Scientists at the European Molecular Biology Laboratory (EMBL) in Heidelberg have developed software that is claimed to bring machine learning to microscopy.
The so-called Micropilot software learns what researchers are looking for and automatically performs complex microscopy experiments.
It analyses low-resolution images taken by a microscope and, once it has identified a cell or structure the scientists are interested in, it automatically instructs the microscope to start the experiment. This can be recording high-resolution time-lapse video or using lasers to interfere with fluorescently tagged proteins and recording the results.
In four nights of unattended microscope operation, Micropilot is said to have detected 232 cells in two particular stages of cell division and performed a complex imaging experiment on them
The same task undertaken by an experienced microscopist would have taken considerably longer.
Jan Ellenberg and Rainer Pepperkok, whose teams at EMBL designed Micropilot, have used the software to deploy several different microscopy experiments, investigating various aspects of cell division.
They determined when structures known as endoplasmic reticulum exit sites form and uncovered the roles of two proteins, CBX1 and CENP-E, in condensing genetic material into tightly-wound chromosomes and in forming the spindle that helps align those chromosomes.
The Micropilot software, decribed in Nature Methods, is available as open source code at: http://www.embl.de/almf/almf_services/hc_screeing/micropilot/.