SEARCH · Engineering Papers

Results for “graph sequence”

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 127 records · Page 7

All sources NASA NTRS DOE OSTI

DOE OSTI2016

Automatic segmentation of the glenohumeral cartilages from magnetic resonance images

Purpose: Magnetic resonance (MR) imaging plays a key role in investigating early degenerative disorders and traumatic injuries of the glenohumeral cartilages. Subtle morphometric and biochemical changes of potential relevance to clinical diagnosis, treatment planning, and evaluation can be assessed from measurements derived from in vivo MR segmentation of the cartilages. However, segmentation of the glenohumeral cartilages, using approaches spanning manual to automated methods, is technically challenging, due to their thin, curved structure and overlapping intensities of surrounding tissues. Automatic segmentation of the glenohumeral cartilages from MR imaging is not at the same level compared to the weight-bearing knee and hip joint cartilages despite the potential applications with respect to clinical investigation of shoulder disorders. In this work, the authors present a fully automated segmentation method for the glenohumeral cartilages using MR images of healthy shoulders. Methods: The method involves automated segmentation of the humerus and scapula bones using 3D active shape models, the extraction of the expected bone–cartilage interface, and cartilage segmentation using a graph-based method. The cartilage segmentation uses localization, patient specific tissue estimation, and a model of the cartilage thickness variation. The accuracy of this method was experimentally validated using a leave-one-out scheme on a database of MR images acquired from 44 asymptomatic subjects with a true fast imaging with steady state precession sequence on a 3 T scanner (Siemens Trio) using a dedicated shoulder coil. The automated results were compared to manual segmentations from two experts (an experienced radiographer and an experienced musculoskeletal anatomist) using the Dice similarity coefficient (DSC) and mean absolute surface distance (MASD) metrics. Results: Accurate and precise bone segmentations were achieved with mean DSC of 0.98 and 0.93 for the humeral head and glenoid fossa, respectively. Mean DSC scores of 0.74 and 0.72 were obtained for the humeral and glenoid cartilage volumes, respectively. The manual interobserver reliability evaluated by DSC was 0.80 ± 0.03 and 0.76 ± 0.04 for the two cartilages, implying that the automated results were within an acceptable 10% difference. The MASD between the automatic and the corresponding manual cartilage segmentations was less than 0.4 mm (previous studies reported mean cartilage thickness of 1.3 mm). Conclusions: This work shows the feasibility of volumetric segmentation and separation of the glenohumeral cartilages from MR images. To their knowledge, this is the first fully automated algorithm for volumetric segmentation of the individual glenohumeral cartilages from MR images. The approach was validated against manual segmentations from experienced analysts. In future work, the approach will be validated on imaging datasets acquired with various MR contrasts in patients.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2017

Distributed Energy Resources Customer Adoption Model - Graphical User Interface, Version 2.1.8

The DER-CAM Graphical User Interface has been redesigned to consist of a dynamic tree structure on the left side of the application window to allow users to quickly navigate between different data categories and views. Views can either be tables with model parameters and input data, the optimization results, or a graphical interface to draw circuit topology and visualize investment results. The model parameters and input data consist of tables where values are assigned to specific keys. The aggregation of all model parameters and input data amounts to the data required to build a DER-CAM model, and is passed to the GAMS solver when users initiate the DER-CAM optimization process. Passing data to the GAMS solver relies on the use of a Java server that handles DER-CAM requests, queuing, and results delivery. This component of the DER-CAM GUI can be deployed either locally or remotely, and constitutes an intermediate step between the user data input and manipulation, and the execution of a DER-CAM optimization in the GAMS engine. The results view shows the results of the DER-CAM optimization and distinguishes between a single and a multi-objective process. The single optimization runs the DER-CAM optimization once and presents the results as a combination of summary charts and hourly dispatch profiles. The multi-objective optimization process consists of a sequence of runs initiated by the GUI, including: 1) CO2 minimization, 2) cost minimization, 3) a user defined number of points in-between objectives 1) and 2). The multi-objective results view includes both access to the detailed results of each point generated by the process as well as the generation of a Pareto Frontier graph to illustrate the trade-off between objectives. DER-CAM GUI 2.1.8 also introduces the ability to graphically generate circuit topologies, enabling support to DER-CAM 5.0.0. This feature consists of: 1) The drawing area, where users can manually create nodes and define their properties (e.g. point of common coupling, slack bus, load) and connect them through edges representing either power lines, transformers, or heat pipes, all with user defined characteristics (e.g., length, ampacity, inductance, or heat loss); 2) The tables, which display the user-defined topology in the final numerical form that will be passed to the DER-CAM optimization. Finally, the DER-CAM GUI is also deployed with a database schema that allows users to provide different energy load profiles, solar irradiance profiles, and tariff data, that can be stored locally and later used in any DER-CAM model. However, no real data will be delivered with this version.

