Cell sorting machine

Channels in a glass-and-silicon chip (clear vertical panel) carry cells past microscope lenses (silver and black barrels), which help sort the cells into categories. Credit: Keisuke Goda

Biological techniques

A ‘smart’ system could upend a decades-old method of cell analysis

Artificial intelligence is brought to bear on the complex job of sorting cells into subtypes.

Cell sorting, one of the most widely used procedures in biology, is based on technology invented decades ago. But a new method that separates cells on the basis of their shape and molecular make-up could improve the speed and accuracy of the task, which is used in biomedical tests as well as fundamental research.

Current sorting techniques require researchers to add fluorescent tags to key proteins that distinguish cell types. Automated instruments then sort the cells by detecting the presence of fluorescence.

Keisuke Goda at the University of Tokyo and his colleagues devised a new technique that uses imaging to refine the fluorescent-sorting process. A mixture of fluorescently tagged cells passes under a microscope, which photographs them in 2D. An artificial-intelligence system recognizes patterns in the cells’ physical and chemical structures and either counts or sorts them into tubes, processing up to 100 cells per second.

The researchers plan to adapt their system for more-complex cellular structures, including tissues and whole organisms.