Scientists have developed a new way to hunt for hidden signals of past life, and say it could assist in the search for extraterrestrial organisms on other planets.
Using cutting-edge chemical techniques paired with artificial intelligence, researchers found evidence of ancient life in 3.3-billion-year-old rocks from Earth. They hope the same approach could one day be applied to samples from Mars or icy ocean worlds such as Europa.
The research, published in Proceedings of the National Academy of Sciences, involved analysing more than 400 samples from ancient sediments, fossils, modern plants and animals, fungi, and even meteorites to stress-test the new detection model.
The result? A system capable of distinguishing material left behind by life from non-biological samples with more than 90 per cent accuracy.
“This represents an inspiring example of how modern technology can shine a light on the planet’s most ancient stories and could reshape how we search for ancient life on Earth and other worlds,” said Dr Michael Wong, an astrobiologist and planetary scientist who co-authored the study. “This is a powerful new tool for astrobiology.”
To uncover faint chemical fingerprints left behind by ancient organisms, the team used a technique known as pyrolysis–gas chromatography–mass spectrometry to release molecular fragments from samples.
Those complex chemical patterns were then analysed using a machine-learning model to identify biosignatures that would otherwise be too degraded to interpret.
Co-author Dr Robert Hazen told BBC Science Focus that the technique represented “a paradigm shift” in the field because the algorithm isn’t looking for specific molecules – such as DNA or a lipid – that could be evidence of past life.
Instead, it’s looking at the distribution of what’s present, and whether that hints there could have once been life present.
“For the very first time, we’re just looking for a distribution function,” he said. “That allows us to be much more general when examining highly degraded samples with very little information.”
The oldest biosignature signal detected dated back 3.3 billion years – almost twice as old as the previous limit of around 1.7 billion years.
The team also found molecular evidence that oxygen-producing photosynthesis was occurring at least 2.5 billion years ago, extending the chemical record of photosynthesis by more than 800 million years.
Scientists have previously traced life back 3.5 billion years using two main types of evidence: ancient rock structures created by communities of microbes that grew in sticky, layered ‘mats’ and left behind mound-like formations known as stromatolites; and telltale shifts in the ratios of isotopes in the rocks.
However, suitable samples for such detections are rare. The new machine-learning approach avoids the need for intact fossils or surviving biomolecules, offering a complementary line of evidence that can be applied to a far wider range of rocks.
It also goes beyond a simple life-versus-no-life test. The algorithm can already distinguish photosynthetic from non-photosynthetic organisms, and even separate broad groups of cells known as eukaryotes and prokaryotes.
“It looks at patterns in large numbers of data, and in this case, found very distinct differences between biotics and abiotics,” Hazen said. That ability could prove crucial on Mars, where scientists aren’t sure what life may have looked like, biochemically speaking, if it existed.
If getting samples back from Mars proves too expensive, Hazen believes a rover carrying a suite of instruments could apply the same machine-learning approach directly on the Martian surface. His team has recently received NASA funding to develop such an instrument package.
In the meantime, the team are planning to apply the technique to samples from Earth’s Mars-like deserts, which could help lay the groundwork for future analysis of Martian rocks.
“What’s exciting is that this approach doesn’t rely on finding recognisable fossils or intact biomolecules,” said co-first author Dr Anirudh Prabhu.
“AI didn’t just help us analyse data faster, it allowed us to make sense of messy, degraded chemical data. It opens the door to exploring ancient and alien environments with a fresh lens, guided by patterns we might not even know to look for ourselves.”
The authors cautioned that the model remains complementary to existing techniques and is not a standalone confirmation of life yet, but say it could become a key analytical tool in both terrestrial and planetary science.
“For decades, we’ve searched ancient rocks for traces of life using a limited set of tools,” said co-author and paleobiologist Prof Andrew Knoll.
“What’s remarkable about this study is that it adds whole new dimensions – not just better instruments, but better questions. Machine learning helps us uncover biological signals that were effectively invisible before. It’s a leap forward in our ability to read the deep-time record of life on Earth.”
Read more: