Search NASA⌕ Search

SEARCH · Search NASA

Results for “deep”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 541 records · Page 30

Deep Learning Coordinate-Free Quantum Chemistry

Computing quantum chemical properties of small molecules and polymers can provide insights valuable to physicists, chemists and biologists when designing new materials, catalysts, biological probes and drugs. Deep learning can compute quantum chemical properties accurately in a fraction of the time required by commonly used methods such as density functional theory (DFT). However, many of these deep learning architectures require energy minimized molecular geometries as input, which is also computationally expensive, and decreasing the reproducibility and throughput of these methods. In this study, we demonstrate that accurate quantum chemical computations can be performed without optimized geometries by operating in the coordinate-free domain using deep learning on graph encodings. Furthermore, we also find that the choice of graph-encoding architecture substantially affects the performance of these methods. The Wave architecture outperforms graph convolution architectures, particularly on complex molecules. Furthermore, the structures of these graph encoding architectures provide an opportunity to probe an important, outstanding question in quantum mechanics: What types of quantum chemical properties can be represented by local-variable models? We find that Wave, a local-variable model, is more accurately calculates quantum chemical properties. Graph convolutional architectures require global variables, and are not as effective as as Wave. We anticipate that coordinate-free, deep-learning models of quantum chemistry will become valuable tools in chemistry and biology, enabling researchers to rapidly screen chemical databases or identify new molecules using automated, de-novo design algorithms.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Dual‐Transformer Deep Learning Framework for Seasonal Forecasting of Great Lakes Water Levels

Abstract The Great Lakes of North America form one of the largest freshwater systems on Earth, and their lake‐wide average water levels (lake levels) can fluctuate by more than 0.5 m on a seasonal scale. These fluctuations pose substantial challenges for coastal resilience, flood risk management, and navigation planning. Accurate seasonal forecasting of lake levels using traditional mechanistic models is challenging due to the complex physical mechanisms and coupled hydroclimatic processes involved. Recently, deep learning has gained prominence in geoscience applications for its ability to recognize intricate patterns within multiphysical data sets. Here, we introduce a novel Dual‐Transformer deep learning framework, tested on the Great Lakes. This architecture integrates two modified Transformer models: the Prophet, which predicts underlying trends, and the Critic, which refines the Prophet's predictions. The final lake level prediction is derived by weighting the outputs of both models through a multi‐layer perceptron, jointly trained with the Prophet and Critic to enhance overall accuracy. Our results demonstrate that the innovative learning framework achieves the highest prediction accuracy compared to established deep learning models when using identical input features. It attains a root mean square error of 4–7 cm in predicting lake levels up to 6 months in advance across the lakes. Additionally, the Dual‐Transformer model runs six orders of magnitude faster than conventional mechanistic models, producing results in less than one second on a typical personal computer. These findings suggest that our deep learning framework has strong potential to advance lake level prediction and carries important implications for water management and disaster mitigation, thereby enhancing the quality of life in coastal regions.

Chen, Yi [Great Lakes Research Center Michigan Tec↗

Deep Roots Supply Reactivity and Enhance Silicate Weathering in the Bedrock Vadose Zone

In upland environments, roots commonly extend deep below soil into partially saturated bedrock. This Bedrock Vadose Zone (BVZ) has been shown to store and circulate water, host organic carbon respiration and serve as a critical source of rock-derived nutrients. However, the extent to which deep roots influence chemical weathering rates remains poorly understood. Here, we report 4 years of depth-resolved major ion chemistry over a 16-m thick BVZ hosting a deep rhizosphere in a catchment subject to a Mediterranean climate. These data allow development and validation of a reactive transport model (RTM), revealing that the timescales of water storage and drainage in the BVZ are sufficient to facilitate substantial chemical weathering of the shale bedrock. However, observed solute concentrations are only reproduced by the RTM when we explicitly include measured rates of CO 2(g) production meters below soil driven by the deeply rooted forest. By combining direct observations and a process-based RTM we conclude that the carbon respiration promoted by deep roots significantly enhances chemical weathering rates in the BVZ, constituting 43% ± 3% of total solute flux from the base of the BVZ to the water table.

