Artificial intelligence allows you to look inside any living cell

Institute of Cell Sciences Allen (Seattle) presented his new project to create the first predictive (predictable, probabilistic) 3D model of a living human cell.

The Allen Integrated Cell project combines two technologies that expand our understanding of how induced pluripotent stem cells (hiPSC) can differ. The first technology is a model showing the organization of organelles in specific cells, while the other, predictable, reflects the likely shape and location of organelles in any cell, even unknown to science.

Before creating a 3D model of a cell, scientists "edited" a huge array of human cells, after which they inserted fluorescent protein labels into some of them to highlight specific intracellular structures. Then the scientists filmed tens of thousands of frames of "highlighted" cells and "passed" them through AI, which based on them created a probabilistic model of a living cell.

In the next step, the researchers applied a different machine learning algorithm to the resulting images. Based on information from experiments with "highlighted" cells, AI came to the conclusion that cells that are not labeled with protein labels are practically no different from them. This is very important when conducting such experiments, since fluorescence microscopy is quite toxic and expensive.

The Allen Integrated Cell project is freely available on the Internet. With its help, scientists will be able to better understand intracellular processes, which will help predict and track various negative changes leading to serious, in particular, cancer.