Ewald, Friedrich↗

DOE OSTI2026

LibraryX: A Framework for Cross-Library-Call Optimization

Scientific applications utilize performance libraries as a software engineering concept: these libraries encapsulate important and well-understood (mathematical) operations, allow for reuse, and are implemented and tuned by experts. Domain scientists then implement complex algorithms based on these domainspecific libraries. While individual library calls are optimized, larger performance gains across sequences of calls—sometimes spanning multiple libraries—are often unrealized, forcing a trade-off between performance and implementation complexity.To overcome this issue, we propose LibraryX, an approach and a system that allows for cross-library-call optimization even when library calls stem from multiple performance libraries. LibraryX annotates library calls with semantic information and optimizes entire directed acyclic graphs (DAGs) of calls dynamically using the SPIRAL code generation system. We demonstrate its effectiveness across a range of memory bound workloads, achieving significant speedups on Nvidia, AMD, and Intel accelerators compared to code using native libraries without cross-call optimization.

Rao, Sanil [Carnegie Mellon University,Department ↗

DOE OSTI2016

Final Technical Report on Quantifying Dependability Attributes of Software Based Safety Critical Instrumentation and Control Systems in Nuclear Power Plants

With the current transition from analog to digital instrumentation and control systems in nuclear power plants, the number and variety of software-based systems have significantly increased. The sophisticated nature and increasing complexity of software raises trust in these systems as a significant challenge. The trust placed in a software system is typically termed software dependability. Software dependability analysis faces uncommon challenges since software systems’ characteristics differ from those of hardware systems. The lack of systematic science-based methods for quantifying the dependability attributes in software-based instrumentation as well as control systems in safety critical applications has proved itself to be a significant inhibitor to the expanded use of modern digital technology in the nuclear industry. Dependability refers to the ability of a system to deliver a service that can be trusted. Dependability is commonly considered as a general concept that encompasses different attributes, e.g., reliability, safety, security, availability and maintainability. Dependability research has progressed significantly over the last few decades. For example, various assessment models and/or design approaches have been proposed for software reliability, software availability and software maintainability. Advances have also been made to integrate multiple dependability attributes, e.g., integrating security with other dependability attributes, measuring availability and maintainability, modeling reliability and availability, quantifying reliability and security, exploring the dependencies between security and safety and developing integrated analysis models. However, there is still a lack of understanding of the dependencies between various dependability attributes as a whole and of how such dependencies are formed. To address the need for quantification and give a more objective basis to the review process -- therefore reducing regulatory uncertainty -- measures and methods are needed to assess dependability attributes early on, as well as throughout the life-cycle process of software development. In this research, extensive expert opinion elicitation is used to identify the measures and methods for assessing software dependability. Semi-structured questionnaires were designed to elicit expert knowledge. A new notation system, Causal Mechanism Graphing, was developed to extract and represent such knowledge. The Causal Mechanism Graphs were merged, thus, obtaining the consensus knowledge shared by the domain experts. In this report, we focus on how software contributes to dependability. However, software dependability is not discussed separately from the context of systems or socio-technical systems. Specifically, this report focuses on software dependability, reliability, safety, security, availability, and maintainability. Our research was conducted in the sequence of stages found below. Each stage is further examined in its corresponding chapter. Stage 1 (Chapter 2): Elicitation of causal maps describing the dependencies between dependability attributes. These causal maps were constructed using expert opinion elicitation. This chapter describes the expert opinion elicitation process, the questionnaire design, the causal map construction method and the causal maps obtained. Stage 2 (Chapter 3): Elicitation of the causal map describing the occurrence of the event of interest for each dependability attribute. The causal mechanisms for the “event of interest” were extracted for each of the software dependability attributes. The “event of interest” for a dependability attribute is generally considered to be the “attribute failure”, e.g. security failure. The extraction was based on the analysis of expert elicitation results obtained in Stage 1. Stage 3 (Chapter 4): Identification of relevant measurements. Measures for the “events of interest” and their causal mechanisms were obtained from expert opinion elicitation for each of the software dependability attributes. The measures extracted are presented in this chapter. Stage 4 (Chapter 5): Assessment of the coverage of the causal maps via measures. Coverage was assessed to determine whether the measures obtained were sufficient to quantify software dependability, and what measures are further required. Stage 5 (Chapter 6): Identification of “missing” measures and measurement approaches for concepts not covered. New measures, for concepts that had not been covered sufficiently as determined in Stage 4, were identified using supplementary expert opinion elicitation as well as literature reviews. Stage 6 (Chapter 7): Building of a detailed quantification model based on the causal maps and measurements obtained. Ability to derive such a quantification model shows that the causal models and measurements derived from the previous stages (Stage 1 to Stage 5) can form the technical basis for developing dependability quantification models. Scope restrictions have led us to prioritize this demonstration effort. The demonstration was focused on a critical system, i.e. the reactor protection system. For this system, a ranking of the software dependability attributes by nuclear stakeholders was developed. As expected for this application, the stakeholder ranking identified safety as the most critical attribute to be quantified. A safety quantification model limited to the requirements phase of development was built. Two case studies were conducted for verification. A preliminary control gate for software safety for the requirements stage was proposed and applied to the first case study. The control gate allows a cost effective selection of the duration of the requirements phase.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2025

