SEARCH · Engineering Papers

Results for “SCORE”

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.

SourceResearch connectionFrom yearThrough year

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

At least 847 records · Page 47

All sources NASA NTRS DOE OSTI

DOE OSTI2020

A Numerical Study of Hypoxia in Chesapeake Bay Using an Unstructured Grid Model: Validation and Sensitivity to Bathymetry Representation

A three-dimensional unstructured-grid hydrodynamic and water quality model (Semi-implicit Cross-scale Hydroscience Intergrated System Model-Integrated Compartment Model) is applied successfully for Chesapeake Bay. The model is validated with observations of salinity, chlorophyll-a, dissolved oxygen, nutrients, and phytoplankton productions from the year 1991 to 1995 for the mainstem and some major tributaries, based on multiple model skill scores. Model experiments are conducted to test the importance of having (1) an accurate representation of bathymetry to correctly predict hypoxia and other processes and (2) a high-resolution model grid for tributaries to correctly simulate water quality variables. Further, comparison with the model experiment results with bathymetry smoothing indicates that bathymetry smoothing, as commonly used for many systems, changes the stratification and lateral circulation pattern, resulting in more salt intrusion into shallow water regions, and an increase in the freshwater age. Consequently, a model with bathymetry smoothing can lead to an unrealistic prediction of the distribution of hypoxia and phytoplankton production. Local grid refinement shows significant improvement of model simulations on local stratification and water quality variables. Overall, the use of high-resolution unstructured grid model leads to a faithful representation of the complex geometry, and thus a seamless cross-scale capability for simulating water quality processes in the Bay including tributaries and tidal creeks.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Factors associated with treatment limitations in two Swedish intensive care units: Prevalence and patient involvement

Abstract The aim was to study the prevalence, documentation, and patient involvement in treatment limitations (TLs) in two Swedish intensive care units (ICUs). All patients admitted to the ICUs of two Swedish regional hospitals in 2019 were screened for inclusion. Exclusion criteria included postanesthesia care <24 h. Patients were identified using the Swedish Intensive Care Registry (SIR) and data were extracted from SIR and hospital charts. Uni‐ and multivariable logistic analysis was performed to investigate associations with the presence of TLs. A total of 3090 patients were admitted to the two ICUs in 2019. After exclusion, 1019 patients were included in the study. 45.5% were women and the mean age was 62.9 years. 26.5% of the patients had one or several TLs. Age (OR 1.04 per one year increase 95% confidence interval (CI) 1.02–1.05), SAPS3‐score (OR 1.08 per one unit increase 95% CI 1.06–1.09) and ICU length of stay (OR 1.11 per one day increase 95% CI 1.05–1.17) were independently associated with an increased likelihood of receiving a TL. 17% of the patients were involved in the decision‐making process and in >30% of cases neither the patient nor next‐of‐kin were informed. Women were to a larger extent involved in the decision process than men (24.5 vs. 12.5% p < .05). When the intensivist documented why a TL was established, patient autonomy was four times more commonly stated as the motivation for the TL among women compared to men (15.5% vs. 3.8% p < .05). TLs were common in two Swedish ICUs but a substantial number of patients and next‐of‐kin were not involved in the decision‐making process or informed of the decision. Women were more often than men engaged in the decision to establish a TL.

Jönsson, Nino↗

DOE OSTI2021

A Visual Designer of Layer‐wise Relevance Propagation Models

Abstract Layer‐wise Relevance Propagation (LRP) is an emerging and widely‐used method for interpreting the prediction results of convolutional neural networks (CNN). LRP developers often select and employ different relevance backpropagation rules and parameters, to compute relevance scores on input images. However, there exists no obvious solution to define a “best” LRP model. A satisfied model is highly reliant on pertinent images and designers' goals. We develop a visual model designer, named as VisLRPDesigner, to overcome the challenges in the design and use of LRP models. Various LRP rules are unified into an integrated framework with an intuitive workflow of parameter setup. VisLRPDesigner thus allows users to interactively configure and compare LRP models. It also facilitates relevance‐based visual analysis with two important functions: relevance‐based pixel flipping and neuron ablation. Several use cases illustrate the benefits of VisLRPDesigner. The usability and limitation of the visual designer is evaluated by LRP users.

