Researchers Use AI to Discover Evidence of Unknown Human Ancestor In Our DNA

Scientists discovered evidence pointing to a previously unknown ancestor to humans after scouring the human genome using artificial intelligence.

According to a paper published in Nature Communications by a team of researchers in Spain and Estonia, the human ancestor existed around 80,000 years ago. This new evidence could help to explain the genetic links between modern humans and our ancestral cousins, the Denisovans and Neanderthals.

An abundance of evidence suggests that early Homo sapiens interbred with Neanderthals, which left Africa more than 200,000 years earlier than Homo sapiens did. Denisovans were only added to the human lineage in 2008 after the discovery of a pinkie bone and tooth in a Siberian cave, but genetic analysis has shown that interbreeding also occurred between Homo sapiens and Denisovans.

But, the study’s authors note, Homo sapiens breeding with these two other hominid species alone couldn’t account for all of the unexplained genetic remnants found in the modern human genome. A third human ancestor interbreeding with ancient humans seemed plausible, but until quite recently there wasn’t any evidence to support the existence of a third ancestor in the mix.

Last summer, a team of researchers found a bone fragment in Russia that belonged to a child conceived by a Neanderthal mother and Denisovan father. This remarkable finding suggested that not only were Homo sapiens interbreeding with Neanderthals and Denisovans, but these two species were also interbreeding with one another.

This discovery appeared to point to the hypothetical missing third species that could account for the inexplicable parts of the modern human genome. The obstacle geneticists faced was to map not only Neanderthal and Denisovan interbreeding, but also interbreeding between Homo sapiens and a Neanderthal-Denisovan hybrid. According to a statement from the Barcelona Center for Genomic Regulation, one of the three institutions involved in the research, mapping these demographics was “much more complex than anything else considered to date” as far as analyzing human evolution is concerned.

The usual statistical tools used by geneticists simply wouldn’t cut it, and so the researchers turned to deep learning.

Deep learning is a type of machine learning that uses a network architecture loosely modeled on the human brain in order to analyze massive amounts of information for complex patterns. Working backwards with deep learning, the researchers fed the network various demographic models of ancient populations that included this the Neanderthal-Denisovan hybrid until it produced a genome that matched the modern human genome.

This indicated that the missing ancestor to humans was could very well be the byproduct of Neanderthal-Denisovan interbreeding.

“Whenever we run a simulation, we are traveling along a possible path in the history of humankind,” Oscar Lao, a population geneticist at the Center for Genomic Regulation and a co-lead on the study, said in a statement. “Of all simulations, deep learning allows us to observe what makes the ancestral puzzle fit together.”

Researchers still need more evidence to conclusively piece together our ancient history, however. Mayukh Mondal, an evolutionary biologist at the University of Tartu and a co-author of the study, acknowledged in a statement that “as yet we cannot rule out other possibilities.”