Leveraging Natural Language Processing and Generative Models in Molecular Chemistry: Property Prediction and Novel Compound Generation

The accurate prediction of molecular properties is important for the rational design and the advancement of green chemistry and sustainable materials research. However, the predictive power of traditional computational chemistry methods is limited due to computational restrictions. Here, in this study, we examine an alternative approach to the accurate prediction of properties of organic compounds: natural language processing (NLP)-based molecular embedding. Using viscosity, partition coefficient (log P), and enthalpy of vaporization as test properties through a survey of comprehensive datasets comprising 5695 data points for viscosity, 25 870 data points for log P, and 2296 data points for enthalpy of vaporization. These are important properties for the design of greener, safer, and sustainable chemical processes. Models were trained using NLP methods such as Mol2vec and fine-tuned ChemBERTa, and results were compared with traditional input featurization techniques such as Morgan fingerprints and quantum chemistry derived sigma profiles and DFT features. Among the various machine learning models, Mol2vec demonstrated superior predictive capabilities, achieving the highest correlation coefficient (R 2 = 0.945) and lowest RMSE (0.106 mPa s) for viscosity, as well as high accuracy for log P and enthalpy of vaporization predictions. These findings establish the Mol2vec featurization technique, graph-convolutional neural networks (GCNN), and fine-tuned ChemBERTa model as powerful tools for predictive modeling of organic compounds properties, offering a significant improvement over previously used featurization techniques and opening up strategies for very-high-throughput computational screening. Finally, we integrated ML models with hybrid language-model-based generative adversarial networks (LM-GAN) to generate novel molecular sequences with desirable properties for different research applications. The ability to computationally design solvents with lower viscosity, lower log P, and lower enthalpy of vaporization offers a data-driven route to accelerating the discovery of sustainable alternatives to traditionally toxic solvents.

ChemBERTa↗

DOE OSTI2024

DNA parts and gene constructs for plant biodesign

Plant biodesign requires the knowledge of DNA parts (e.g., genes, promoters, terminators), along with their combinations (as gene constructs) linked to engineered traits. DNA parts with validated or predicted functions in plants have been deposited in various online databases. However, these existing databases focus on basic biological functions of individual DNA parts, leaving a gap between basic knowledge and bioengineering applications. To fill this knowledge gap, we have created a user-friendly, open-ended database as a knowledge graph linking DNA parts to gene constructs to traits. This database contains experimentally validated DNA parts and gene constructs documented in peer-reviewed publications. The DNA parts include 1) molecular components with biological functions, such as genes involved in various biological processes (e.g., metabolic and signal transduction pathways) and 2) molecular components with technical functions, such as gene expression, genome engineering and sequence splicing. The gene constructs deposited in this database include both single-gene and multi-gene constructs. This database allows users to submit DNA parts and gene construct compositions linked to engineered traits described in peer-reviewed publications, providing a public digital repository for sharing the biodesign information among the researchers in the fields of plant biotechnology and plant synthetic biology.