Huang, Xinyi↗

DOE OSTI2022

Interactively Assessing Disentanglement in GANs

Abstract Generative adversarial networks (GAN) have witnessed tremendous growth in recent years, demonstrating wide applicability in many domains. However, GANs remain notoriously difficult for people to interpret, particularly for modern GANs capable of generating photo‐realistic imagery. In this work we contribute a visual analytics approach for GAN interpretability, where we focus on the analysis and visualization of GAN disentanglement. Disentanglement is concerned with the ability to control content produced by a GAN along a small number of distinct, yet semantic, factors of variation. The goal of our approach is to shed insight on GAN disentanglement, above and beyond coarse summaries, instead permitting a deeper analysis of the data distribution modeled by a GAN. Our visualization allows one to assess a single factor of variation in terms of groupings and trends in the data distribution, where our analysis seeks to relate the learned representation space of GANs with attribute‐based semantic scoring of images produced by GANs. Through use‐cases, we show that our visualization is effective in assessing disentanglement, allowing one to quickly recognize a factor of variation and its overall quality. In addition, we show how our approach can highlight potential dataset biases learned by GANs.

Jeong, Sangwon↗

DOE OSTI2020

Do Programmers Prefer Predictable Expressions in Code?

Source code is a form of human communication, albeit one where the information shared between the programmers reading and writing the code is constrained by the requirement that the code executes correctly. Programming languages are more syntactically constrained than natural languages, but they are also very expressive, allowing a great many different ways to express even very simple computations. Still, code written by developers is highly predictable, and many programming tools have taken advantage of this phenomenon, relying on language model surprisal as a guiding mechanism. Additionally, while surprisal has been validated as a measure of cognitive load in natural language, its relation to human cognitive processes in code is still poorly understood. In this paper, we explore the relationship between surprisal and programmer preference at a small granularity—do programmers prefer more predictable expressions in code? Using meaning-preserving transformations, we produce equivalent alternatives to developer-written code expressions and run a corpus study on Java and Python projects. In general, language models rate the code expressions developers choose to write as more predictable than these transformed alternatives. Then, we perform two human subject studies asking participants to choose between two equivalent snippets of Java code with different surprisal scores (one original and transformed). We find that programmers do prefer more predictable variants, and that stronger language models like the transformer align more often and more consistently with these preferences.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Effects on the Ocular Surface from Reading on Different Smartphone Screens: A Prospective Randomized Controlled Study

The purpose of this study was to investigate the influence of smartphone reading on the ocular surface and to compare the various effects of different screens and light conditions on the ocular surface. One hundred nineteen volunteers were randomly divided into: light + organic light‐emitting diode (OLED), light + electronic ink (eINK), dark + OLED, and dark + eINK. Ocular surface examinations, including noninvasive break‐up time (NIBUT), noninvasive keratograph tear meniscus height (NIKTMH), ocular redness, fluorescein break‐up time (FBUT), corneal fluorescein staining, meibomian gland assessment, Schirmer I Test, and blinking frequency, were performed before and after a reading task. Symptoms were evaluated using the Ocular Surface Disease Index (OSDI) and Computer Vision Syndrome Questionnaire (CVS‐Q). NIBUT and FBUT were decreased statistically significantly after participants read on an OLED screen for 2 hours compared with the baseline in light and dark environments, whereas no statistically significant decrease was observed on an eINK screen. NIKTMH was statistically significantly decreased after reading on an OLED screen in light and dark settings, and the eINK screen had a lesser effect on NIKTMH. An obvious increase in the ocular redness, OSDI and CVS‐Q scores was observed after reading on an OLED screen, whereas the eINK screen had a lesser effect on these indicators. Blink rate increased gradually in OLED subgroups during the reading task, whereas no statistically significant difference was observed in the eINK subgroups. Our research suggested that reading on an OLED screen can cause ocular surface disorder and obvious subjective discomfort, whereas reading on an eINK screen can minimize ocular surface disorder in both dark and light environments.

