Scientists Search for Archaic DNA With Machine Learning

COPENHAGEN, DENMARK—According to a statement released by the University of Copenhagen, a team of researchers including Graham Gower and Fernando Racimo has developed a new method to search the modern human genome for genetic material inherited from Neanderthals and Denisovans. The researchers trained a convolutional neural network (CNN), a kind of deep learning framework, to identify patterns in images of known beneficial mutations in modern human genomes. These mutations are thought to have been introduced through mixing with other human species, a process known as adaptive introgression. Some of these mutations may have affected modern human skin development and metabolism, for example. Racimo said that while using the new technique, the team members were able to confirm already discovered mutations, and identify several additional gene variants. These variants may be involved in modern human metabolism, blood cell counts, immunity, tumor suppression, and neurological diseases, the researchers explained. For more, go to "Denisovans at Altitude," one of ARCHAEOLOGY's Top 10 Discoveries of 2019.