plant biodesign synthetic biology gene constructs ↗

DOE OSTI2016

Development of the FHR advanced natural circulation analysis code and application to FHR safety analysis

The University of California, Berkeley (UCB) is performing thermal hydraulics safety analysis to develop the technical basis for design and licensing of fluoride-salt-cooled, high-temperature reactors (FHRs). FHR designs investigated by UCB use natural circulation for emergency, passive decay heat removal when normal decay heat removal systems fail. The FHR advanced natural circulation analysis (FANCY) code has been developed for assessment of passive decay heat removal capability and safety analysis of these innovative system designs. The FANCY code uses a one-dimensional, semi-implicit scheme to solve for pressure-linked mass, momentum and energy conservation equations. Graph theory is used to automatically generate a staggered mesh for complicated pipe network systems. Heat structure models have been implemented for three types of boundary conditions (Dirichlet, Neumann and Robin boundary conditions). Heat structures can be composed of several layers of different materials, and are used for simulation of heat structure temperature distribution and heat transfer rate. Control models are used to simulate sequences of events or trips of safety systems. A proportional-integral controller is also used to automatically make thermal hydraulic systems reach desired steady state conditions. A point kinetics model is used to model reactor kinetics behavior with temperature reactivity feedback. The underlying large sparse linear systems in these models are efficiently solved by using direct and iterative solvers provided by the SuperLU code on high performance machines. Input interfaces are designed to increase the flexibility of simulation for complicated thermal hydraulic systems. In conclusion, this paper mainly focuses on the methodology used to develop the FANCY code, and safety analysis of the Mark 1 pebble-bed FHR under development at UCB is performed.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

DOE OSTI2021

Network analysis of earthquake ground motion spatial correlation: a case study with the San Jacinto seismic nodal array

SUMMARY The spatial correlation of earthquake ground motion intensity can be measured from strong motion data; however, the data used in past studies is sparsely sampled in space, and only the interstation distance was considered as a correlation variable. These limitations mean that we have only weak constraints on the true correlation structure of ground motion and that potentially important aspects of spatial correlation are unconstrained. In this study, we combine a large-N seismic array and graph analytics to explore this issue at a local scale using small local and regional earthquakes. Our result suggests site conditions, and how they interact with the incident seismic wavefield, strongly condition the spatial correlation of ground motion. Future progress in characterizing ground motion spatial variability will require dense wavefield measurements, either through nodal deployments, or perhaps distributed acoustic sensing measurements, of seismic wavefields. Aftershock sequences of major earthquakes would provide particularly data-rich targets of opportunity.

58 GEOSCIENCES↗

DOE OSTI2024

Estimating Sparse Direct Effects in Multivariate Regression With the Spike-and-Slab LASSO

The multivariate regression interpretation of the Gaussian chain graph model simultaneously parametrizes (i) the direct effects of p predictors on q outcomes and (ii) the residual partial covariances between pairs of outcomes. We introduce a new method for fitting sparse versions of these models with spike-and-slab LASSO (SSL) priors. We develop an Expectation Conditional Maximization algorithm to obtain sparse estimates of the p × q matrix of direct effects and the q × q residual precision matrix. Our algorithm iteratively solves a sequence of penalized maximum likelihood problems with self-adaptive penalties that gradually filter out negligible regression coefficients and partial covariances. Because it adaptively penalizes individual model parameters, our method is seen to outperform fixed-penalty competitors on simulated data. We establish the posterior contraction rate for our model, buttressing our method’s excellent empirical performance with strong theoretical guarantees. Using our method, we estimated the direct effects of diet and residence type on the composition of the gut microbiome of elderly adults.

