Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “compound screening”

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

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

At least 199 records · Page 11
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Whole‐Cell MALDI‐ToF MS Coupled with Untargeted Metabolomics Facilitates Investigations of Microbial Chemical Interactions

Abstract The emergence of drug‐resistant pathogens necessitates the development of new countermeasures. In this regard, the introduction of probiotics to directly attack or competitively exclude pathogens presents a useful strategy. Application of this approach requires an understanding of how a probiotic and its target pathogen interact. A key means of probiotic‐pathogen interaction involves the production of small molecules called natural products (NPs). Here, we report the use of whole‐cell matrix‐assisted laser desorption/ionization time‐of‐flight (MALDI‐ToF) mass spectrometry to characterize NP production by candidate probiotics (mouse airway microbiome isolates) when co‐cultured with the respiratory pathogen Burkholderia . We found that a Bacillus velezensis strain inhibits growth of and elicits NP production by Burkholderia thailandensis . Dereplication of known NPs detected in the metabolome of this B. velezensis strain suggests that a previously unannotated bioactive compound is involved. Thus, we present the use of whole‐cell MALDI as a broadly applicable method for screening the NP composition of microbial co‐cultures; this can be combined with other ‐omics methods to characterize probiotic‐pathogen and other microbe‐microbe interactions.

Aiosa, Nicole↗
DOE OSTI2025

Mechanistic Insights into the Demethylation of Lignin‐Derived Structures Using Protic Ionic Liquids: A Density Functional Theory Study

Lignin valorization is restricted by the stability of its methoxy groups, creating a critical need for efficient demethylation strategies. Here, in this study, density functional theory (DFT) is employed to dissect the mechanistic pathways of demethylation in lignin model compounds, guaiacol and syringol, using protic ionic liquids (PILs) that act as both solvent and catalyst. Conductor‐like Screening Model for Real Solvents (COSMO‐RS) analysis identifies monoethanolammonium acetate ([MEOA][Ace]) as the most promising medium, attributed to its strong hydrogen bonding network and solvation ability. By integrating implicit and explicit solvation models, it is revealed that an acid‐catalyzed hydrolytic mechanism governs demethylation, with PILs stabilizing crucial transition states and intermediates. Complementary electronic structure evaluations, including highest occupied molecular orbital–lowest unoccupied molecular orbital (HOMO‐LUMO) gap analysis, charge distribution, and electrostatic potential mapping, demonstrate how PILs lower energetic barriers and enhance reactivity. To simulate realistic environments, this study is extended to lignin dimer complexes with varying water content, uncovering how water‐bridged solvation reverses demethylation preference from guaiacyl to syringyl units. This mechanistic shift aligns with experimental observations showing faster S‐unit reactivity in hydrated systems. Together, these findings provide atomic‐level insight into lignin demethylation dynamics and highlight how tuning acid concentration in PILs can accelerate kinetics, enabling a rational pathway toward next‐generation biomass conversion technologies.

COSMO-RS↗
DOE OSTI2022

Pose Classification Using Three-Dimensional Atomic Structure-Based Neural Networks Applied to Ion Channel–Ligand Docking

The identification of promising lead compounds showing pharmacological activities toward a biological target is essential in early stage drug discovery. With the recent increase in available small-molecule databases, virtual high-throughput screening using physics-based molecular docking has emerged as an essential tool in assisting fast and cost-efficient lead discovery and optimization. However, the best scored docking poses are often suboptimal, resulting in incorrect screening and chemical property calculation. We address the pose classification problem by leveraging data-driven machine learning approaches to identify correct docking poses from AutoDock Vina and Glide screens. To enable effective classification of docking poses, we present two convolutional neural network approaches: a three-dimensional convolutional neural network (3D-CNN) and an attention-based point cloud network (PCN) trained on the PDBbind refined set. We demonstrate the effectiveness of our proposed classifiers on multiple evaluation data sets including the standard PDBbind CASF-2016 benchmark data set and various compound libraries with structurally different protein targets including an ion channel data set extracted from Protein Data Bank (PDB) and an in-house KCa3.1 inhibitor data set. Our experiments show that excluding false positive docking poses using the proposed classifiers improves virtual high-throughput screening to identify novel molecules against each target protein compared to the initial screen based on the docking scores.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

In crystallo screening for proline analog inhibitors of the proline cycle enzyme PYCR1

