ABSTRACT
Decision making during Off-shore hydrocarbon exploration campaigns often relies on the geochemical analysis of piston cores extracted from marine sediments, which need to be sent to specialized laboratories onshore in a logistically complex, expensive and time demanding process. Alternatively, we propose the usage of a new generation of Microbial Exploration Techniques which, based on the combination of NGS and machine learning classifiers, would be able to perform a rapid assessment of the probability of hydrocarbon presence in a piston core solely based on their microbial DNA fingerprint. Specifically, we performed a taxonomical characterization of the microbial populations living in 20 piston cores extracted for standard geochemical analysis during an off-shore campaign. Through multivariate statistics, we have identified a set of bioindicators of hydrocarbon presence which show strong correspondence between biological and geochemical anomalies. In addition, these data have also been used to build a robust machine learning classification model that could be used to classify new samples in a human independent manner.