Yuan, Kelan↗

DOE OSTI2020

Linkage mapping evidence for a syntenic QTL associated with flowering time in perennial C 4 rhizomatous grasses Miscanthus and switchgrass

Flowering in perennial species is directed via complex signalling pathways that adjust to developmental regulations and environmental cues. Synchronized flowering in certain environments is a prerequisite to commercial seed production, and so the elucidation of the genetic architecture of flowering time in Miscanthus and switchgrass could aid breeding in these underdeveloped species. In this context, we assessed a mapping population in Miscanthus and two ecologically diverse switchgrass mapping populations over 3 years from planting. Multiple flowering time quantitative trait loci (QTL) were identified in both species. Remarkably, the most significant Miscanthus and switchgrass QTL proved to be syntenic, located on linkage groups 4 and 2, with logarithm of odds scores of 17.05 and 21.8 respectively. These QTL regions contained three flowering time transcription factors: Squamosa Promoter-binding protein-Like, MADS-box SEPELLATA2 and gibberellin-responsive bHLH137. The former is emerging as a key component of the age-related flowering time pathway.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2025

Comparative Economic Analysis Between Bioenergy and Forage Types of Switchgrass for Sustainable Biofuel Feedstock Production: A Data Envelopment Analysis and Cost–Benefit Analysis Approach

ABSTRACT The capacity to produce switchgrass efficiently and cost‐effectively across diverse environments can be pivotal in achieving the short‐ and medium‐term Sustainable Aviation Fuel targets set by the U.S. Department of Energy. This study evaluated the economic performance of forage‐ and bioenergy‐type switchgrass cultivars and their response to N fertilization under diverse marginal environments across the US Midwest that included Illinois (IL), Iowa (IA), Nebraska (NE), and South Dakota (SD). Data Envelopment Analysis (DEA) was used to evaluate the efficiency of 23 Decision‐Making Units (DMUs)—cultivar types and N fertilization rate combinations—while a cost–benefit analysis calculated their profitability over 5 years. Results showed that two energy‐type cultivars—“Independence” and “Liberty”—were superior economically to the forage cultivars. Independence performed best with the highest profit margin when fertilized at 56 kg N ha −1 , particularly in the US hardiness zone 6a (Urbana, IL). Liberty exhibited the highest profit margins in hardiness zone 5b (Madrid, IA, and Ithaca, NE) at 56 kg N ha −1 and showed exceptional profitability with 28 kg N ha −1 in hardiness zone 6b (Brighton, IL). Switchgrass cultivar “Carthage” showed better efficiency score and profitability results in hardiness zone 4b (South Shore, SD) at 56 kg N ha −1 . The profit trends observed in current study sites may indicate broader patterns across similar US hardiness zones. This study provides valuable insights for decision‐makers to optimize input strategies for biomass production of bioenergy switchgrass to meet renewable energy demands.

Arshad, Muhammad Umer [Department of Crop Sciences↗

DOE OSTI2022

Mechanistic assessment of tolerance to iron deficiency mediated by Trichoderma harzianum in soybean roots

Abstract Aims Iron (Fe) deficiency in soil is a continuing problem for soybean (Glycine max L.) production, partly as a result of continuing climate change. This study elucidates how Trichoderma harzianum strain T22 (TH) mitigates growth retardation associated with Fe-deficiency in a highly sensitive soybean cultivar. Methods and Results Soil TH supplementation led to mycelial colonization and the presence of UAOX1 gene in roots that caused substantial improvement in chlorophyll score, photosynthetic efficiency and morphological parameters, indicating a positive influence on soybean health. Although rhizosphere acidification was found to be a common feature of Fe-deficient soybean, the upregulation of Fe-reductase activity (GmFRO2) and total phenol secretion were two of the mechanisms that substantially increased the Fe availability by TH. Heat-killed TH applied to soil caused no improvement in photosynthetic attributes and Fe-reductase activity, confirming the active role of TH in mitigating Fe-deficiency. Consistent increases in tissue Fe content and increased Fe-transporter (GmIRT1, GmNRAMP2a, GmNRAMP2b and GmNRAMP7) mRNA levels in roots following TH supplementation were observed only under Fe-deprivation. Root cell death, electrolyte leakage, superoxide (O2•–) and hydrogen peroxide (H2O2) substantially declined due to TH in Fe-deprived plants. Further, the elevation of citrate and malate concentration along with the expression of citrate synthase (GmCs) and malate synthase (GmMs) caused by TH suggest improved chelation of Fe in Fe-deficient plants. Results also suggest that TH has a role in triggering antioxidant defence by increasing the activity of glutathione reductase (GR) along with elevated S-metabolites (glutathione and methionine) to stabilize redox status under Fe-deficiency. Conclusions TH increases the availability and mobilization of Fe by inducing Fe-uptake pathways, which appears to help provide resistance to oxidative stress associated with Fe-shortage in soybean. Significance and Impact of the Study These findings indicate that while Fe deficiency does not affect the rate or degree of TH hyphal association in soybean roots, the beneficial effects of TH alone may be Fe deficiency-dependent.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2022