EM algorithm↗

DOE OSTI2023

Efficient 3D Molecular Design with an E(3) Invariant Transformer VAE

This work introduces a three-dimensional (3D) invariant graph-to-string transformer variational autoencoders (VAE) (Vagrant) for generating molecules with accurate density functional theory (DFT)-level properties. Vagrant learns to model the joint probability distribution of a 3D molecular structure and its properties by encoding molecular structures into a 3D-aware latent space. Directed navigation through this latent space implicitly optimizes the 3D structure of a molecule, and the latent embedding can be used to condition a generative transformer to predict the candidate structure as a one-dimensional (1D) sequence. Additionally, we introduce two novel sampling methods that exploit the latent characteristics of a VAE to improve performance. Here, we show that our method outperforms comparable 3D autoregressive and diffusion methods for predicting quantum chemical property values of novel molecules in terms of both sample quality and computational efficiency.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2019

PHANOTATE: a novel approach to gene identification in phage genomes

Motivation: Currently there are no tools specifically designed for annotating genes in phages. Several tools are available that have been adapted to run on phage genomes, but due to their underlying design, they are unable to capture the full complexity of phage genomes. Phages have adapted their genomes to be extremely compact, having adjacent genes that overlap and genes completely inside of other longer genes. This non-delineated genome structure makes it difficult for gene prediction using the currently available gene annotators. Here we present PHANOTATE, a novel method for gene calling specifically designed for phage genomes. Although the compact nature of genes in phages is a problem for current gene annotators, we exploit this property by treating a phage genome as a network of paths: where open reading frames are favorable, and overlaps and gaps are less favorable, but still possible. We represent this network of connections as a weighted graph, and use dynamic programing to find the optimal path. Results: We compare PHANOTATE to other gene callers by annotating a set of 2133 complete phage genomes from GenBank, using PHANOTATE and the three most popular gene callers. We found that the four programs agree on 82% of the total predicted genes, with PHANOTATE predicting more genes than the other three. We searched for these extra genes in both GenBank’s non-redundant protein database and all of the metagenomes in the sequence read archive, and found that they are present at levels that suggest that these are functional protein-coding genes. Availability and implementation: https://github.com/deprekate/PHANOTATE. Supplementary information: Supplementary data are available at Bioinformatics online.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2022

Illuminating the pathways to carbon liberation: a systems approach to characterizing the consequential unknowns of carbon transformation and loss from thawing permafrost peatlands (Final Report)

The IsoGenie3 Project delivered new systems-level insights into carbon cycling in thawing permafrost landscapes, with an emphasis on methane and carbon dioxide emissions. From >200 samples from the site collected over a decade, co-analyzed for geochemistry and microbiology, the team recovered ~1,500 assembled microbial genomes and ~1,900 viral population genomes, revealing appreciable genetic novelty - from a new highly abundant bacterial phylum, to novel methane consumers and their activities, to rampant viral novelty. IsoGenie3 linked these organisms to carbon compound transformations (which define the cycling of organic matter in soils, and the loss of the greenhouse gases carbon dioxide and methane), and saw that the microbes at each stage of permafrost thaw had different genetic potential to degrade categories of compounds, expressed that genetic potential differently, and actually transformed carbon compounds into greenhouse gases in different ways. IsoGenie 3 identified that some of the thaw-stage differences were due to plant-microbiome relationships; the plant species across the thaw gradient contributed different carbon compounds into the soil, and hosted distinct microbiota (differing among parts of plants as well as species). Lastly, microbes in the saturated post-thaw conditions appeared likely to contribute to the mobilization and toxification of mercury released during thaw. In parallel with ongoing field sampling and analysis, hypotheses arising from field observations were tested via lab incubation experiments. When communities are taken out of their native habitats, they behave differently, and the team first rigorously quantified the magnitude of this effect on microbiome composition and functional capacity, organic matter composition, and gas production; overall the main system processes were maintained in the lab incubations under the conditions tested. Further, the microbial data could inform geochemical reaction network models of those processes. Then, the team ran experiments with additions of compounds, varying temperature, and “live” vs. “dead” peat (the latter having been gamma irradiated, with a few additional variants to control for methodological artifacts). From these, we (a) determined the importance of plant-derived soluble phenolic compounds in bogs’ extraordinary recalcitrance of organic matter, and carbon gas emissions skewed to carbon dioxide; (b) proposed an abiotic ‘tanning’ mechanism, which could contribute to Sphagnum’s inhibitory effect on anaerobic decomposition through alteration of N availability. IsoGenie3 illuminated longer-term and landscape-scale interactions of permafrost thaw and carbon cycling, advancing knowledge of the drivers of methane dynamics not only across in the permafrost-associated peatland (where hydrology and plant communities dictate microbiomes) but also their interconnected lakes (where sediment carbon quality and resident microbiota are determined by position within lake, and lake features). By leveraging observations of site methane dynamics extending well before this project, the team was able to construct a 44-year portrait of the interplay of permafrost thaw, hydrology, vegetation dynamics, and carbon gas emissions, and the doubling of the fully-thawed fens over this time. From the detailed study of this focal site, IsoGenie3 also aimed to improve model representation of these kinds of sites and processes. To improve predictions of methane transformations, we incorporated acetate and isotope dynamics into the ‘DNDC’ biogeochemistry model. In addition, recovered genomes were grouped into ‘functional groups’, i.e. the genomes that perform a specific function of interest, then used to parameterize maximum growth rate and optimum growth temperature (via signatures in their sequence composition) for the BioCrunch model. The BioCrunch model was then in turn used to test the impact of increasing functional resolution of the microbes, on the carbon gas emissions. Lastly for modeling, the ecosys model was parameterized from the microbial and other data, and used to evaluate drivers of e.g. change in methane emissions. Finally, this project also led to the development of a range of new methods and tools, a new metric of organic matter decomposability, as well as a graph-database solution to multidisciplinary data storage and querying. This project’s ongoing analyses at our focal site also contributed to broader advancements in understanding elements of genetic plasticity and methane metabolism, climate change microbiology and community assembly, global peatland geochemistry and Arctic lakes’ roles in climate feedbacks.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2020

