evolutionary inference, among other areas. The lectures will highlight various phase transitions for this model and their connections to modern statistical inference. Finally we show that, perhaps unexpectedly, this ``simple algorithm" requires complex computation in a number of models.
hide

MIT

Tue 14 Jul 2020, 9:00am
Probability Seminar
Online (see Angel, Murugan or Perkins for link)

Simplicity and Complexity in Belief Propagation II

Online (see Angel, Murugan or Perkins for link)
Tue 14 Jul 2020, 9:00am10:00am
Abstract
There is a very simple algorithm for the inference of posteriors for probability Markov models on trees. Asymptotic properties of this algorithm were first studied in statistical physics and have later played a role in coding theory, in machine learning, and in evolutionary inference, among other areas. The lectures will highlight various phase transitions for this model and their connections to modern statistical inference. Finally we show that, perhaps unexpectedly, this ``simple algorithm" requires complex computation in a number of models.
hide

MIT

Wed 15 Jul 2020, 9:10am
Probability Seminar
Online (see Angel, Murugan or Perkins for link)

Simplicity and Complexity in Belief Propagation III

Online (see Angel, Murugan or Perkins for link)
Wed 15 Jul 2020, 9:10am10:00am
Abstract
There is a very simple algorithm for the inference of posteriors for probability Markov models on trees. Asymptotic properties of this algorithm were first studied in statistical physics and have later played a role in coding theory, in machine learning, and in evolutionary inference, among other areas. The lectures will highlight various phase transitions for this model and their connections to modern statistical inference. Finally we show that, perhaps unexpectedly, this ``simple algorithm" requires complex computation in a number of models.
hide

Vanderbilt University

Wed 15 Jul 2020, 1:45pm
Mathematical Biology Seminar
Zoom

Something's wrong in the (cellular) neighborhood: Mechanisms of epithelial wound detection

Zoom
Wed 15 Jul 2020, 1:45pm2:45pm
Abstract
The first response of epithelial cells to local wounds is a dramatic increase in cytosolic calcium. This increase occurs quickly – calcium floods into damaged cells within 0.1 s, moves into adjacent cells over ~20 s, and appears in a much larger set of surrounding cells via a delayed second expansion over 40300 s – but calcium is nonetheless a reporter: cells must detect wounds even earlier. Using the calcium response as a proxy for wound detection, we have identified an upstream Gproteincoupledreceptor (GPCR) signaling pathway, including the receptor and its chemokine ligand. We present experimental and computational evidence that multiple proteases released during cell lysis/wounding serve as the instructive signal, proteolytically liberating active ligand to diffuse to GPCRs on surrounding epithelial cells. Epithelial wounds are thus detected by the activation of a protease bait. We will discuss the experimental evidence and a corresponding computational model developed to test the plausibility of these hypothesized mechanisms. The model includes calcium currents between each cell’s cytosol and its endoplasmic reticulum (ER), between cytosol and extracellular space, and between the cytosol of neighboring cells. These calcium currents are initiated in the model by cavitationinduced microtears in the plasma membranes of cells near the wound (initial influx), by flow through gap junctions into adjacent cells (first expansion), and by the activation of GPCRs via a proteolytically activated diffusible ligand (second expansion). We will discuss how the model matches experimental observations and makes experimentally testable predictions.
Supported by NIH Grant 1R01GM130130.
hide

Seminar Information Pages