Osorio‐Leon, Ivan D. [University of Illinois Urban↗

Assessing the Effect of a Deep‐Rooted Grass on Belowground Carbon Storage in Cultivated Land: Insights From a Multi‐Site US Study

Agriculture depletes soil organic carbon (SOC), partly due to the exclusion of deep-rooted perennials. Reintroducing deep-rooted perennials to cultivated land may help to mitigate SOC loss. We quantified the effect of deep roots on SOC by comparing 8 to 30 year-old stands of switchgrass ( Panicum virgatum L .) with paired annual row crop fields at 12 sites across the central and eastern USA. We hypothesized that switchgrass would store more root C and SOC than neighboring shallow-rooted annual crops, and that these effects would extend deeper than 30 cm. We also evaluated whether switchgrass stimulates decomposition of SOC at depth using radiocarbon ( 14 C). Finally, we explored whether the effect of switchgrass on SOC is moderated by soil chemical and physical properties. While the effect of switchgrass on SOC in the surface 100 cm was positive at most sites, the average effect was not statistically significant (difference in SOC = 0.6 kg C m −2 [95% CI −0.8 to +1.9 kg C m −2 ]). By contrast, we found that root C was consistently more abundant under switchgrass, yielding an estimated additional 0.6 kg C m −2 in the surface 100 cm of soil [95% CI +0.5 to +0.7 kg C m −2 ]. 14 C measurements suggested that root C inputs were adding to existing SOC without stimulating decomposition. The effect of switchgrass on belowground C was not strongly related to any of the soil properties that we evaluated. Our observations show that root C can contribute substantially to belowground C stocks when deep-rooted perennials replace shallow-rooted crops.

60 APPLIED LIFE SCIENCES↗

Impact of Penetrating Kelvin Waves on MJO Rainfall Propagation Through Deep Convection Modulation

The Madden‐Julian Oscillation (MJO) is a dominant intraseasonal oscillation in the tropics. As MJOs propagate over the Indian Ocean, some of them can be penetrated by fast‐moving equatorial Kelvin waves originating to the west of the MJOs. Through a series of sensitivity experiments, we demonstrate that these penetrating Kelvin waves can modulate the eastward propagation speed of MJO rainfall. Deep convection in the western MJO rainfall envelope initiates more frequently and produces greater lifetime rainfall when a convectively active Kelvin phase is present, while the opposite occurs during a convectively suppressed Kelvin phase. This alteration in deep convection activity can lead to notable changes in the propagation of the large‐scale MJO rainfall envelope. The modulation of deep convection lifetime rainfall by penetrating Kelvin waves primarily occurs through their impact on deep convection rainfall area, which is closely related to the variations in lower‐tropospheric moisture induced by the Kelvin waves.

Chen, Xingchao [Pennsylvania State Univ., Universi↗

Ecological and genomic profiling of anaerobic methane-oxidizing archaea in a deep granitic environment

Recent single-gene-based surveys of deep continental aquifers demonstrated the widespread occurrence of archaea related to Candidatus Methanoperedens nitroreducens (ANME-2d) known to mediate anaerobic oxidation of methane (AOM). However, it is unclear whether ANME-2d mediates AOM in the deep continental biosphere. We found the dominance of ANME-2d in groundwater enriched in sulfate and methane from a 300-m deep underground borehole in granitic rock. A near-complete genome of one representative species of the ANME-2d obtained from the underground borehole has most of functional genes required for AOM and assimilatory sulfate reduction. The genome of the subsurface ANME-2d is different from those of other members of ANME-2d by lacking functional genes encoding nitrate and nitrite reductases and multiheme cytochromes. In addition, the subsurface ANME-2d genome contains a membrane-bound NiFe hydrogenase gene putatively involved in respiratory H 2 oxidation, which is different from those of other methanotrophic archaea. Short-term incubation of microbial cells collected from the granitic groundwater with 13 C-labeled methane also demonstrates that AOM is linked to microbial sulfate reduction. Finally, given the prominence of granitic continental crust and sulfate and methane in terrestrial subsurface fluids, we conclude that AOM may be widespread in the deep continental biosphere.

