• Switches and Kernels Allow for Novelty in Early Development

Most animals are created in a similar way, starting with a fertilized egg that divides into different cell types. Signaling pathways called gene regulatory networks (GRNs) tell cells what to become in a precise process. Changes in this process can mean death or evolution. To find out how an organism can survive change, or novelty, Veronica Hinman, Frederick A. Schwertz Distinguished Professor and head of Biological Sciences, compared the GRNs of a sea urchin and a sea star — two organisms that begin life in an almost identical fashion. The addition of a novel gene, pmar1, signals the embryo to develop into a sea urchin. In a paper published in Nature Communications, Hinman showed that the inclusion of pmar1 created a switch between two stable modes of the Delta-Notch signaling pathway and that each species maintained evolutionarily conserved network motifs, called kernels, that locked down essential elements of development. The switches and kernels may be a common evolutionary mechanism that allows organisms to survive novelty in the early stages of development.

• By Changing Their Shape, Some Bacteria Can Grow More Resilient to Antibiotics

New research led by Assistant Professor of Physics Shiladitya Banerjee demonstrates how certain types of bacteria can adapt to long-term exposure to antibiotics by changing their shape.

Banerjee’s research has focused on the physics behind various cellular processes, with a common theme being that the shape of a cell can have major effects on its reproduction and survival. He and his collaborators decided to dig into how exposure to antibiotics affects the growth and morphologies of the bacterium Caulobacter crescentus.

“Using single-cell experiments and theoretical modeling, we demonstrate that cell shape changes act as a feedback strategy to make bacteria more adaptive to surviving antibiotics,” Banerjee said.

When exposed to less than lethal doses of the antibiotic chloramphenicol over multiple generations, the researchers found that the bacteria dramatically changed their shape by becoming wider and more curved.

• Mathematics Helps Machines Learn from Few Examples

We live in the world where a wealth of data is available about many phenomena. The challenge for researchers is to extract the information they need from a large dataset to answer a specific question. Often, researchers have the information they need for a small fraction of the samples in the dataset, called labeled points. The challenge is to learn the labels for the rest of the sample points, called unlabeled points.

Professor of Mathematical Sciences Dejan Slepčev has worked with a number of students, postdocs and collaborators to develop ways that take advantage of the geometry of the dataset, provided by the unlabeled points, to assign the correct label to as many points as possible. Partial differential equations and applied analysis provide the language to properly model the problem and design solutions.

Slepčev, The University of Minnesota’s Jeff Calder and Ben Cook and former CMU postdoc Matthew Thorpe developed a method and efficient algorithms based on partial differential equations that assign labels based on very few labeled points. In many cases their method outperforms the available approaches for identifying small subsets of information in large datasets. Their findings were published at the International Conference on Machine Learning.