Threat Reduction Research Networks: Fostering Sustainable Collaborations Through Trainings for Genomics for Biosurveillance

Scientific research communities can be represented as heterogeneous or multidimensional networks encompassing multiple types of entities and relationships. These networks might include researchers, institutions, meetings, and publications, connected by relationships like authorship, employment, and attendance. We describe a method for efficiently and flexibly capturing, storing, and extracting information from multidimensional scientific networks using a graph database. The database structure is based on an ontology that captures allowable types of entities and relationships. This allows us to construct a variety of projections of the underlying multidimensional graph through database queries to answer specific research questions. We demonstrate this process through a study of the U.S. Biological Threat Reduction Program (BTRP), which seeks to develop Threat Reduction Networks to build and strengthen a sustainable international community of biosecurity, biosafety, and biosurveillance experts to address shared biological threat reduction challenges. Networks like these create connectional intelligence among researchers and institutions around the world, and are central to the concept of cooperative threat reduction. Our analysis focuses on a series of seven BTRP genome sequencing training workshops, showing how they created a growing network of participants and countries over time, which is also reflected in coauthorship relationships among attendees. By capturing concept and relationship hierarchies, our ontology-based approach allows us to pose general or specific questions about networks within the same framework. This approach can be applied to other research communities or multidimensional social networks to capture, analyze, and visualize different types of interactions and how they change over time.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2016

Epigraph: A Vaccine Design Tool Applied to an HIV Therapeutic Vaccine and a Pan-Filovirus Vaccine

Epigraph is an efficient graph-based algorithm for designing vaccine antigens to optimize potential T-cell epitope (PTE) coverage. Functionally, epigraph vaccine antigens are similar to Mosaic vaccines, which have demonstrated effectiveness in preliminary HIV non-human primate studies. In contrast to the Mosaic algorithm, Epigraph is substantially faster, and in restricted cases, provides a mathematically optimal solution. Furthermore, epigraph has new features that enable enhanced vaccine design flexibility. These features include the ability to exclude rare epitopes from a design, to optimize population coverage based on inexact epitope matches, and to apply the code to both aligned and unaligned input sequences. Epigraph was developed to provide practical design solutions for two outstanding vaccine problems. The first of these is a personalized approach to a therapeutic T-cell HIV vaccine that would provide antigens with an excellent match to an individual’s infecting strain, intended to contain or clear a chronic infection. The second is a pan-filovirus vaccine, with the potential to protect against all known viruses in the Filoviradae family, including ebolaviruses. A web-based interface to run the Epigraph tool suite is available (http://www.hiv.lanl.gov/content/sequence/EPIGRAPH/epigraph.html).