59 BASIC BIOLOGICAL SCIENCES↗

Lipid analysis of CO 2 -rich subsurface aquifers suggests an autotrophy-based deep biosphere with lysolipids enriched in CPR bacteria

Sediment-hosted CO 2 -rich aquifers deep below the Colorado Plateau (USA) contain a remarkable diversity of uncultivated microorganisms, including Candidate Phyla Radiation (CPR) bacteria that are putative symbionts unable to synthesize membrane lipids. The origin of organic carbon in these ecosystems is unknown and the source of CPR membrane lipids remains elusive. We collected cells from deep groundwater brought to the surface by eruptions of Crystal Geyser, sequenced the community, and analyzed the whole community lipidome over time. Characteristic stable carbon isotopic compositions of microbial lipids suggest that bacterial and archaeal CO 2 fixation ongoing in the deep subsurface provides organic carbon for the complex communities that reside there. Coupled lipidomic-metagenomic analysis indicates that CPR bacteria lack complete lipid biosynthesis pathways but still possess regular lipid membranes. These lipids may therefore originate from other community members, which also adapt to high in situ pressure by increasing fatty acid unsaturation. An unusually high abundance of lysolipids attributed to CPR bacteria may represent an adaptation to membrane curvature stress induced by their small cell sizes. Our findings provide new insights into the carbon cycle in the deep subsurface and suggest the redistribution of lipids into putative symbionts within this community.

59 BASIC BIOLOGICAL SCIENCES↗

Active anaerobic methane oxidation and sulfur disproportionation in the deep terrestrial subsurface

Abstract Microbial life is widespread in the terrestrial subsurface and present down to several kilometers depth, but the energy sources that fuel metabolism in deep oligotrophic and anoxic environments remain unclear. In the deep crystalline bedrock of the Fennoscandian Shield at Olkiluoto, Finland, opposing gradients of abiotic methane and ancient seawater-derived sulfate create a terrestrial sulfate-methane transition zone (SMTZ). We used chemical and isotopic data coupled to genome-resolved metaproteogenomics to demonstrate active life and, for the first time, provide direct evidence of active anaerobic oxidation of methane (AOM) in a deep terrestrial bedrock. Proteins from Methanoperedens (formerly ANME-2d) are readily identifiable despite the low abundance (≤1%) of this genus and confirm the occurrence of AOM. This finding is supported by 13C-depleted dissolved inorganic carbon. Proteins from Desulfocapsaceae and Desulfurivibrionaceae, in addition to 34S-enriched sulfate, suggest that these organisms use inorganic sulfur compounds as both electron donor and acceptor. Zerovalent sulfur in the groundwater may derive from abiotic rock interactions, or from a non-obligate syntrophy with Methanoperedens, potentially linking methane and sulfur cycles in Olkiluoto groundwater. Finally, putative episymbionts from the candidate phyla radiation (CPR) and DPANN archaea represented a significant diversity in the groundwater (26/84 genomes) with roles in sulfur and carbon cycling. Our results highlight AOM and sulfur disproportionation as active metabolisms and show that methane and sulfur fuel microbial activity in the deep terrestrial subsurface.

