Posted by Researchers have developed a machine-learning method that uses Raman spectroscopy to distinguish biologically formed minerals from those produced by non-biological processes, offering a potential new tool for future searches for life on Mars and other rocky worlds.
The study, published in PNAS Nexus, focuses on apatite – a phosphate mineral found on Earth, Mars and other rocky bodies. Minerals can preserve evidence of biological activity over geological timescales, potentially making them valuable biosignatures when organic molecules or fossils have degraded. The researchers assembled 331 Raman spectra of apatite from biological, geological and synthetic sources and extracted 21 spectral features. A random forest classifier distinguished biotic from abiotic apatite with 96.8% accuracy on an independent test set. A more demanding leave-one-source-out validation, spanning data from 60 independent sources, still achieved 93.8% accuracy, suggesting the results were not simply caused by differences between laboratories or instruments.
Two spectral properties proved especially important: broadening of the main phosphate Raman band and the intensity of a carbonate-related band. The researchers linked these features to reduced crystallinity, structural disorder and carbonate incorporation associated with biological mineral formation. Computational modeling showed that carbonate substitution can roughly double distortions within phosphate structures and make formation of well-ordered apatite energetically less favorable. The team also created a simplified two-feature decision map. About 85.8% of biological apatite samples entered its high-confidence biotic region, while only one abiotic sample was incorrectly placed there. Tests on apatite from three Martian meteorites produced low probabilities of biological origin, while apatite standards measured by Perseverance’s SuperCam also fell within the abiotic region.
The approach could eventually help planetary missions rapidly screen mineral samples for possible signs of ancient life. However, the authors caution that Martian radiation and unusual extraterrestrial mineral-forming environments could complicate interpretation and require broader training datasets.
For more details, read the full article.
