The first AI-based method for dating ancient genomes has been created

A new study published in Methods for cell reports outlines how artificial intelligence (AI) can support DNA analysis of ancient human remains.

Ancient DNA reveals the history of humanity

Advances in next-generation sequencing (NGS) technologies allow scientists to analyze samples that contain extremely small amounts of DNA. Applying these technologies to ancient samples where DNA is significantly degraded has helped researchers study and understand the history of human evolution and our planet.

A critical aspect of collecting historical events is the ability to date ancient samples. The traditional “gold standard” dating method in archeology is radiocarbon dating.

What is radiocarbon dating?

The premise behind radiocarbon dating is that all living organisms absorb carbon from the environment around them – including stable carbon 12 (12C) and radioactive carbon-14 (14° C). When organisms die, this absorption stops, the level of 12C at the time of death remains, and 14C begins to break down. The rate of this decay can be used as a “clock” to determine when the organism has died. Because atmospheric carbon levels change over time, the method requires reliable historical records of carbon change levels.

Problems with the accuracy of radiocarbon dating – and how AI can offer a solution

Although radiocarbon dating arguably “revolutionized” the field of archaeological science, it is not without its drawbacks, including an accuracy that scientists are working to improve. “Unreliable dating is a major problem, leading to unclear and conflicting results,” says Dr. Eran Elhayk, Associate Professor of Molecular Cell Biology at Lund University.

Elhaik is part of a research team that has developed a new method for dating ancient genomic data using advanced AI technology. The method is called Temporal Population Structure – or TPS – and is an example of Supervised Machine Learning (SML) technology.

“The rationale for TPS is that because most human variation is within continental populations and is subject to processes such as selection and genetic drift that modulate allele frequencies over time, markers exist that show significantly different allele frequencies between different periods , regardless of geography, that can be used to assess temporal trends. We termed these markers temporal informative markers (TIMs),” the authors they write in the post outlining their research.

Changes in allele frequencies—whether through genetic drift or natural selection—create “unique allele combinations” that characterize the historical period in which individuals lived, the research team explained. They call these frequency combinations “time components.”

“Because of their relationship to time, temporal components can be used to transform genomic data into time and predict the age of a sample from genotype data alone,” say Elhaik and colleagues. TPS is trained on the temporal components of thousands of ancient – ​​and modern – genomes and “learns” how to predict their ages.

The research team tested their method by analyzing ~5,000 human remains from the late Mesolithic period (approximately 10,000–8,000 BCE) to the present day. Compared to known sample dates, the dates obtained using TPS correlate with “high accuracy”.

“We show that information about the period in which people lived is encoded in the genetic material. By figuring out how to interpret it and position it in time, we were able to date it with the help of AI,” Elhaik says.

The researchers emphasize that TPS will not eliminate the use of radiocarbon dating, but rather can be used as an additional tool for analysis, especially when there is uncertainty surrounding radiocarbon dating prediction.

“Radiocarbon dating can be very unstable and is affected by the quality of the material being examined. Our method is based on DNA, which makes it very robust. We can now seriously begin to trace the origins of ancient humans and map their migration routes,” concludes Elhayk.

This article is a rework of a press release issued by Lund University. The material has been edited for length and content.

Reference: Behnamian S, Esposito U, Holland G, et al. Temporal population structure, a method for genetic dating of ancient Eurasian genomes from the past 10,000 years. Cell presentation methodp. 2022; 2 (8). doi: 10.1016/j.crmeth.2022.100270.

Leave a Comment