59 BASIC BIOLOGICAL SCIENCES↗

Spiraling pathways of global deep waters to the surface of the Southern Ocean

Upwelling of global deep waters to the sea surface in the Southern Ocean closes the global overturning circulation and is fundamentally important for oceanic uptake of carbon and heat, nutrient resupply for sustaining oceanic biological production, and the melt rate of ice shelves. However, the exact pathways and role of topography in Southern Ocean upwelling remain largely unknown. Here in this paper we show detailed upwelling pathways in three dimensions, using hydrographic observations and particle tracking in high-resolution models. The analysis reveals that the northern-sourced deep waters enter the Antarctic Circumpolar Current via southward flow along the boundaries of the three ocean basins, before spiraling southeastward and upward through the Antarctic Circumpolar Current. Upwelling is greatly enhanced at five major topographic features, associated with vigorous mesoscale eddy activity. Deep water reaches the upper ocean predominantly south of the Antarctic Circumpolar Current, with a spatially nonuniform distribution. The timescale for half of the deep water to upwell from 30° S to the mixed layer is ~60–90 years.

58 GEOSCIENCES↗

Scaling deep learning for material imaging with a pseudo 3D model for domain transfer

The recent introduction of deep learning methods for image processing has greatly advanced the characterization of materials using three-dimensional (3D) X-ray imaging techniques. However, deep learning models often have difficulty performing consistently across images owing to unavoidable variations in imaging conditions, which create inconsistencies even for the same material. As a result, networks must frequently be retrained for new datasets, limiting their applicability and generalization. Thus, it is critical to reduce the variations between images to enable a single model to process multiple datasets. Herein, we introduce P3T-Net, a pseudo-3D domain transfer network that transfers diverse 3D images into a uniform domain before processing using deep learning models. Remarkably, P3T-Net enables the reuse of previously trained networks for processing new images and considerably reduces the computational cost of transferring 3D images across domains. These unique capabilities were demonstrated in the following scenarios: (i) image enhancement of fast scans for geological rock and hydrogen fuel cells, (ii) enhancement of images to match the quality of multi-source imaging for lithium-ion batteries, (iii) accurate segmentation of images captured under different conditions, and (iv) tera-scale 3D transfer (10 11 voxels) on a single GPU. Overall, the proposed approach addresses cross-domain inconsistencies across various materials and conditions, thereby enabling more robust and generalizable deep learning solutions for a wide range of material imaging tasks.

25 ENERGY STORAGE↗

Off-the-shelf deep learning is not enough, and requires parsimony, Bayesianity, and causality

Abstract Deep neural networks (‘deep learning’) have emerged as a technology of choice to tackle problems in speech recognition, computer vision, finance, etc. However, adoption of deep learning in physical domains brings substantial challenges stemming from the correlative nature of deep learning methods compared to the causal, hypothesis driven nature of modern science. We argue that the broad adoption of Bayesian methods incorporating prior knowledge, development of solutions with incorporated physical constraints and parsimonious structural descriptors and generative models, and ultimately adoption of causal models, offers a path forward for fundamental and applied research.

97 MATHEMATICS AND COMPUTING↗

Materials Graph Library (MatGL), an open-source graph deep learning library for materials science and chemistry

Graph deep learning models, which incorporate a natural inductive bias for atomic structures, are of immense interest in materials science and chemistry. Here, we introduce the Materials Graph Library (MatGL), an open-source graph deep learning library for materials science and chemistry. Built on top of the popular Deep Graph Library (DGL) and Python Materials Genomics (Pymatgen) packages, MatGL is designed to be an extensible “batteries-included” library for developing advanced model architectures for materials property predictions and interatomic potentials. At present, MatGL has efficient implementations for both invariant and equivariant graph deep learning models, including the Materials 3-body Graph Network (M3GNet), MatErials Graph Network (MEGNet), Crystal Hamiltonian Graph Network (CHGNet), TensorNet and SO3Net architectures. MatGL also provides several pre-trained foundation potentials (FPs) with coverage of the entire periodic table, and property prediction models for out-of-box usage, benchmarking and fine-tuning. Finally, MatGL integrates with PyTorch Lightning to enable efficient model training.

