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 343 records · Page 19

All sources NASA NTRS DOE OSTI

DOE OSTI2023

Spatio-Temporal Deep Graph Network for Event Detection, Localization, and Classification in Cyber-Physical Electric Distribution System

This work proposes a deep graph learning framework to identify, locate, and classify power, cyber, and cyber power events at the distribution system level. The proposed algorithm jointly exploits spatial, temporal, and node-level cyber and physical data features. The developed graph neural network, together with a deep autoencoder, utilizes physical measurements from distribution level phasor measurement units and cyber data from communication network logs. The spatial structure of the synchrophasor measurements and network is incorporated through a weighted adjacency matrix. The temporal structure is incorporated by defining a spatial operation in the gated recurrent unit. This spatio-temporal learning element resides inside a power event detection, localization, and classification module that provides the degree of confidence for an event label. To accurately pinpoint the location of an event to the nearest bus equipped with a measurement unit, a combination of squared error and proximity score is utilized. Also included is a cyber event detection module that employs heteroskedasticity to analyze the significance of various cyber features during different types of attacks. Finally, a dual-bit cyber-power decision table determines the nature of the event. The proposed method is validated on two distribution systems modeled in OPAL-RT/Hypersim with limited phasor measurement units for different possible physical and cyber events. Further analyses include comparison with other state-of-the-art methods and validation in the presence of measurement noise. As a result, our method outperforms existing approaches and achieves an average detection accuracy of 97.97%, F1-score of 96.88%, precision of 96.53%, and recall of 98.57%.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2021

Predicting nepheline precipitation in waste glasses using ternary submixture model and machine learning

Nepheline precipitation in nuclear waste glasses during vitrification can be detrimental due to its negative effect on chemical durability. Developing models to accurately predict nepheline precipitation from compositions is important to increase waste loading since existing models can be overly conservative. In this study, an expanded dataset containing 955 glasses was compiled from literature data, where 355 glasses are for high-level waste (HLW). Previously developed submixture models were refitted using the new dataset, where a misclassification rate of 7.8% was achieved. Nine machine learning (ML) algorithms (e.g., k-nearest neighbor, Gaussian process regression, artificial neural network, support vector machine, decision tree, etc.) were applied to evaluate their ability of predicting nepheline precipitation from compositions. Model accuracy, precision, recall/sensitivity, and F1 score were systemically compared between different ML algorithms and modeling protocols. Good model prediction with an accuracy ~0.9 (misclassification rate of ~10%) was observed with different algorithms under certain protocol. This study evaluated various ML models to predict nepheline precipitations in waste glasses, highlighting the importance of data preparation, modeling protocol, and their effect on model stability and reproducibility. The results provide insights into applying ML to predict glass properties and suggest areas for future research on modeling nepheline precipitations.

Lu, Xiaonan↗

DOE OSTI2023

Amoeba species colonizing the gills of rainbow trout ( Oncorhynchus mykiss ) in Swiss aquaculture

Nodular gill disease (NGD) is an infectious condition characterized by proliferative gill lesions leading to respiratory problems, oxygen deficiency and mortality in fish. Globally, NGD primarily impacts freshwater salmonids in intensive aquaculture systems. In recent years, numerous outbreaks of severe gill disease have affected more than half of the larger rainbow trout ( Oncorhynchus mykiss ) farms in Switzerland, mainly during spring and early summer. Mortality has reached up to 50% in cases where no treatment was administered. Freshwater amoeba are the presumed aetiologic agent of NGD. The gross gill score (GS) categorising severity of gill pathology is a valuable first-line diagnostic tool aiding fish farmers in identifying and quantifying amoebic gill disease (AGD) in farmed marine salmonids. In this study, the GS was adapted to the NGD outbreak in farmed trout in Switzerland. In addition to scoring disease severity, gill swabs from NGD-affected rainbow trout were sampled and amoeba were cultured from these swabs. Morphologic and molecular methods identified six amoeba strains: Cochliopodium sp., Naegleria sp., Vannella sp., Ripella sp., Saccamoeba sp. and Mycamoeba sp. However, the importance of the different amoeba species for the onset and progression of NGD still has to be evaluated. This paper presents the first description of NGD with associated amoeba infection in farmed rainbow trout in Switzerland.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2025

