Unexpected Evidence of Life Found in 3.3-Billion-Year-Old Rocks Using AI
Life on Earth may have appeared much earlier than scientists thought, according to new chemical evidence preserved in rocks more than 3.3 billion years old. An international team led by the Carnegie Institution for Science has discovered molecular signals suggesting that oxygen-producing photosynthesis appeared nearly a billion years earlier than previous records.
The results, published in Proceedings of the National Academy of Sciencesrely on high-resolution chemistry combined with artificial intelligence to detect biological patterns long after their original molecules have disappeared.
“Ancient rocks are full of interesting puzzles that tell us the story of life on Earth, but there are always a few pieces missing,” co-author Katie Maloney said in a press release. “Combining chemical analysis and machine learning has revealed biological clues about ancient life that were previously invisible.”
Learn more: First evidence of Proto Earth may be a chemical imbalance hidden inside ancient rocks
Why early childhood is difficult to detect
The early Earth was home to microbial mats and rarely fossilized simple cells. Over billions of years, these materials were buried, heated, crushed and fractured as the Earth’s crust shifted. These transformations have virtually erased biosignatures that once held clues to the origins and early evolution of life. For this reason, scientists have traditionally only been able to identify reliable molecular traces in rocks less than 1.7 billion years old.
This made it difficult to reconstruct Earth’s early biosphere and to time major events such as the rise of photosynthesis.
The new study calls this limitation into question. This shows that even when the original biomolecules are gone, molecular fragments preserved in ancient rocks may still contain information indicating whether life was once present.
Identifying ancient life using AI
To discover these patterns, the team analyzed organic and inorganic matter from ancient rocks by breaking them down into molecular fragments. The machine learning model was trained on more than 400 samples, including plants, animals, billion-year-old fossils, microbial mats and meteorites, allowing it to detect the chemical fingerprints of life.
Among the samples were exceptionally well-preserved, billion-year-old algae fossils from the Yukon Territory, which helped the AI discover what early photosynthetic organisms looked like in molecular form.
Once trained, the AI system distinguished biological from non-biological chemistry with over 90% accuracy. They also identified molecular signs of photosynthesis in rocks at least 2.5 billion years old, pushing chemical evidence for this process to hundreds of millions of years earlier than previous work and showing that the distribution of degraded molecular fragments can still reveal whether life was once present.
“Ancient life leaves behind more than fossils; it leaves behind chemical echoes,” Dr. Robert Hazen, co-lead author of the study, said in the press release. “Thanks to machine learning, we can now interpret these echoes reliably for the first time.”
In search of life on other worlds
Overall, the work provides a clearer view of Earth’s early biosphere and expands the tools available to study it. And because the method can detect biological chemistry even after billions of years of alteration, it could prove useful far beyond Earth. The same analytical approach could be applied to samples from Mars or other worlds to assess whether they once supported life.
“This innovative technique helps us read the deep-time fossil record in a new way,” Maloney said. “This could help guide the search for life on other planets.”
Learn more: The Earth was formed 4.54 billion years ago – How do scientists know?
Article sources
Our Discovermagazine.com editors use peer-reviewed research and high-quality sources for our articles, and our editors review the articles for scientific accuracy and editorial standards. See the sources used below for this article:
