An international team of researchers has demonstrated a new biosignature detection method for distinguishing between abiotic and biotic apatite mineral samples on Mars using instruments already aboard the Perseverance Rover, rather than returning samples to Earth.
The team said their mineral-based biosignature detection approach was successful in identifying samples with nearly 98% accuracy and allows for “deeper time” searches for signs of life, given the longer-lasting nature of minerals compared to biological materials. They also suggest their approach could broaden the capabilities of upcoming rover and life-hunting missions without adding new tools or sensors.
Advantages of a Mineral-Based Approach for Hunting Biosignatures
In a study detailing the team’s technique, Robert M. Hazen from the Carnegie Institution for Science and colleagues explained that searching for life and biosignatures beyond Earth “is a central objective of astrobiology.” They also note that this research is complemented by studies exploring the origin and early evolution of terrestrial life. Still, they add, although identifying biosignatures remains a key challenge, “robust mineral biosignatures remain limited.”
For example, most proposed biosignature missions focus on organic molecules or searching for fossils preserved in sedimentary environments or aquatic environments, including the subsurface oceans of some of the solar system’s moons. However, there are few, if any, proposed missions searching for signs of life preserved in mineral records, which the team notes “can retain records of biological activity over much longer geological timescales”
On Mars, NASA’s Perseverance rover has meticulously collected soil and rock samples deemed most likely to contain biosignatures. However, efforts to send a future sample-return mission to the red planet to retrieve samples are currently stalled and may never be completed.
According to the study authors, there are flight-ready instruments “already deployed on the “Perseverance” rover and planned for deployment on the “Rosalind Franklin” rover capable of searching for signs of life in place rather than returning the samples to labs on Earth. However, they add, “to fully exploit these mission-limited yet information-rich tools, it is essential to efficiently extract as much diagnostic value as possible from the data they generate.”
To search for mineral-based biosignatures, the team focused its research on Raman spectroscopy.
Using Instruments Already Aboard Mars Rovers to Search for Life
“Raman spectroscopy is particularly attractive because it provides bond-specific information on crystal structure and chemical substitution in a nondestructive manner,” they explain.
The team also notes that the Raman spectra acquired by the rover’s instruments “are highly comparable to those obtained under laboratory conditions,” increasing the robustness of any possible life signal.
“This consistency enables direct translation of laboratory reference datasets to planetary environments,” they write.
However, they add, Raman spectra from natural geological samples are often complex, including “overlapping vibrational modes, heterogeneous backgrounds, and chemistry-dependent band shifts.” Until now, this complexity has made interpreting samples based on individual spectral peaks or qualitative inspection inherently challenging.
“Identifying reliable biosignatures that can be detected by flight-ready instruments, such as Raman spectroscopy, remains a central challenge in astrobiology,” the researchers explain. “Developing robust mineralogical indicators of life therefore represents a critical and complementary approach for planetary exploration.”
Approach Distinguishes Abiotic and Biotic Features
To test their approach, the team focused on the mineral apatite. A ubiquitous phosphate, the team notes that apatite can be found “in terrestrial and extraterrestrial environments.” Although apatite is a major component of biological formations such as teeth and bones, it also forms naturally, providing an opportunity to search for biosignatures over much longer timescales if scientists can differentiate the two.
“Here, we integrate Raman spectroscopy with interpretable machine learning to distinguish biotic from abiotic apatite, a ubiquitous phosphate mineral in terrestrial and extraterrestrial environments,” the researchers write.
First, Hazen and colleagues compiled 331 Raman spectra samples. Critically, they included only specimens with known origin attributes: biogenic, abiogenic, or synthetic.
Next, they used this data to extract 21 “band-resolved features.” This was accomplished by training a random forest classifier to identify the “most diagnostic” features of each sample.
“Determining the origin of a sample involves assessing multiple independently varying features of a Raman spectrum, including band positions, widths, and relative intensities, the type of multivariate analysis well suited to machine learning,” the study authors explain.
After some comparative analysis, the team ultimately determined that a sample’s carbonate band and the width of the dominant phosphate band, “reflecting chemical composition and crystal structure, respectively,” were the strongest indicators of biological, natural, or synthetic origin.
After using their method to check the 311 samples, the team said the model successfully distinguished biotic from abiotic apatite “with classification accuracy exceeding 96%.” In fact, the team reported that their mineral-based biosignature detection method misidentified only one biotic apatite sample as abiotic and one abiotic sample as biological.
“Here, we show that biologically formed apatite can be distinguished from abiotic counterparts using interpretable machine learning applied to Raman spectra,” the study authors write.
Life Detection Across Diverse Planetary Environments
When discussing the potential implications of their new mineral-based biosignature detection approach, the team noted that their findings establish a “broadly applicable and mission-relevant strategy for deep-time archives and future planetary missions.” More broadly, the researchers note that their findings highlight the “growing importance of combining flight-ready analytical instruments with AI” to fully leverage datasets already captured by Mars rovers and other “in situ” planetary missions.
“As planetary exploration increasingly shifts toward autonomous decision-making and target selection, the integration of interpretable AI with mineralogical biosignatures offers a powerful and extensible framework,” they write.
Beyond mineral-based biosignature detection strategies, the research team notes that future life hunting missions will “increasingly rely on the joint interpretation of mineralogical and organic spectroscopic signatures.” These include on-board py-GC-MS measurements, infrared spectroscopy, luminescence, laser-induced breakdown spectroscopy, X-ray fluorescence, and “other analytic methods.”
Future approaches will increasingly rely on AI to analyze these disparate datasets, which the authors note will allow for biologically mediated patterns to be distinguished from natural or synthetic backgrounds “in a physically and chemically consistent manner.”
“In this context, such an integrative and interpretable framework provides a pathway toward autonomous, mechanism-informed life detection across diverse planetary environments, advancing the broader goals of planetary science in understanding habitability, biosignature preservation, and the coevolution of geology and biology,” the authors conclude.
The study “Mineral biosignature identification from Raman spectroscopy using machine learning” was published in PNAS Nexus.
Christopher Plain is a Science Fiction and Fantasy novelist and Head Science Writer at The Debrief. Follow and connect with him on X, learn about his books at plainfiction.com, or email him directly at christopher@thedebrief.org.