Illuminated nights accelerate the migration‐linked phenology in a passerine finch redheaded bunting

Abstract The excessive use of artificial light is altering the natural light–dark cycles, consequently impacting animal behavior and physiology. Dim light at night (dLAN) can disrupt migratory patterns, alter hormone levels, and impact breeding success in birds. The present research aims to address the effects of dLAN on the metabolic and reproductive tissues of migratory redheaded bunting ( Emberiza bruniceps ). For this, buntings under short winter‐like days (10L:14D) were exposed to either dark nights ( D = 0.00014 W/m 2 ) or dLAN ( D = 0.058 W/m 2 ), and their locomotor activity, body mass, fat score, food intake, testicular volume, and plasma testosterone levels were measured. The histological architecture of the muscle, intestine, testis, and liver tissues was assessed. Birds exposed to dark nights confined their activity to the daytime only, whereas the dLAN group showed nocturnal activity and initiated Zugunruhe (nighttime restlessness). The body mass, food intake, fat score, and testicular volume significantly increased under dLAN. Histomorphometry revealed increased muscle width, epithelium thickness, and lumen diameter in the testis, proximal and distal muscularis thickness in the intestine, hepatic lipid droplet size, and decreased proximal villi length and intestinal diameter under dLAN. Further, plasma testosterone levels also increased under dLAN. Our results suggest that dLAN can induce migration‐linked phenotypes even under non‐stimulatory short days leading to mistimed seasonal activities.

Tiwari, Jyoti [Department of Zoology University of↗

DOE OSTI2023

A common resequencing‐based genetic marker data set for global maize diversity

SUMMARY Maize ( Zea mays ssp. mays ) populations exhibit vast ranges of genetic and phenotypic diversity. As sequencing costs have declined, an increasing number of projects have sought to measure genetic differences between and within maize populations using whole‐genome resequencing strategies, identifying millions of segregating single‐nucleotide polymorphisms (SNPs) and insertions/deletions (InDels). Unlike older genotyping strategies like microarrays and genotyping by sequencing, resequencing should, in principle, frequently identify and score common genetic variants. However, in practice, different projects frequently employ different analytical pipelines, often employ different reference genome assemblies and consistently filter for minor allele frequency within the study population. This constrains the potential to reuse and remix data on genetic diversity generated from different projects to address new biological questions in new ways. Here, we employ resequencing data from 1276 previously published maize samples and 239 newly resequenced maize samples to generate a single unified marker set of approximately 366 million segregating variants and approximately 46 million high‐confidence variants scored across crop wild relatives, landraces as well as tropical and temperate lines from different breeding eras. We demonstrate that the new variant set provides increased power to identify known causal flowering‐time genes using previously published trait data sets, as well as the potential to track changes in the frequency of functionally distinct alleles across the global distribution of modern maize.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2025

Deep Learning Super-Resolution X-Ray Computed Tomography Algorithms for Additive Manufacturing

