Fossilized remnants of ancient carbon from the heart of South Africa’s Mpumalanga province have just yielded the earliest chemical evidence yet of life on Earth.
According to a new analysis using machine learning, fragmentary traces of carbon from the Josefsdal Chert, dating back 3.33 billion years, are the earliest and most confident detection of biotic chemistry found on Earth to date.
In addition, the team’s work identified the oldest evidence for photosynthesis to date in rocks 2.52 and 2.3 billion years old, from South Africa and Canada, respectively – pushing back the documented timeline for the process by more than 800 million years.
Related: 1 Billion-Year-Old Fossil Could Be The Oldest Multicellular Animal on Record
“Our results show that ancient life leaves behind more than fossils; it leaves chemical ‘echoes’,” says mineralogist and astrobiologist Robert Hazen of the Carnegie Institution for Science in the US. “Using machine learning, we can now reliably interpret these echoes for the first time.”
Time, decay, and geology are not kind to the traces life leaves behind – and the greater the passage of time, the greater the opportunity for degradation.
In addition, the first life to emerge on Earth would have been tiny microbes, scientists believe, whose physical remnants would have been dramatically altered in the billions of years since they first wiggled around in the primordial damp.
That’s not to say they left no traces. Based on their physical structure, formations such as stromatolites are interpreted as the remains of microbial mats, vast communities of microbes so numerous that they left behind layers in ancient rock. There is also black chert and shale, as well as carbonate formations, within which ancient, fragmentary traces of fossilized carbon have been retained over eons.
It’s difficult to determine with certainty, however, whether these sooty remnants of highly altered carbon were produced by biological or non-biological processes.
Now, a team led by Hazen, in a paper with Carnegie Science astrobiologists Michael Wong and Anirudh Prabhu as first authors, developed a way to positively identify ancient carbon produced by life.
First, they identified specific, subtle patterns unique to biology, left behind by biological molecules, as seen in younger samples. Then, they trained a machine learning algorithm to identify those patterns below the threshold of human discernment.
“Think of it like showing thousands of jigsaw puzzle pieces to a computer and asking whether the original scene was a flower or a meteorite,” Hazen says. “Rather than focus on individual molecules, we looked for chemical patterns, and those patterns could be true elsewhere in the universe.”
Finally, the researchers collected 406 samples of both modern organisms and ancient fossils that ranged from stromatolites to carbon traces in a silica matrix, and subjected them to a technique called pyrolysis-gas chromatography-mass spectrometry (Py-GC-MS).
Py-GC-MS involves heating the sample to break its organic material into fragments, separating those fragments, and measuring their mass signatures.
The machine learning model then pored through the data looking for biotic patterns, returning an accuracy rate of more than 90 percent.
“These samples and the spectral signatures they produce have been studied for decades, but AI offers a powerful new lens that allows us to extract critical information and better understand their nature,” explains Prabhu, an expert in machine learning.
“Even when degradation makes it difficult to spot signs of life, our machine learning models can still detect the subtle traces left behind by ancient biological processes.”
The samples ranged in age from now back to about 3.8 billion years ago, including Greenland carbon from 3.7 billion years ago and 3.5 billion-year-old stromatolites from the Australian desert.
Younger samples – anything under about 500 million years – produced strong, clear biological signatures. But the older the samples grew, the more the biotic signals faded as geological processes stripped away chemical detail.
The oldest sample that returned a positive identification was from the Josefsdal Chert, dating back 3.33 billion years.
That doesn’t mean the older samples aren’t biological. The samples could be so degraded that the pattern is no longer discernible, even to the algorithm. But now, scientists believe, we can positively say that life on Earth had emerged and spread by 3.33 billion years ago, with the possibility that it happened earlier.
“This study represents a major leap forward in our ability to decode Earth’s oldest biological signatures,” Hazen says.
“By pairing powerful chemical analysis with machine learning, we have a way to read molecular ‘ghosts’ left behind by early life that still whisper their secrets after billions of years. Earth’s oldest rocks have stories to tell and we’re just beginning to hear them.”
The research has been published in the Proceedings of the National Academy of Sciences.