Search for: All records

Analyses of gene expression of subsurface bacteria and archaea provide insights into their physiological adaptations to in situ subsurface conditions. We examined patterns of expressed genes in hydrothermally heated subseafloor sediments with distinct geochemical and thermal regimes in Guaymas Basin, Gulf of California, Mexico. RNA recovery and cell counts declined with sediment depth, however, we obtained metatranscriptomes from eight sites at depths spanning between 0.8 and 101.9m below seafloor. We describe the metabolic potential of sediment microorganisms, and discuss expressed genes involved in tRNA, mRNA, and rRNA modifications that enable physiological flexibility of bacteria and archaea in the hydrothermal subsurface. Microbial taxa in hydrothermally influenced settings like Guaymas Basin may particularly depend on these catalytic RNA functions since they modulate the activity of cells under elevated temperatures and steep geochemical gradients. Expressed genes for DNA repair, protein maintenance and circadian rhythm were also identified. The concerted interaction of many of these genes may be crucial for microorganisms to survive and to thrive in the Guaymas Basin subsurface biosphere. 

Deep neural networks are powerful tools to detect hidden patterns in data and leverage them to make predictions, but they are not designed to understand uncertainty and estimate reliable probabilities. In particular, they tend to be overconfident. We begin to address this problem in the context of multi-class classification by developing a novel training algorithm producing models with more dependable uncertainty estimates, without sacrificing predictive power. The idea is to mitigate overconfidence by minimizing a loss function, inspired by advances in conformal inference, that quantifies model uncertainty by carefully leveraging hold-out data. Experiments with synthetic and real data demonstrate this method can lead to smaller conformal prediction sets with higher conditional coverage, after exact calibration with hold-out data, compared to state-of-the-art alternatives.