Pyrroline-5-carboxylate reductase 1 (PYCR1) catalyzes the biosynthetic half-reaction of the proline cycle by reducing Δ 1 -pyrroline-5-carboxylate (P5C) to proline through the oxidation of NAD(P)H. Many cancers alter their proline metabolism by up-regulating the proline cycle and proline biosynthesis, and knockdowns of PYCR1 lead to decreased cell proliferation. Thus, evidence is growing for PYCR1 as a potential cancer therapy target. Inhibitors of cancer targets are useful as chemical probes for studying cancer mechanisms and starting compounds for drug discovery; however, there is a notable lack of validated inhibitors for PYCR1. To fill this gap, we performed a small-scale focused screen of proline analogs using X-ray crystallography. Five inhibitors of human PYCR1 were discovered: l-tetrahydro-2-furoic acid, cyclopentanecarboxylate, l-thiazolidine-4-carboxylate, l-thiazolidine-2-carboxylate, and N-formyl l-proline (NFLP). The most potent inhibitor was NFLP, which had a competitive (with P5C) inhibition constant of 100 μm. The structure of PYCR1 complexed with NFLP shows that inhibitor binding is accompanied by conformational changes in the active site, including the translation of an α-helix by 1 Å. These changes are unique to NFLP and enable additional hydrogen bonds with the enzyme. NFLP was also shown to phenocopy the PYCR1 knockdown in MCF10A H-RAS V12 breast cancer cells by inhibiting de novo proline biosynthesis and impairing spheroidal growth. In summary, we generated the first validated chemical probe of PYCR1 and demonstrated proof-of-concept for screening proline analogs to discover inhibitors of the proline cycle.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Polarity-Tolerant Chloride Binding in Foldamer Capsules by Programmed Solvent-Exclusion

Persistent anion binding in a wide range of solution environments is a key challenge that continues to motivate and demand new strategies in synthetic receptor design. Though strong binding in low-polarity solvents has become routine, our ability to maintain high affinities in high-polarity solvents has not yet reached the standard set by nature. Anions are bound and transported regularly in aqueous environments by proteins that use secondary and tertiary structure to isolate anion binding sites from water. Inspired by this principle of solvent exclusion, we created a sequence-defined foldameric capsule whose global minimum conformation displays a helical folded state and is preorganized for 1:1 anion complexation. The high stability of the folded geometry and its ability to exclude solvent were supported by solid-state and solution phase studies. This capsule then withstood a 4-fold increase in solvent dielectric constant (εr) from dichloromethane (9) to acetonitrile (36) while maintaining a high and solvent-independent affinity of 10 5 M –1 ; ΔG ~ 28 kJ mol –1 . This behavior is unusual. More typical of solvent-dependent behavior, Cl – affinities were seen to plummet in control compounds, such as aryl-triazole macrocycles and pentads, with their solvent-exposed binding cavities susceptible to dielectric screening. Finally, dimethyl sulfoxide denatures the foldamer by putative solvent binding, which then lowers the foldamer’s Cl – affinity to normal levels. Furthermore, the design of this capsule demonstrates a new prototype for the development of potent receptors that can operate in polar solvents and has the potential to help manage hydrophilic anions present in the hydrosphere and biosphere.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Ambient mechanosynthesis of flexible two-dimensional covalent organic frameworks

We present the first ambient mechanosynthesis of 16 flexible covalent organic frameworks (COFs) within an hour. Notably, one representative COF exhibited a high iodine uptake capacity of ∼4.3 g g −1 from aqueous solutions and 5.97 g g −1 from vapor. Flexible two-dimensional covalent organic frameworks (2D COFs) constructed from nonplanar building blocks represent an emerging paradigm in COF design. Nevertheless, the prevailing solvothermal synthesis suffers from low time efficiency, environmental unfriendliness, and cumbersome protocols. Here, we address these challenges by developing the first ambient mechanosynthesis of a diverse library of flexible 2D COFs. Sixteen distinct triazine-cored Schiff-base COFs, including five as-yet-unreported ones, were rapidly synthesized via ball milling using 2,4,6-tris(4-aminophenoxy)-1,3,5-triazine (TPT-NH 2 ) and 2,4,6-tris(4-formylphenoxy)-1,3,5-triazine (TPT-CHO) as building blocks. Notably, the representative COF, MC-flexible-COF-1, was synthesized in as little as one hour under mechanosynthesis conditions, whereas it remained unattainable via the traditional solvothermal method, despite prolonged heating and extensive solvent screening. The highly dynamic nature of the imine linkage was unequivocally demonstrated through mechanochemical “scrambling” experiments using molecular model compounds. Furthermore, MC-flexible-COF-1 exhibited a high iodine uptake capacity of ∼4.30 g g −1 from aqueous solutions and 5.97 g g −1 from the vapor phase. This work underscores the immense potential of mechanochemistry as a powerful and sustainable tool for the rapid synthesis of advanced 2D COFs, including those inaccessible via conventional solution-based methods.