chemistry↗

Substantial and overlooked greenhouse gas emissions from deep Arctic lake sediment

Thermokarst lakes cause abrupt and sustained permafrost degradation and have the potential to release large quantities of ancient carbon to the atmosphere. Despite concerns about how lakes will affect the permafrost carbon feedback, the magnitude of carbon dioxide and methane emissions from deep permafrost soils remains poorly understood. Here we incubated a very deep sediment core (20 m) to constrain the potential productivity of thawed Yedoma and underlying Quaternary sand and gravel deposits. Through radiocarbon dating, sediment incubations and sediment facies classifications, we show that extensive permafrost thaw can occur beneath lakes on timescales of decades to centuries. Although it has been assumed that shallow, aerobic carbon dioxide production will dominate the climate impact of permafrost thaw, we found that anaerobic carbon dioxide and methane production from deep sediments was commensurate with aerobic production on a per gram carbon basis, and had double the global warming potential at warmer temperatures. Carbon release from deep Arctic sediments may thus have a more substantial impact on a changing climate than currently anticipated. These environments are presently overlooked in estimates of the permafrost carbon feedback.

54 ENVIRONMENTAL SCIENCES↗

Viruses control dominant bacteria colonizing the terrestrial deep biosphere after hydraulic fracturing

The deep terrestrial biosphere harbors a significant fraction of earth’s biomass and remains understudied compared to other ecosystems. Deep biosphere life primarily consists of bacteria and archaea, yet knowledge of their co-occurring viruses is poor. Here we temporally catalogued viral diversity from five deep terrestrial subsurface locations (hydraulically fractured wells), examined virus-host interaction dynamics, and experimentally assessed metabolites from cell lysis to better understand viral roles in this ecosystem. We uncovered high viral diversity, rivaling that of peatland soil ecosystems, despite low host diversity. Many viral OTUs were predicted to infect Halanaerobium, the dominant microbe in these ecosystems. Examination of CRISPR-Cas spacers elucidated lineage-specific virus-host dynamics suggesting active in situ viral predation of Halanaerobium. These dynamics indicate an ongoing arms-race with repeated viral encounters and changing viral host range across temporally and geographically distinct shale formations. Laboratory experiments showed that prophage-induced Halanaerobium lysis releases intracellular metabolites that can sustain key fermentative metabolisms, supporting the persistence of microorganisms in this ecosystem. Together these findings suggest that diverse and active viral populations play critical roles in driving strain-level microbial community development and resource turnover within this deep terrestrial subsurface ecosystem.

Daly, Rebecca A.↗

Protein model accuracy estimation empowered by deep learning and inter-residue distance prediction in CASP14

Abstract The inter-residue contact prediction and deep learning showed the promise to improve the estimation of protein model accuracy (EMA) in the 13th Critical Assessment of Protein Structure Prediction (CASP13). To further leverage the improved inter-residue distance predictions to enhance EMA, during the 2020 CASP14 experiment, we integrated several new inter-residue distance features with the existing model quality assessment features in several deep learning methods to predict the quality of protein structural models. According to the evaluation of performance in selecting the best model from the models of CASP14 targets, our three multi-model predictors of estimating model accuracy (MULTICOM-CONSTRUCT, MULTICOM-AI, and MULTICOM-CLUSTER) achieve the averaged loss of 0.073, 0.079, and 0.081, respectively, in terms of the global distance test score (GDT-TS). The three methods are ranked first, second, and third out of all 68 CASP14 predictors. MULTICOM-DEEP, the single-model predictor of estimating model accuracy (EMA), is ranked within top 10 among all the single-model EMA methods according to GDT-TS score loss. The results demonstrate that inter-residue distance features are valuable inputs for deep learning to predict the quality of protein structural models. However, larger training datasets and better ways of leveraging inter-residue distance information are needed to fully explore its potentials.