Industrial X-ray computed tomography (XCT) is a nondestructive method for inspection and characterization of additively manufactured (AM) materials and parts. In practice, the resolution of XCT can be limited by factors such as detector binning, restricted field of view for large-scale objects, system blur, motion during scanning, and acquisition settings. These limitations can reduce the detectability of critical flaws such as pores, cracks, and lack of fusion. Super-resolution (SR) techniques offer a promising solution for improving the effective resolution and image quality of XCT reconstructions without the need for expensive hardware upgrades or laborious, time-consuming scans. In particular, deep learning-based SR methods have garnered attention in recent years as powerful tools for reconstructing high-resolution volumes from low-resolution inputs. In this work, a novel deep learning-based SR method is proposed for XCT scans of AM parts, and compared against several existing state-of-the-art (SOTA) methods. The proposed method, Simurgh-SR, is built on the pre-existing Simurgh framework and consists of a 2.5D U-Net trained to map low-quality inputs containing noise and artifacts to high-quality reconstructions characterized by higher flaw contrast, better noise texture, and reduced artifacts. The experimental results demonstrate superior performance of Simurgh-SR in performing 4× SR on real industrial XCT scans of thick 316L components, enhancing the structural similarity score and peak signal-to-noise ratio (>7dB) compared to the LR counterpart while improving the F1-score for flaw detection by more than 2.3× when compared to alternative SOTA SR methods. This improvement enables more accurate and significantly faster characterization of metal AM components. Additionally, Simurgh-SR was trained for both 2X and 4X SR and performs effectively at both levels, enabling the use of a single model for various SR factors.

Rahman, Obaid [ORNL] (ORCID:0000000277810840)↗

DOE OSTI2025

Benchmarking large language models for materials synthesis: The case of atomic layer deposition

In this work, we introduce an open-ended question benchmark, ALDbench, to evaluate the performance of large language models (LLMs) in materials synthesis, and, in particular, in the field of atomic layer deposition, a thin film growth technique used in energy applications and microelectronics. Our benchmark comprises questions with a level of difficulty ranging from the graduate level to domain expert current with the state of the art in the field. Human experts reviewed the questions along the criteria of difficulty and specificity, and the model responses along four different criteria: overall quality, specificity, relevance, and accuracy. We ran this benchmark on an instance of OpenAI’s GPT-4o. The responses from the model received a composite quality score of 3.7 on a 1–5 scale, consistent with a passing grade. However, 36% of the questions received at least one below average score. An in-depth analysis of the responses identified at least five instances of suspected hallucination. Finally, we observed statistically significant correlations between the difficulty of the question and the quality of the response, the difficulty of the question and the relevance of the response, the specificity of the question, and the accuracy of the response as graded by the human experts. Furthermore, this emphasizes the need to evaluate LLMs across multiple criteria beyond difficulty or accuracy.

Artificial intelligence↗

DOE OSTI2022

Optimal band selection for target detection with a LWIR multispectral imager

Multispectral imaging can offer many benefits in cost, complexity, resolution, size, weight, and power, relative to hyperspectral imaging. When designing a multispectral system, spectral bandpasses can be selected using optimization algorithms configured to maximally separate target detection scores between target and background regions. A hyperspectral image (HSI) can serve as the source of data from which band groupings can be tested for optimality. The output of an adaptive cosine estimator target detection algorithm is used in an objective function. Three optimization algorithms are compared: particle swarm, dual annealing, and differential evolution. A global optimum is also found using a brute force approach on the Livermore Computing Syrah supercomputer. Three materials are investigated: calcite, gypsum, and limestone. This is done for 3-, 4-, and 5-band systems. The data originate from a longwave infrared HSI of a material display board. The optimization algorithms were run 30 times for every scenario. Performance statistics (maximum, minimum, mean, standard deviation, and median) based on the separation values are given. Additional characterization was performed using receiver operator characteristic (ROC) curves and the area under the ROC curve. While good performance was obtained for the three optimization algorithms, the dual annealing algorithm produced the highest and most consistent detection separation scores on average.

47 OTHER INSTRUMENTATION↗

DOE OSTI2023

A comparison of histopathology imaging comprehension algorithms based on multiple instance learning