Nailwal, Yogendra↗
NASA NTRS2012

Visible Light Responsive Catalysts Using Quantum Dot-Modified Ti02 for Air and Water Purification

The method of photocatalysis utilizing titanium dioxide, TiO2, as the catalyst has been widely studied for trace contaminant control for both air and water applications because of its low energy consumption and use of a regenerable catalyst. Titanium dioxide requires ultraviolet light for activation due to its band gap energy of 3.2 eV. Traditionally, Hg-vapor fluorescent light sources are used in PCO reactors and are a setback for the technology for space application due to the possibility of Hg contamination. The development of a visible light responsive (VLR) TiO2-based catalyst could lead to the use of solar energy in the visible region (approx.45% of the solar spectrum lies in the visible region; > 400 nm) or highly efficient LEDs (with wavelengths > 400 nm) to make PCO approaches more efficient, economical, and safe. Though VLR catalyst development has been an active area of research for the past two decades, there are few commercially available VLR catalysts; those that are available still have poor activity in the visible region compared to that in the UV region. Thus, this study was aimed at the further development of VLR catalysts by a new method - coupling of quantum dots (QD) of a narrow band gap semiconductor (e.g., CdS, CdSe, PbS, ZnSe, etc.) to the TiO2 by two preparation methods: 1) photodeposition and 2) mechanical alloying using a high-speed ball mill. A library of catalysts was developed and screened for gas and aqueous phase applications, using ethanol and 4-chlorophenol as the target contaminants, respectively. Both target compounds are well studied in photocatalytic systems serve as model contaminants for this research. Synthesized catalysts were compared in terms of preparation method, type of quantum dots, and dosage of quantum dots.

Coutts, Janelle L.↗
NASA NTRS2014

Visible-Light-Responsive Catalysts Using Quantum Dot-Modified TiO2 for Air and Water Purification

Photocatalysis, the oxidation or reduction of contaminants by light-activated catalysts, utilizing titanium dioxide (TiO2) as the catalytic substrate has been widely studied for trace contaminant control in both air and water applications. The interest in this process is due primarily to its low energy consumption and capacity for catalyst regeneration. Titanium dioxide requires ultraviolet light for activation due to its relatively large band gap energy of 3.2 eV. Traditionally, Hg-vapor fluorescent light sources are used in PCO reactors; however, the use of mercury precludes the use of this PCO technology in a spaceflight environment due to concerns over crew Hg exposure. The development of a visible-light responsive (VLR) TiO2-based catalyst would eliminate the concerns over mercury contamination. Further, VLR development would allow for the use of ambient visible solar radiation or highly efficient LEDs, both of which would make PCO approaches more efficient, flexible, economical, and safe. Though VLR catalyst development has been an active area of research for the past two decades, there are few commercially available VLR catalysts. Those VLR catalysts that are commercially available do not have adequate catalytic activity, in the visible region, to make them competitive with those operating under UV irradiation. This study was initiated to develop more effective VLR catalysts through a novel method in which quantum dots (QD) consisting of narrow band gap semiconductors (e.g., CdS, CdSe, PbS, ZnSe, etc.) are coupled to TiO2 via two preparation methods: 1) photodeposition and 2) mechanical alloying using a high-speed ball mill. A library of catalysts was developed and screened for gas and aqueous phase applications using ethanol and 4-chlorophenol as the target contaminants, respectively. Both target compounds are well studied in photocatalytic systems and served as model contaminants for this research. Synthesized catalysts were compared in terms of preparation method, nature of the quantum dots, and dosage of quantum dots.

visible light↗
DOE OSTI2020

Use of high throughput screening methods to study dual-functional crosslinkable latexes

Herein, structure-formulation-property relationships for crosslinkable core-shell latexes were studied initially using high throughput screening methods and then using conventional experimental methods to determine optimum coating formulations. Two approaches that are used within coatings technology to improve sustainability are reduction in volatile organic compounds (VOC) and reduction in energy needed to cure coatings. Hybrid crosslinkable core-shell latexes have shown potential in meeting these challenges where film properties provided by particle coalescence are not sufficient and chemical crosslinking is needed. However, with numerous parameters involved such as latex particle architecture, nature and amount of crosslinker, and curing conditions, the screening of formulations via conventional coating testing methods becomes laborious and time-consuming. The results from high-throughput dye extraction and puncture testing were compared to data from conventional testing methods such as chemical resistance (MEK double rubs), pendulum hardness, and tensile testing to identify correlations. The dye extraction results have shown a strong correlation to solvent resistance and pendulum hardness data whereas toughness from puncture testing data correlated to tensile testing. Both high throughput dye extraction and puncture testing methods can be used for the accelerated screening of crosslinkable coating formulations to optimize curing parameters to achieve the best coating performance.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Gene Amplification, Laboratory Evolution, and Biosensor Screening Reveal MucK as a Terephthalic Acid Transporter in Acinetobacter baylyi ADP1

