Skip to main content

New cell tool uses AI to look inside human cells

New cell tool uses AI to look inside human cells


This could help us better understand disease

Share this story

GIF courtesy of Allen Institute

Using artificial intelligence, scientists have created a tool that lets anyone visualize what the structures inside a cell look like — even when you only have images of the outside. The Allen Integrated Cell, available for free online, creates 3D visualizations that could help researchers better understand disease.

This cell tool focuses on human stem cells, or cells that haven’t yet turned into, say, muscle cells or heart cells. If we better understand the inner workings of a healthy cell, then we can better understand what goes wrong when it turns into something like a cancer cell, says Greg Johnson, a scientist at the Allen Institute. “This can help us ‘roll back in time’ from a cancer cell to observe the changes that are happening, and detect them as early as possible,” he adds.

Image of cell and its internal structures.
Image of cell and its internal structures.
Image: Allen Institute

First, the scientists genetically engineered cells so that their internal structures (like the mitochondria) glowed. Then, they took thousands of photos of these glowing cells and fed them to machine learning algorithms. The algorithms learned to predict the shape and location of structures in any cell, not just the ones it had already seen or the ones that had various other structures labeled.

This is important because it will let researchers study cells more effectively and cheaply. When it comes to cell imagining, there are a couple different approaches. The simplest and least expensive kind is bright-field imaging. According to Johnson, this is like looking at pond water through a microscope in high school biology. You’ll see a really, really bright image and some dark spots that are internal structures, but there’s not much more detail than that. That makes it hard to really understand the delicate intricacies of the cell.

The other approach is what the Allen scientists did: engineering the cells to make them glow. This method costs far more money, and, of course, the more structures you make glow, the more expensive the process is. Plus, the cells don’t react too well to the dye that makes them glow, meaning that they die quickly and you can’t observe them over time.

Now, the Allen Integrated Cell can look at a bright-field image and then tell the scientists exactly what it looks like inside. “Because we can get these images from just bright-field, it’s low-cost and it won’t be toxic,” says Johnson. “And that means we can take many images and observe the dynamics over a long period of time, which is not something that would have been possible before.”