60 APPLIED LIFE SCIENCES↗

DOE OSTI2024

Invariant Features for Accurate Predictions of Quantum Chemical UV-vis Spectra of Organic Molecules

Including invariance of global properties of a phys-ical system as an intrinsic feature in graph neural networks (GNNs) enhances the model's robustness and generalizability and reduces the amount of training data required to obtain a desired accuracy for predictions of these properties. Existing open source GNN libraries construct invariant features only for specific GNN architectures. This precludes the generalization of invariant features to arbitrary message passing neural network (MPNN) layers which, in turn, precludes the use of these libraries for new, user-specified predictive tasks. To address this limitation, we implement invariant MPNNs into the flexible and scalable HydraGNN architecture. HydraGNN enables a seamless switch between various MPNNs in a unified layer sequence and allows for a fair comparison between the predictive performance of different MPNNs. We trained this enhanced HydraGNN archi-tecture on the ultraviolet-visible (UV-vis) spectrum of GDB-9 molecules, a feature that describes the molecule's electronic exci-tation modes, computed with time-dependent density functional tight binding (TD-DFTB) and available open source through the GDB-9-Ex dataset. We assess the robustness (i.e., accuracy and generalizability) of the predictions obtained using different invariant MPNNs with respect to different values of the full width at half maximum (FWHM) for the Gaussian smearing of the theoretical peaks. Our numerical results show that incorporating invariance in the HydraGNN architecture significantly enhances both accuracy and generalizability in predicting UV-vis spectra of organic molecules.

Baker, Justin↗

DOE OSTI2021

TX$^2$: Transformer eXplainability and eXploration

The Transformer eXplainability and eXploration (Martindale & Stewart, 2021), or TX 2 software package, is a library designed for artificial intelligence researchers to better understand the performance of transformer models (Vaswani et al., 2017) used for sequence classification. The tool is capable of integrating with a trained transformer model and a dataset split into training and testing populations to produce an ipywidget (Project Jupyter Contributors, 2021) dashboard with a number of visualizations to understand model performance with an emphasis on explainability and interpretability. The TX 2 package is primarily intended to integrate into a workflow centered around Jupyter Notebooks (Kluyver et al., 2016), and currently assumes the use of PyTorch (Paszke et al., 2019) and Hugging Face transformers library (Wolf et al., 2020). The dashboard includes visualization and data exploration features to aid researchers, including an interactive UMAP embedding graph (McInnes et al., 2018) to understand classification clusters, a word salience map that can be updated as researchers alter textual entries in near real time, a set of tools to understand word frequency and importance based on the clusters in the UMAP embedding graph, and a set of traditional confusion matrix analysis tools.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Replace Human Intelligence with Fast and Smart Geometric Reasoning and Graph Neural Network to Accelerate Next Gen ModSim Workflows

We present an agent-guided approach to CAD geometry decomposition that automates hex/hybrid meshing with graph neural networks (GNNs) to accelerate next-generation ModSim workflows. Our end-to-end pipeline (i) reduces 3D boundary-representation (B-Rep) models to a 2D chordal axis skeleton (CAT) and then to a 1D bipartite graph of surface and curve nodes, (ii) assigns per node labels as Cubit® WebCut actions, (iii) trains a multi-action GNN under supervised learning, and (iv) predicts five surface-node and three curve-node actions on out-of-distribution test geometries. Each graph node carries geometric, topological, and meshing attributes drawn from the B-Rep “skin” and CAT “skeleton,” with two-way mappings across 3D↔2D↔1D representations to maintain traceability back to 3D CAD. The supervised learning model exhibits stable convergence of the binary cross-entropy loss and achieves 98.7% accuracy on unseen lattice models. To operationalize decision-making, we rank predicted commands by geometric significance and prototyped the agent-guided workflow through the Cubit® Meshing PowerTool GUI. As a stretch goal, we explore reinforcement learning (RL) to reduce or remove label requirements and to learn policies for action sequences that maximize total reward (e.g., size of hex-meshable regions and resulting hex mesh quality). When all-hex meshing is not feasible, the agent assists in producing hybrid meshes—prioritizing hex in critical regions and transitioning to tetrahedral elements (tets) elsewhere—maintaining fidelity while ensuring robustness. The overarching objective is to replace manual, heuristics-based decomposition with data-driven, reproducible automation, cutting meshing turnaround time by orders of magnitude. We anticipate direct impact on simulation workflows through intelligent, scalable decomposition of complex CAD models into hex-meshable subdomains.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Closing the Loop between In Situ Stress Complexity and EGS Fracture Complexity

42 ENGINEERING↗