The structures of stem cells can vary wildly, even if they're genetically identical -- and that could be critical to predicting the onset of diseases like cancer. But how do you know what a stem cell will look like until it's already formed? That's where the Allen Institute wants to help: it's launching an online database, the Allen Cell Explorer, where deep learning AI predicts the layout of human stem cells. You only need a pair of identifying structures, like the position of the nucleus, to fill out the rest of the cell's innards.
The team learned how to train its AI thanks to gene editing. After reverting adult cells to stem cells, researchers tagged genes to make cell structures glow and track their layout. This helped identify a clear relationship between the locations of cell structures, making it possible to predict how a stem cell would develop. It was just a matter of teaching the AI to understand this relationship, using real-world examples to verify that it was on the right track.
There are thousands of images at the Cell Explorer right now, and scientists plan to expand it with snapshots of stem cells as they grow into their adult forms. That could pinpoint key processes and help understand how certain diseases develop. If the database proves successful, you could see more effective treatments for conditions that were once relatively mysterious.