Whole slide imaging (WSI), also called digital virtual microscopy, is a new imaging modality. It allows for the application of AI and machine learning methods to cancer pathology to help establish a means for the automatic diagnosis of cancer cases. However, designing machine-learning models for WSI is computationally challenging due to its required ultra-high resolution. The current state-of-the-art models use multiple instance learning (MIL). MIL is a weakly-supervised learning method in which the model uses an array of inferences from many smaller instances to make a final classification about the entire set. In the context of WSI, researchers divide the ultra-high-resolution image into many patches. The model then classifies the slide based on an array of inferences from the patches. Among several ways of making the final classification, attention-based mechanisms have resulted in superb accuracy scores. The Transformer, one attention-based algorithm, has reported substantial improvements for WSI comprehension tasks. In this project, we studied and compared several WSI comprehension algorithms. We used the following three datasets: CAMELYON16+17, TCGALung, and TCGA-Kidney. We found that attention-based MIL algorithms performed better than standard MIL algorithms for classifying WSI images, achieving a higher mean accuracy and AUC. However, none of the attention-based algorithms performed significantly better than the others, reporting accuracy scores that varied widely. Presumably, it is due to the limited availability of training samples in the data corpus. Since it is not easy to increase the samples from human subjects, some machine learning techniques like transfer learning could help mitigate this issue.

Saunders, Adam↗

DOE OSTI2022

Direction-optimizing Label Propagation Framework for Structure Detection in Graphs: Design, Implementation, and Experimental Analysis

Label Propagation is not only a well-known machine learning algorithm for classification but also an effective method for discovering communities and connected components in networks. We propose a new Direction-optimizing Label Propagation Algorithm (DOLPA) framework that enhances the performance of the standard Label Propagation Algorithm (LPA), increases its scalability, and extends its versatility and application scope. As a central feature, the DOLPA framework relies on the use of frontiers and alternates between label push and label pull operations to attain high performance. It is formulated in such a way that the same basic algorithm can be used for finding communities or connected components in graphs by only changing the objective function used. Additionally, DOLPA has parameters for tuning the processing order of vertices in a graph to reduce the number of edges visited and improve the quality of solution obtained. We present the design and implementation of the enhanced algorithm as well as our shared-memory parallelization of it using OpenMP. We also present an extensive experimental evaluation of our implementations using the LFR benchmark and real-world networks drawn from various domains. Compared with an implementation of LPA for community detection available in a widely used network analysis software, we achieve at most five times the F-Score while maintaining similar runtime for graphs with overlapping communities. We also compare DOLPA against an implementation of the Louvain method for community detection using the same LFR-graphs and show that DOLPA achieves about three times the F-Score at just 10% of the runtime. For connected component decomposition, our algorithm achieves orders of magnitude speedups over the basic LP-based algorithm on large-diameter graphs, up to 13.2× speedup over the Shiloach-Vishkin algorithm, and up to 1.6× speedup over Afforest on an Intel Xeon processor using 40 threads.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

On the Abuse and Detection of Polyglot Files

A polyglot is a file that is valid in two or more formats. Polyglot files pose a problem for file-upload and generative AI web interfaces that rely on format identification to determine how to securely handle incoming files. In this work we found that existing file-format and embedded-file detection tools, even those developed specifically for polyglot files, fail to reliably detect polyglot files used in the wild. To address this issue, we studied the use of polyglot files by malicious actors in the wild, finding 30 polyglot samples and 15 attack chains that leveraged polyglot files. Using knowledge from our survey of polyglot usage in the wild---the first of its kind---we created a novel data set based on adversary techniques. We then trained a machine learning detection solution, PolyConv, using this data set. PolyConv achieves a precision-recall area-under-curve score of 0.999 with an F1 score of 99.20% for polyglot detection and 99.47% for file-format identification, significantly outperforming all other tools tested. We developed a content disarmament and reconstruction tool, ImSan, that successfully sanitized 100% of the tested image-based polyglots, which were the most common type found via the survey. Our work provides concrete tools and suggestions to enable defenders to better defend themselves against polyglot files, as well as directions for future work to create more robust file specifications and methods of disarmament.

