Machine learning helps archaeologists identify the source of ancient poop
Ancient feces can be a gold mine of information. But only if scientists know what left it behind.
For archaeologists, fossilized feces is a valuable tool that can provide useful information about how ancient creatures lived, what they ate, and how healthy they were. But because humans have lived alongside dogs for thousands of years, it can often be difficult to tell who left the feces behind.
"When you find paleo-feces archaeologically it's undergone a lot of changes, it's often been compressed, it's lost much of its colour," anthropologist Christina Warinner told Quirks & Quarks host Bob McDonald. "All those other features that are usually make it obvious if it's human or dog are lost."
So Warinner and her colleague Maxime Borry, of the Max Planck Institute for the Science of Human History, have developed a tool that uses machine learning to scan the feces for gut bacteria as well as remnants of host DNA, to provide a species match within minutes.
To test the system, they applied it to both new and old samples of paleo-feces, some dating back 7,200 years, from archeological sites in Mexico, China, and across Europe. The team discovered that many samples that had previously been attributed to humans, actually came from dogs.
"We have discovered that actually the archeological record is full of dog poop and this is giving us an entirely new window into the relationship between humans and dogs," said Warinner.
"I know that for the average person it might seem a little bit strange, but latrines and trash pits are some of the most interesting things that an archeologist can excavate."
The research was published in the journal PeerJ.
Produced and written by Amanda Buckiewicz