AI Uncovers Evidence of Life in 3.3-Billion-Year-Old Rocks

The findings, published Monday in the journal Proceedings of the National Academy of Sciences, include the detection of chemical “fingerprints†left behind by microbes in 3.3-billion-year-old rocks. On top of this, they found chemical signatures of photosynthetic life in rocks as old as 2.5 billion years, extending the chemical record of photosynthesis preserved in carbon molecules by over 800 million years.

“Scientists have developed many different ways to infer life in ancient samples—looking at the textures of rocks, their minerals, the isotopes—but using complex molecules to come up with an unambiguous record of life only extended previously to about 1.6 billion years ago,†co-lead author Michael L. Wong, a research scientist at Carnegie Science’s Earth & Planets Laboratory, told Gizmodo. “We’re taking that all the way to 3.3 [billion], so doubling that age.â€

Harnessing AI for ancient chemical analysis

Wong led the project alongside Anirudh Prabhu, another research scientist at Carnegie Science’s Earth & Planets Laboratory. While Wong specializes in astrobiology and planetary science, Prabhu is an AI and machine learning expert.

To understand how this new model accurately distinguishes biosignatures from abiotic materials, you can think of it like facial-recognition software, Prabhu told Gizmodo.

The model is trained on GC-MS (gas chromatography mass spectrometry) data. This 3D spectral data looks kind of like a landscape, with peaks, valleys, hills, and other features, Prabhu explained. The model identifies patterns among these features that correspond to biological materials, similarly to how facial recognition software is trained to identify the shapes that make up a person’s eyes, mouth, nose, and bone structure.

“We’re looking at the entire data[set], and the model is able to pick out specific features that are very key to a sample being photosynthetic or not—or biogenic or not—in a manner that humans just can’t do because of how vast the data is,†Prabhu explained.

The model is currently able to do this with 90% accuracy, and the researchers hope it will improve as it trains on more data from an increasingly diverse set of samples. This new technique could be a game changer for paleobiologists, allowing them to detect ancient biomarkers even in badly degraded or deformed samples. It’s already opening up a whole new world of opportunity for ancient chemical analysis, and Earth is only just the beginning.

Out-of-this-world possibilities

The search for ancient life extends far beyond our home planet. Astrobiologists like Wong look for evidence of life elsewhere in the solar system, such as on Mars or Saturn’s icy moons.

The fact that the AI was able to accurately detect signs of ancient life on Earth “boosted my confidence that we’re on the right track for developing the kinds of instrumentation and machine learning algorithms that we need to try to find evidence of life in, say, ancient Mars rocks,†Wong said. “I’m full of optimism for the applications elsewhere, beyond Earth.â€

Wong, Prabhu, and their colleagues chose to train the AI on GC-MS data largely because it is a flight-ready instrument. “It has spaceflight heritage, there’s one of these pyrolysis GC-MS instruments sitting in the belly of the Curiosity rover on Mars right now,†Wong said.

The model’s design also prioritizes computational lightweightness and interpretability, which is critical for conducting analyses in real-time as rovers collect geological samples, Prabhu explained.

“So you have a rover on Mars or some other planet, it picks up a sample, zaps it, and produces the spectra. You can quickly get a preliminary prediction—a highly accurate, but preliminary prediction—that scientists can use to understand that area and make decisions,†he said.

Both Wong and Prabhu hope to see this technology applied across the solar system, and they’ll be seeking NASA partnerships to expand its capabilities and ultimately send it to space. For now, the model will continue to deepen our understanding of the emergence of life on Earth, helping us unravel the mysteries of our origin.