Oesch, T [ORNL] (ORCID:0000000269091022)↗

DOE OSTI2025

Preliminary Study on Fine-Grained Power and Energy Measurements on Grace Hopper GH200 with Open-Source Performance Tools

The increasing adoption of tightly integrated, heterogeneous architectures, combined with the slowdown of Moore’s law, has made application power and energy-driven optimizations critical to efficiently use high-performance computing systems. This paper introduces a newly developed open-source toolkit that seamlessly integrates the Linux real-time hardware monitoring program hwmon with the Performance Application Programming Interface and the Score-P performance measurement system, thereby enabling fine-grained power and energy measurements for high-performance computing applications. Our primary target platform is the Wombat test bed, which is a system based on the NVIDIA GH200 superchip. The toolkit can capture transient power peaks with high temporal resolution (50 ms) and, thanks to Score-P integration, can map power metrics to specific code regions, thereby providing actionable information on power-intensive operations and inefficiencies. The toolkit also provides a holistic view of both the power and the energy consumption of the entire GH200 superchip by covering all major components: the Grace CPU, the Hopper GPU, and the I/O subsystem. Experiments that use Locally Self-consistent Multiple Scattering, which is an application for first-principles calculations of materials developed at Oak Ridge National Laboratory, have demonstrated the tool’s ability to identify transient power spikes and uncover opportunities for energy-aware optimizations. Additionally, we introduce a Python-based utility for converting Open Trace Format 2 traces to Parquet format, thus enabling advanced data analysis for numerical integration methods applied to power data for accurate energy profiling.

Hernandez Mendoza, Oscar [ORNL] (ORCID:00000002538↗

DOE OSTI2026

SetGo: Metadata Readiness for Scientific AI Datasets

Scientific datasets intended for AI use require both computational readiness for model training and metadata readiness for discovery, sharing, and reuse. The Readiness Engine for Data Integration (REDI) addresses computational readiness, but no corresponding tool evaluates whether a dataset’s metadata are sufficiently complete, governed, and standards-compliant for publication and agent-based consumption. Existing FAIR assessors operate only on published repository records, and no single system covers FAIR compliance, licensing, provenance, governance, reproducibility, and catalog readiness together. We present SetGo, an open-source Python toolkit that assesses and repairs metadata readiness across these six dimensions before a dataset is published or archived. Applied to four scientific corpora, SetGo surfaces deficiencies that general-purpose tools do not detect: ERA5 climate metadata scores 4% on ACDD 1.3 compliance; materials datasets fail OPTIMADE species-definition requirements; and PDB-derived proteomics data carries licensing terms incompatible with standard SPDX identifiers. Guided enrichment raises overall FAIR scores from 52–57% to 81–91%, and a single setgo publish command pushes to Hugging Face Hub, CKAN, or OpenMetadata with ML Commons Croissant 1.0 metadata sidecars. To support interactive and automated workflows, SetGo integrates with coding agents powered by large language models (LLMs) through a /setgo skill that enables natural-language execution of the full assess–enrich–publish loop, with user involvement limited to supplying missing metadata values.

Wilkinson, Sean [ORNL] (ORCID:0000000214437479)↗

DOE OSTI2020

Inner Engineering Practices and Advanced 4-day Isha Yoga Retreat Are Associated with Cannabimimetic Effects with Increased Endocannabinoids and Short-Term and Sustained Improvement in Mental Health: A Prospective Observational Study of Meditators