Microbial terephthalic acid (TPA) catabolic pathways are conserved among the few bacteria known to turnover this xenobiotic aromatic compound. However, to date there are few reported cases in which this pathway has been successfully expressed in heterologous hosts to impart efficient utilization of TPA as a sole carbon source. In this work, we aimed to engineer TPA conversion in Acinetobacter baylyi ADP1 via the heterologous expression of catabolic and transporter genes from a native TPA-utilizing bacterium. Specifically, we obtained ADP1-derived strains capable of growing on TPA as the sole carbon source using chromosomal insertion and targeted amplification of the tph catabolic operon from Comamonas sp. E6. Adaptive laboratory evolution was then used to improve growth on this substrate. TPA consumption rates of the evolved strains, which retained multiple copies of the tph genes, were ~0.2 g/L/h (or ~1 g TPA/g cells/h), similar to that of Comamonas sp. E6 and almost 2-fold higher than that of Rhodococcus jostii RHA1, another native TPA-utilizing strain. To evaluate TPA transport in the evolved ADP1 strains, we engineered a TPA biosensor consisting of the transcription factor TphR and a fluorescent reporter. In combination with whole-genome sequencing, the TPA biosensor revealed that transport of TPA was not mediated by the heterologous proteins from Comamonas sp. E6. Instead, the endogenous ADP1 muconate transporter MucK, a member of the major facilitator superfamily, was responsible for TPA transport in several evolved strains in which MucK variants were found to enhance TPA uptake. Furthermore, the IclR-type transcriptional regulator DcaS was identified as a repressor of mucK expression. Overall, this work presents an unexpected function of a native protein identified through gene amplification, adaptive laboratory evolution, and a combination of screening methods. This study also provides a TPA biosensor for application in ADP1 and identifies transporter variants for use in metabolic engineering applications focused on plastic upcycling of polyesters.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Bridging Experiment and Theory to Reveal Compounds in K–Zn(Cd)–Bi Systems

This study investigates the facile hydride synthesis method guided by theoretical predictions to explore the K–T–Bi (T = Zn, Cd) phase spaces. Using an adaptive genetic algorithm (AGA) and density functional theory (DFT), candidate compositions are identified for experimental validation via a facile hydrides route, permitting experimental screening of K–Zn–Bi and “empty” K–Cd–Bi systems. The previously reported KZnBi and KZn 2 Bi 2 are synthesized alongside newly discovered KCdBi and KCd 2 Bi 2 . While the AGA and DFT predict the stability of these compounds, structural predictions align with the experiment only for KZnBi and KZn 2 Bi 2 . Single-crystal X-ray structure refinements confirm that KZnBi and KZn 2 Bi 2 adopt the hexagonal ZrBeSi- and tetragonal ThCr 2 Si 2 -structure types, respectively. KCdBi has tetragonal PbClF-structure type and KCd 2 Bi 2 belongs to the ThCr 2 Si 2 -structure type. A trend based on the ratio of the metal ionic radii allows to rationalize variation in the structure types within the ATBi family (A = Li–Cs), correctly identifying KCdBi as isostructural to NaZnBi. Thermal stability studied by high-temperature powder X-ray diffraction reveals that Zn-containing compounds melt at higher temperatures (821 K for KZn 2 Bi 2 ) than Cd-containing KCd 2 Bi 2 (635 K). This study highlights the efficacy of combining rapid synthesis techniques with predictive modeling, though structural predictions show some limitations in accuracy.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Integrated data-driven and experimental approaches to accelerate lead optimization targeting SARS-CoV- 2 main protease