A Novel Anatomic Reconstruction for Posterior Tibialis Tendon in Treatment of Flexible Adult‐Acquired Flatfoot Deformity

Objective To present a novel approach for the anatomic reconstruction of the posterior tibialis tendon (PTT) in restoring plantar insertions and evaluate its efficiency in treating flexible adult‐acquired flatfoot deformity (AAFD) caused by PTT dysfunction. Methods For AAFD treatment, a novel PTT reconstruction method was presented. The current study involved 16 patients, including three men, and 13 women, from August 2017 to July 2019. The mean age was 43.2 ± 15.1 years (21–64 years). The innovative PTT repair method was used on all patients. The treatment involved performing a traditional Flexor Digitorum Longus (FDL) transfer in the navicular tuberosity and suturing the plantar insertions to FDL as tension was applied to tighten the plantar structures of the foot. The results were retrospectively analyzed. The clinical outcome was assessed using the pain visual analogue scale (VAS), the satisfaction VAS, and the American Orthopedic Foot and Ankle Society ankle–hindfoot scale (AOFAS‐AH). Isokinetic testing was performed using a dynamometer at 60°/s and 120°/s for inversion/eversion and plantarflexion/dorsiflexion, respectively, to determine the mean peak torque. Radiographic measurements were employed to assess the outcomes. Results Bone surgeries combined with the modified anatomic PTT reconstruction were performed on patients with medializing calcaneal osteotomy in 12 (75%) patients and subtalar joint fusion in four (25%) patients. The branch linking to the plantar insertions was detected in every case, with an average width of 3.5 ± 0.8 mm (3.1–4.3 mm). All patients were followed up for the mean of 16.8 ± 1.8 months (range, 15–20 months). The average postoperative functional scores, including pain VAS, satisfaction VAS, total AOFAS‐AH, and all AOFAS‐AH sub‐scales, steadily improved during the follow‐up. In the last follow‐up, isokinetic testing revealed no loss of plantarflexion strength ( p = 0.350 and 0.098) and significant improvement in the inversion strength ( p = 0.007 and 0.008) in the operated ankles at 60°/s and 120°/s. Radiographic outcomes, particularly the talar head uncovering, improved significantly after more than a year ( p < 0.001 for all). Conclusions The novel technique for PTT reconstruction in restoring the plantar insertions serves as an effective procedure in treating AAFD caused by PTT dysfunction in terms of delivering a consistent improvement in ankle inversion strength, medial longitudinal arch restoring, and satisfactory clinical outcomes.

Hu, Yong↗

DOE OSTI2022

Sorghum Association Panel whole‐genome sequencing establishes cornerstone resource for dissecting genomic diversity

SUMMARY Association mapping panels represent foundational resources for understanding the genetic basis of phenotypic diversity and serve to advance plant breeding by exploring genetic variation across diverse accessions. We report the whole‐genome sequencing (WGS) of 400 sorghum ( Sorghum bicolor (L.) Moench) accessions from the Sorghum Association Panel (SAP) at an average coverage of 38× (25–72×), enabling the development of a high‐density genomic marker set of 43 983 694 variants including single‐nucleotide polymorphisms (approximately 38 million), insertions/deletions (indels) (approximately 5 million), and copy number variants (CNVs) (approximately 170 000). We observe slightly more deletions among indels and a much higher prevalence of deletions among CNVs compared to insertions. This new marker set enabled the identification of several novel putative genomic associations for plant height and tannin content, which were not identified when using previous lower‐density marker sets. WGS identified and scored variants in 5‐kb bins where available genotyping‐by‐sequencing (GBS) data captured no variants, with half of all bins in the genome falling into this category. The predictive ability of genomic best unbiased linear predictor (GBLUP) models was increased by an average of 30% by using WGS markers rather than GBS markers. We identified 18 selection peaks across subpopulations that formed due to evolutionary divergence during domestication, and we found six F st peaks resulting from comparisons between converted lines and breeding lines within the SAP that were distinct from the peaks associated with historic selection. This population has served and continues to serve as a significant public resource for sorghum research and demonstrates the value of improving upon existing genomic resources.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2024

