Algorithm able to predict viral infections

Researchers working at the University of Zurich (UZH) have managed to develop an artificial neural network that is able to reliably detect when cells have been infected with herpes or adenoviruses. Moreover, it is capable of predicting that acute and severe infections will occur up to 24 hours in advance with 95 percent accuracy. Further details can be found in a press release issued by UZH.

First of all, the research group used changes in the fluorescence of a cell nucleus infected by a virus to their advantage. Using a fluorescence microscope, this process can be visualized, with microscopy images then used to train a deep-learning algorithm, otherwise known as an artificial neural network. This extracts patterns typical of infected or healthy cells. “After training and validation are complete, the neural network automatically detects virus-infected cells”, explains Urs Greber, Professor at the Department of Molecular Life Sciences at UZH, who heads up the research group.

However, this method “not only reliably identifies virus-infected cells, but also accurately detects virulent infections in advance”, according to Greber, who concludes that this development will open up “new ways to better understand infections and to discover new active agents against pathogens such as viruses or bacteria”.

Despite the high level of accuracy in identifying infections, it is not yet clear which features of infected cell nuclei that the artificial neural network is able to recognize. This is the next item to tick off the research team’s ‘to do’ list.