Anxiety and depression are common in the modern world, and there is growing demand for alternative therapies such as meditation. Meditation can decrease perceived stress and increase general well-being, although the physiological mechanism is not well-characterized. Endocannabinoids (eCBs), lipid mediators associated with enhanced mood and reduced anxiety/depression, have not been previously studied as biomarkers of meditation effects. Our aim was to assess biomarkers (eCBs and brain-derived neurotrophic factor [BDNF]) and psychological parameters after a meditation retreat. Methods. This was an observational pilot study of adults before and after the 4-day Isha Yoga Bhava Spandana Program retreat. Participants completed online surveys (before and after retreat, and 1 month later) to assess anxiety, depression, focus, well-being, and happiness through validated psychological scales. Voluntary blood sampling for biomarker studies was done before and within a day after the retreat. The biomarkers anandamide, 2-arachidonoylglycerol (2-AG), 1-arachidonoylglycerol (1-AG), docosatetraenoylethanolamide (DEA), oleoylethanolamide (OLA), and BDNF were evaluated. Primary outcomes were changes in psychological scales, as well as changes in eCBs and BDNF. Results. Depression and anxiety scores decreased while focus, happiness, and positive well-being scores increased immediately after retreat from their baseline values (P < 0.001). All improvements were sustained 1 month after BSP. All major eCBs including anandamide, 2-AG, 1-AG, DEA, and BDNF increased after meditation by > 70% (P < 0.001). Increases of ≥20% in anandamide, 2-AG, 1-AG, and total AG levels after meditation from the baseline had weak correlations with changes in happiness and well-being. Conclusions. A short meditation experience improved focus, happiness, and positive well-being and reduced depression and anxiety in participants for at least 1 month. Participants had increased blood eCBs and BDNF, suggesting a role for these biomarkers in the underlying mechanism of meditation. Meditation is a simple, organic, and effective way to improve well-being and reduce depression and anxiety.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2020

Collective Risk Ranking of Highway Segments on the Basis of Severity-Weighted Crash Rates

This study is intended to focus on the major factors affecting traffic crash rates and severity levels, in addition to identifying crash-prone locations (i.e., black spots) based on the two indicators. The available crash data for different road segments used for the analysis were obtained from the Washington state database provided by the Highway Safety Information System (HSIS) for the years 2006 to 2011. A Random Forest (RF) classifier was used to predict the outcome level of crash severity, while crash rates were predicted by applying RF regressor. Certain features were selected for each model besides the abstraction of new features to check if there are unobserved correlations affecting the independent variables, such as accounting for the number and weight of crashes within 1 km2 area by implementing the Getis-Ord Gi∗ index. Moreover, to calculate the collective risk (CR) score, crash rates were adjusted to incorporate crash severity weights (cost per severity type) and regression-to-the-mean (RTM) bias via Empirical Bayes (EB) method. Finally, segments were ranked according to their CR score.

Li, Dawei↗

DOE OSTI2022

Longitudinal Trajectories of Memory Performance in Patients with Early-Stage Breast Cancer

Background. While breast cancer and its treatments may affect cognition, the longitudinal trajectories of cognition among those receiving differing cancer treatment types remain poorly understood. Prior research suggests hippocampal-prefrontal cortex network integrity may influence cognition, although how this network predicts performance over time remains unclear. Methods. We conducted a prospective trial including 69 patients with early-stage breast cancer receiving adjuvant therapy and 12 controls. Longitudinal cognitive testing was conducted at four visits: pretreatment-baseline, 6-7 months, 14-15 months, and 23-24 months. Cognitive composite scores of episodic memory, executive functioning, and processing speed were assessed at each timepoint. Baseline structural MRI was obtained in a subset of these participants, and hippocampal and prefrontal cortex regional volumes were extracted. Results. Longitudinal linear mixed modeling revealed significant group by time interactions on memory performance, controlling for age and education. Post hoc analyses revealed this effect was driven by patients treated with chemotherapy or chemotherapy plus hormone therapy, who demonstrated the least improvement in memory scores over time. Treatment group did not significantly influence the relationship between time and processing speed or executive functioning. Neither pretreatment hippocampal nor prefrontal volume differed between groups, and there were no significant group by time by baseline regional volume effects on cognition. Conclusion. Patients with early-stage breast cancer treated with chemotherapy or chemotherapy plus hormone therapy benefit less from practice effects seen in healthy controls on memory tests. Loss of longitudinal practice effect may be a new and clinically relevant measure for capturing patients’ experience of cognitive difficulties after treatment.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2022

Effect of Low Frequency Repetitive Transcranial Magnetic Stimulation (rTMS) Combined with Hyperbaric Oxygen (HBO) on Awakening of Coma Patients with Traumatic Brain Injury