Population‐level gene expression can repeatedly link genes to functions in maize

SUMMARY Transcriptome‐wide association studies (TWAS) can provide single gene resolution for candidate genes in plants, complementing genome‐wide association studies (GWAS) but efforts in plants have been met with, at best, mixed success. We generated expression data from 693 maize genotypes, measured in a common field experiment, sampled over a 2‐h period to minimize diurnal and environmental effects, using full‐length RNA‐seq to maximize the accurate estimation of transcript abundance. TWAS could identify roughly 10 times as many genes likely to play a role in flowering time regulation as GWAS conducted data from the same experiment. TWAS using mature leaf tissue identified known true‐positive flowering time genes known to act in the shoot apical meristem, and trait data from a new environment enabled the identification of additional flowering time genes without the need for new expression data. eQTL analysis of TWAS‐tagged genes identified at least one additional known maize flowering time gene through trans ‐eQTL interactions. Collectively these results suggest the gene expression resource described here can link genes to functions across different plant phenotypes expressed in a range of tissues and scored in different experiments.

Torres‐Rodríguez, J. Vladimir↗

DOE OSTI2020

The Value of Information From Horizontal Distributed Acoustic Sensing Compared to Multicomponent Geophones Via Machine Learning

Abstract Faults play an important role in recharging many geothermal reservoirs, and seismic information can image the locations of these faults. The value of information (VOI) metric is used to objectively quantify and compare the value of two types of seismic receiver data via a machine learning approach. The demonstrated VOI methodology is novel by including spatial models from seismic data and obtaining the information statistics from machine learning. Our two-dimensional numerical experiments compare images created from sparsely spaced (80 m), two-component geophone sampling to high spatial resolution (1 m), single-component DAS. We used a three-fold cross validation of a U-Net convolutional neural networks to achieve average classification statistics. The results suggest that when horizontal sources are utilized, geophones and DAS identify reflectors and non-reflectors at roughly the same rate. The average F1 score for horizontal DAS is 0.939 and 0.931 for geophones. For images created from a vertical source, DAS performed marginally better (F1 = 0.919) than geophones (F1 = 0.877). Our transferrable methodology can provide guidance on which acquisition scenarios can improve images of important structures in the subsurface and present an efficient method for obtaining reliability statistics from high-dimensional, spatial data.

Energy & Fuels↗

DOE OSTI2021

A Statistical Evaluation of Combining Human Productivity Metrics in the Indoor Environment

The potential of improving human productivity by providing healthy indoor environments has been a consistent interest in the building field for decades. This research field's long-standing challenge is to measure human productivity given the complex nature of office work. Previous studies have diversified productivity metrics, allowing greater flexibility in collecting human data; however, this diversity complicates the ability to combine productivity metrics from disparate studies within a meta-analysis. This study aims to categorize existing productivity metrics and statistically assess which categories show similar behavior when used to measure the impacts of indoor environmental quality. The 106 productivity metrics compiled were grouped into six productivity metric categories: neurobehavioral speed, accuracy, neurobehavioral response time, call handling time, self-reported productivity, and performance score. Then, this study set neurobehavioral speed as the baseline category given its fitness to the efficiency-based definition of productivity (i.e., output versus input) and conducted three statistical analyses with the other categories to evaluate their similarity. As results, the categories of neurobehavioral response time, self-reported productivity, and call handling time were found to have statistical similarity with neurobehavioral speed. This study contributes to creating a constructive research environment for future meta-analyses to understand which human productivity metrics can be combined with each other.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2022

Deep learning categorization of infrasound array data

Here we develop a deep learning-based infrasonic detection and categorization methodology that uses convolutional neural networks with self-attention layers to identify stationary and non-stationary signals in infrasound array processing results. Using features extracted from the coherence and direction-of-arrival information from beamforming at different infrasound arrays, our model more reliably detects signals compared with raw waveform data. Using three infrasound stations maintained as part of the International Monitoring System, we construct an analyst-reviewed data set for model training and evaluation. We construct models using a 4-category framework, a generalized noise vs non-noise detection scheme, and a signal-of-interest (SOI) categorization framework that merges short duration stationary and non-stationary categories into a single SOI category. We evaluate these models using a combination of k-fold cross-validation, comparison with an existing “state-of-the-art” detector, and a transportability analysis. Although results are mixed in distinguishing stationary and non-stationary short duration signals, f-scores for the noise vs non-noise and SOI analyses are consistently above 0.96, implying that deep learning-based infrasonic categorization is a highly accurate means of identifying signals-of-interest in infrasonic data records.

