Does This 7,000-Year-Old Poop Belong to a Human or Dog? Machine Learning Has the Answer

Scientists collect so much ancient poop that they designed an AI to tell ancient dog and human feces apart based on preserved DNA.
​Coprolites from China. Image: Jada Ko, courtesy of the Anhui Provincial Institute of Cultural Relics and Archaeology
Coprolites from China. Image: Jada Ko, courtesy of the Anhui Provincial Institute of Cultural Relics and Archaeology

Humans and domestic dogs share a deep bond that dates back at least 12,000 years. Evidence of this intertwined history can be found in places that people and pups simultaneously occupied—and pooped in—according to a study published on Friday in the journal PeerJ.

It is “not uncommon for human and dog feces to co-occur at archaeological sites,” according to the study, which was led by Maxime Borry, a bioinformatics PhD student at the Max Planck Institute for the Science of Human History.


These fossilized stool samples, which are known as coprolites or paleofeces, are valuable time capsules packed with information about past life. They can even preserve DNA from the host (the pooper), as well as from the original food source and the host gut microbiome.

Because dogs often have similar diets to humans, and produced similar-looking stools, it can be difficult to distinguish between human and dog doo-doo from the past. Figuring out “the origin of feces” is further complicated by other behaviors. For instance, dogs sometimes eat human poop and humans sometimes eat dog meat, both of which lead to paleofeces with mixed DNA.

To solve this problem, Borry and his colleagues developed coproID, a bioinformatics method that can predict whether a coprolite was produced by a dog or a human. As outlined in the new paper, coproID uses machine learning software to compare ancient host DNA with DNA in modern feces, enabling it to distinguish between the two hosts in most cases.

“CoproID is the first line of identification in coprolite analysis to confirm that what we’re looking for is actually human, or a dog if we’re interested in dogs,” Borry said in a call.

The team tested out the new method on 20 samples—some of which date back 7,200 years—sourced from 10 archaeological sites in Mexico, China, and across Europe. Thirteen of them were paleofeces and seven were sedimentary samples that were included to make sure coproID could distinguish non-feces from feces.


CoproID successfully identified five of the paleofeces samples as human, two as canine, and flagged the other six coprolites as unknown (it also accurately ruled out the non-feces samples).

Three of the unidentified coprolites didn’t contain enough preserved microbe DNA to make a clear call. The remaining three, which came from a 1,300-year-old site in Mexico, had a gut microbiome that matched humans, but was also puzzlingly high in dog DNA.

“Classified as ‘uncertain,’ there are two possible explanations for these samples,” Borry’s team said in the study. “First, these feces could have originated from a human who consumed a recent meal of canine meat.”

“Alternatively, these feces could have originated from a canine whose microbiome composition is shifted relative to that of the reference metagenomes used in our training set,” the researchers added.

In other words, the feces may be from a dog with a distinct microbiome compared to most modern dogs. A free-ranging canine with a subsistence diet based on urban scavenging or hunting in the wild would probably develop very different microbes in its gut compared to dogs that eat pet food on a regular schedule.

“There is not so much known about the microbiome of dogs,” Borry said. “We are hoping that someone does definitively look at the diversity of the dog gut microbiome because this would allow us to make a much better estimation of whether coprolites are coming from humans or dogs.”

To that point, Borry and his colleagues intend to continue integrating new information into coproID so it can make better identifications about host identities. In addition to distinguishing between species, the method could also help researchers to reconstruct the evolutionary history of the human microbiome over the past several thousand years.

When movies feature archaeologists, they are usually on the prowl for opulent artifacts, lost cities, or the tombs of charismatic ancient leaders. But in real life, one of the most valuable things that an archaeologist can discover is a good old-fashioned turd. Now, scientists have one more way to study these fascinating poops of the past.