59 BASIC BIOLOGICAL SCIENCES↗

Survival analysis of localized prostate cancer with deep learning

In recent years, data-driven, deep-learning-based models have shown great promise in medical risk prediction. By utilizing the large-scale Electronic Health Record data found in the U.S. Department of Veterans Affairs, the largest integrated healthcare system in the United States, we have developed an automated, personalized risk prediction model to support the clinical decision-making process for localized prostate cancer patients. This method combines the representative power of deep learning and the analytical interpretability of parametric regression models and can implement both time-dependent and static input data. To collect a comprehensive evaluation of model performances, we calculate time-dependent C-statistics C td over 2-, 5-, and 10-year time horizons using either a composite outcome or prostate cancer mortality as the target event. The composite outcome combines the Prostate-Specific Antigen (PSA) test, metastasis, and prostate cancer mortality. Our longitudinal model Recurrent Deep Survival Machine (RDSM) achieved C td 0.85 (0.83), 0.80 (0.83), and 0.76 (0.81), while the cross-sectional model Deep Survival Machine (DSM) attained C td 0.85 (0.82), 0.80 (0.82), and 0.76 (0.79) for the 2-, 5-, and 10-year composite (mortality) outcomes, respectively. In addition to estimating the survival probability, our method can quantify the uncertainty associated with the prediction. The uncertainty scores show a consistent correlation with the prediction accuracy. We find PSA and prostate cancer stage information are the most important indicators in risk prediction. Our work demonstrates the utility of the data-driven machine learning model in prostate cancer risk prediction, which can play a critical role in the clinical decision system.

60 APPLIED LIFE SCIENCES↗

Deep ocean metagenomes provide insight into the metabolic architecture of bathypelagic microbial communities

The deep sea, the largest ocean’s compartment, drives planetary-scale biogeochemical cycling. Yet, the functional exploration of its microbial communities lags far behind other environments. Here we analyze 58 metagenomes from tropical and subtropical deep oceans to generate the Malaspina Gene Database. Free-living or particle-attached lifestyles drive functional differences in bathypelagic prokaryotic communities, regardless of their biogeography. Ammonia and CO oxidation pathways are enriched in the free-living microbial communities and dissimilatory nitrate reduction to ammonium and H 2 oxidation pathways in the particle-attached, while the Calvin Benson-Bassham cycle is the most prevalent inorganic carbon fixation pathway in both size fractions. Reconstruction of the Malaspina Deep Metagenome-Assembled Genomes reveals unique non-cyanobacterial diazotrophic bacteria and chemolithoautotrophic prokaryotes. The widespread potential to grow both autotrophically and heterotrophically suggests that mixotrophy is an ecologically relevant trait in the deep ocean. These results expand our understanding of the functional microbial structure and metabolic capabilities of the largest Earth aquatic ecosystem.

59 BASIC BIOLOGICAL SCIENCES↗

Deep Koopman operators for causal discovery

Causal discovery aims to identify cause-effect mechanisms for better scientific understanding, explainable decision-making, and more accurate modeling. Standard statistical frameworks, such as Granger causality, lack the ability to quantify causal relationships in nonlinear dynamics due to the presence of complex feedback mechanisms, timescale mixing, and nonstationarity. Thus, applying these methods to study causal dynamics in real-world systems, such as the Earth, is a major challenge. Addressing this shortcoming, we leverage deep learning and a Koopman operator-theoretic formalism to present a class of causal discovery algorithms. Kausal uses deep Koopman operator methods to approximate nonlinear dynamics in a linearized vector space in which traditional causal inference methods such as Granger causality can be more easily applied. Our idealized experiments demonstrate Kausal’s superior ability in discovering and characterizing causal signals compared to existing deep learning and non-deep learning state-of-the-art approaches. Finally, the successful identification of major El Niño and La Niña events in observations showcases Kausal’s skill to handle real-world applications.

54 ENVIRONMENTAL SCIENCES↗