Investigation of Raman Spectroscopic Signatures with Multivariate Statistics: An Approach for Cataloguing Microbial Biosignatures.

Spectroscopic instruments are increasingly being implemented in the search for extraterrestrial life. However, microstructural spectral analyses of alien environments could prove difficult without knowledge on the molecular identification of individual spectral signatures. To bridge this gap, we introduce unsupervised K-means clustering as a statistical approach to discern spectral patterns of biosignatures without prior knowledge of spectral regions of biomolecules. Spectral profiles of bacterial isolates from analogous polar ice sheets were measured with Raman spectroscopy. Raman analysis identified carotenoid and violacein pigments, and key cellular features including saturated and unsaturated fats, triacylglycerols, and proteins. Principal component analysis and targeted spectra integration biplot analysis revealed that the clustering of bacterial isolates was attributed to spectral biosignatures influenced by carotenoid pigments and ratio of unsaturated/saturated fat peaks. Unsupervised K-means clustering highlighted the prevalence of the corresponding spectral peaks, while subsequent supervised permutational multivariate analysis of variance provided statistical validation for spectral differences associated with the identified cellular features. Establishing a validated catalog of spectral signatures of analogous biotic and abiotic materials, in combination with targeted supervised tools, could prove effective at identifying extant biosignatures.