Coma caused by craniocerebral injury is a common condition of neurosurgical acute injury. There is no specific method to promote awakening in a clinic. Early comprehensive treatment may be helpful to patients. The common methods are hyperbaric oxygen (HBO) and low-frequency repetitive transcranial magnetic stimulation (rTMS). However, the application effect and mechanism of rTMS combined with HBO on coma patients with traumatic brain injury need to be further studied. The brain stem auditory evoked potential (BAEP) is examined by the Kennedy coma recovery scale (CRS-R), the recovery of brain function and the state of consciousness are evaluated, and the therapeutic effect is evaluated by the Glasgow Coma Scale (GCS). Cerebrospinal fluid NE level, MCA blood flow velocity, and left brainstem and right brainstem auditory evoked potential are used to evaluate brain rehabilitation. RTMS combined with HBO could shorten the wake-up time, improve the wake-up rate, improve the GCS score and CRS-R score, shorten the brain wave latency time of the left and right brainstem, increase the NE level of cerebrospinal fluid, and decrease the blood flow velocity of MCA. RTMS combines with HBO can improve the nerve excitability of brain cells, reduce the disturbance of consciousness, promote the functional recovery of brain injury, and has a certain role in promoting the awakening of patients with traumatic brain injury coma.

Liu, Mei↗

DOE OSTI2023

EI_MS_ML

The unambiguous identification of compounds from their electron ionization mass (EI-MS) spectra remains a significant unsolved problem in the field of metabolomics and analytical chemistry as a whole. Typically EI-MS spectra are compared using various mathematical operations that convert the spectral similarity or differences into a distance-like metric that roughly approximates the similarity of any two spectra. A commonly used metric for this is the cosine similarity metric which has values close to one for very similar spectra and a value of zero for very dissimilar spectra; however, no metric is perfect. Due to the prevalence of structurally-similar compounds such as isomers and the prevalence of certain fragmentation patterns across structurally-dissimilar compounds, the unambiguous assignment of EI-MS spectra compounds remains difficult. Frequently, querying an observed EI-MS spectrum against a large database such as the NIST17 library yields multiple possible assignments requiring the end user to distinguish between multiple high scoring hits, or multiple low scoring hits while keeping in mind that the correct hit may not be in the database at all. Although techniques such as orthogonal information from techniques such as chromatography can greatly aid in unambiguous assignment, this also requires more complicated experimental designs and access to more complicated analytical instrumentation. Substructures can be trivially detected and represented as strings using a previously published technique called node coloring from a known chemical structure. However, for experimentally-derived EI-MS spectra this information must be derived from the spectra itself (i.e., because we do not know what compound it represents). To achieve this, the software uses techniques from the field of machine learning and a large training dataset of EI-MS spectra corresponding to known structures annotated with substructure strings, to build models that can predict the presence of a given chemical substructure from an EI-MS spectrum directly.If these predictions are of high-quality (i.e., are unlikely to be false positives), the presence of one or more predicted substructures can be used to constrain the number of possible hits for a query spectrum. Mathematically, this restriction could be expressed in many forms, but the most straight-forward implementation is to weight the cosine similarity of a query spectrum and a plausible database match with a Tanimoto-like coefficient based on the ratio of the number of substructures predicted to the number of substructures present in the potential database hit. Determining which combination of models best reduces assignment ambiguity will be achieved using a combination of manual curation and optimization techniques such as genetic algorithms. This software will perform all the steps necessary to construct said models from a training dataset and evaluate them using a holdout dataset. Various statistical analyses can be performed to determine if this approach does decrease assignment ambiguity. For example, if this approach works, on average, the rank-order of the correct assignment for the holdout set of EI-MS spectra should decrease and the weighted cosine similarities for most of the possible matches in the database should be better than the unweighted cosine similarities. Furthermore, this same pipeline can be used on real experimental data to generate less ambiguous assignments.

Mitchell, Joshua↗