Machine learning interpretable models of cell mechanics from protein images. (https://pubmed.ncbi.nlm.nih.gov/38194965/)

These scientists wanted to understand how cells stick together and move around. They knew that the inside of a cell is made up of tiny parts called molecules, and they wanted to figure out how these molecules work together to make the cell do things.

But there was a problem - scientists didn't have a good way to study the big picture of how cells behave based on their molecules. So, these scientists came up with a new way to learn about how cells work by using a computer program called a neural network.

First, they trained the neural network by showing it pictures of a specific protein inside the cell called zyxin. They also measured the forces that the cell was making. The neural network learned to predict the forces based on the pictures of zyxin.

What was really cool is that the scientists found that just by looking at the pictures of zyxin, they could predict the forces the cell was making, even in situations they had never seen before. This means that the neural network could understand how the cell was behaving just by looking at one protein!

Using this discovery, the scientists made two different models to understand how the cell's forces were made. One model followed the rules of physics, and the other model didn't have any rules - it just learned from the data. Both models showed that the cell's forces were made up of two different length scales.

This study not only helped us understand how cells stick together and move, but it also showed how we can use computer programs to learn more about cells. This is just the beginning of using these programs to understand the biology of cells.

Schmitt MS., Colen J., Sala S., Devany J., Seetharaman S., Caillier A., Gardel ML., Oakes PW., Vitelli V. Machine learning interpretable models of cell mechanics from protein images. Cell. 2024 Jan 18;187(2):481-494.e24. doi: 10.1016/j.cell.2023.11.041. Epub 2024 Jan 8.