47 OTHER INSTRUMENTATION↗

DOE OSTI2021

Geometric deep learning of RNA structure

RNA molecules adopt three-dimensional structures that are critical to their function and of interest in drug discovery. Few RNA structures are known, however, and predicting them computationally has proven challenging. We introduce a machine learning approach that enables identification of accurate structural models without assumptions about their defining characteristics, despite being trained with only 18 known RNA structures. The resulting scoring function, the Atomic Rotationally Equivariant Scorer (ARES), substantially outperforms previous methods and consistently produces the best results in community-wide blind RNA structure prediction challenges. By learning effectively even from a small amount of data, our approach overcomes a major limitation of standard deep neural networks. Because it uses only atomic coordinates as inputs and incorporates no RNA-specific information, this approach is applicable to diverse problems in structural biology, chemistry, materials science, and beyond.

Townshend, Raphael J. L.↗

DOE OSTI2025

Mutations in the proximal binding site and F-loop of AdeJ confer resistance to efflux pump inhibitors

Multidrug efflux is one of the major mechanisms of antibiotic resistance in gram-negative bacteria. Inhibitors of efflux pumps potentiate the activities of antibiotics, and their discovery could lead to new therapeutic options. The AdeIJK pump in Acinetobacter baumannii is a promising target for efflux pump inhibitors (EPIs) due to its high clinical importance and conservation, and several classes of EPIs targeting this and other A. baumannii efflux transporters have been recently reported. However, the mechanisms of action of these EPIs and their resistance liability remain underexplored. Here, we analyzed the impact of site-specific substitutions in the substrate/EPI translocation path of the inner membrane transporter AdeJ on efflux of substrate antibiotics and fluorescent probes and activities of substituted 4,6-diaminoquinoline EPIs. We found that substitutions in amino acid residues located in the entrance cleft (R701) and the flexible loop (E675) of AdeJ lead to resistance specifically against biphenyl-substituted EPIs, whereas the substitution of F178 in the distal binding pocket increased AdeJ sensitivity to certain naphthyl- and biphenyl-substituted EPIs. No major differences in docking scores and poses of substrates and EPIs were observed between the wild type and corresponding AdeJ variants for any of the mutations considered. This study concludes that substrates and EPIs bound along the translocation path of AdeJ participate in its conformational transitions and can either increase or decrease the rate of transport and therefore the efficiency of EPIs.

Acinetobacter↗

DOE OSTI2023

A bacterial sensor taxonomy across earth ecosystems for machine learning applications

Microbial communities have evolved to colonize all ecosystems of the planet, from the deep sea to the human gut. Microbes survive by sensing, responding, and adapting to immediate environmental cues. This process is driven by signal transduction proteins such as histidine kinases, which use their sensing domains to bind or otherwise detect environmental cues and “transduce” signals to adjust internal processes. We hypothesized that an ecosystem’s unique stimuli leave a sensor “fingerprint,” able to identify and shed insight on ecosystem conditions. To test this, we collected 20,712 publicly available metagenomes from Host-associated, Environmental, and Engineered ecosystems across the globe. We extracted and clustered the collection’s nearly 18M unique sensory domains into 113,712 similar groupings with MMseqs2. We built gradient-boosted decision tree machine learning models and found we could classify the ecosystem type (accuracy: 87%) and predict the levels of different physical parameters (R2 score: 83%) using the sensor cluster abundance as features. Feature importance enables identification of the most predictive sensors to differentiate between ecosystems which can lead to mechanistic interpretations if the sensor domains are well annotated. To demonstrate this, a machine learning model was trained to predict patient’s disease state and used to identify domains related to oxygen sensing present in a healthy gut but missing in patients with abnormal conditions. Moreover, since 98.7% of identified sensor domains are uncharacterized, importance ranking can be used to prioritize sensors to determine what ecosystem function they may be sensing. Furthermore, these new predictive sensors can function as targets for novel sensor engineering with applications in biotechnology, ecosystem maintenance, and medicine.

97 MATHEMATICS AND COMPUTING↗