APPLYING MACHINE LEARNING FOR MOLECULAR BIOSIGNATURES DETECTION

In this poster presentation, we describe machine learning methods for the detection of molecular biosignatures in complex molecular mixtures of natural and synthetic samples analyzed by gas chromatography-mass spectrometry (py-GC-MS). A data pipeline was developed and machine-learning models were trained on three-dimensional chromatographic retention time/mass-to-charge ratio/intensity data to predict the biogenicity of both contemporary and taphonomically altered biological samples. We investigate whether we can discover molecular biosignatures that can differentiate between samples from biotic and abiotic origin in order to test the hypothesis that terrestrial biochemistry differ from abiotic chemistry.