Identification of potential therapeutic candidates can be expedited by integrating computational modeling with domain aware machine learning (ML) approaches followed by experimental validation. Generative deep learning models have been recently developed that can generate thousands of new candidates, but their physiochemical properties are typically not optimized. Using our deep learning models and a scaffold as a starting point, we generated tens of thousands of compounds for SARS-CoV-2 M pro that preserve the core scaffold. Here we utilized and implemented several computational tools such as structural alert and toxicity analysis, high throughput virtual screening, ML-based 3D quantitative structure–activity relationships, multi-parameter optimization, and graph neural networks on libraries of generated candidates to predict biological activity and binding affinity a priori. From these collective computational results, eight promising candidates were identified and tested experimentally using Native Mass Spectrometry (MS) and FRET-based functional assays. Two compounds, with quinazoline-2-thiol and acetylpiperidine core moiety showed IC 50 values in the low micromolar range: 2.95±0.0017 µM and 3.41±0.0015 µM, respectively. The molecular dynamics simulations further highlight that binding of these compounds results in allosteric modulations in the chain B and the interface domains of the M pro . The key fragments from these top hits can be used as input for closed loop lead optimization in the integrated pipeline.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Stability of Pt Skin Intermetallic Core Catalysts and Adsorption Properties for the Oxygen Reduction Reaction

Density functional theory calculations were used to determine the stability of metal slabs consisting of a Pt surface monolayer and intermetallic supporting layers made from combinations of six transition metal elements (Pt, Fe, Co, Ni, Cu, and Ag), as a model system for Pt skin intermetallic core nanoparticle catalysts. The stability of the slabs is largely determined by strain at the interface of the Pt skin and the subsurface intermetallic, which was described by a lattice matching parameter (r). The surface charge on the Pt skin was found to be correlated with the average electronegativity (EN) of the intermetallic core, so this average EN was used as a descriptor for how the electronic coupling (or ligand effect) affects adsorption energies. A total of 46 slabs were investigated in terms of their stability, from which 10 stable slabs were selected for further studies of adsorbate binding (OOH*, O*, and OH*) that are intermediates in the oxygen reduction reaction (ORR). The correlation between all three adsorption energies and descriptors r and EN was found. Using a linear fit between our descriptors and the calculated adsorption energies, the overpotential for the ORR was obtained as a function of r and EN, from which a volcano plot was produced. The volcano peak was found at r = 0.96 or at EN = 2.025. Interestingly, neither r nor EN was a sufficient single reactivity descriptor as the data points were well off the general trend in both linear fits; this implies that both the strain effect and the ligand effect influence the adsorption energies, although they are partly correlated. The (r, EN) target peak parameters were used to screen over 241 intermetallic combinations of transition metal elements as active ORR activity. In conclusion, this analysis identified 11 intermetallic compounds which can support a Pt skin to have a high predicted ORR activity.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Bifunctionality of supported metal hydrodeoxygenation catalysts

The transition to sustainable energy relies on innovative methods to convert biomass-derived compounds into viable biofuels. In this study, the hydrodeoxygenation (HDO) of 6-undecanone is used as a model reaction to screen bifunctional catalysts, where metal sites facilitate hydrogenation and the support promotes deoxygenation, enabling high conversion and selectivity toward desirable alkanes for biofuel production. This reaction is particularly relevant as it represents a critical step in upgrading volatile fatty acids, derived from biomass, into long-chain hydrocarbons suitable for fuel applications. By examining a range of metals (nickel, cobalt, and tin) on different supports, it is revealed that the choice of metal–support combination is critical to catalyst performance. Zeolite beta's 3D microporous structure and adjustable acidity provide an ideal environment for fine-tuning metal–support interactions (MSIs), which are essential for balancing deoxygenation with alkane isomerization, a desirable trait for biofuels.

09 BIOMASS FUELS↗
DOE OSTI2025

Combining MicroED and native mass spectrometry for structural discovery of enzyme–small molecule complexes

With the goal of accelerating the discovery of small molecule–protein complexes, we leverage fast, low-dose, event-based electron counting microcrystal electron diffraction (MicroED) data collection and native mass spectrometry. This approach, which we term electron diffraction with native mass spectrometry (ED-MS), allows assignment of protein target structures bound to ligands with data obtained from crystal slurries soaked with mixtures of known inhibitors and crude biosynthetic reactions. This extends to libraries of printed ligands dispensed directly onto TEM grids for later soaking with microcrystal slurries, and complexes with noncovalent ligands. ED-MS resolves structures of the natural product, epoxide-based cysteine protease inhibitor E-64, and its biosynthetic analogs bound to the model cysteine protease, papain. It further identifies papain binding to its preferred natural products, by showing that two analogs of E-64 outcompete others in binding to papain crystals, and by detecting papain bound to E-64 and an analog from crude biosynthetic reactions, without purification. ED-MS also resolves binding of the CTX-M-14 β-lactamase, a target of active drug development, to the non-β-lactam inhibitor, avibactam, alone or in a cocktail of unrelated compounds. These results illustrate the utility of ED-MS for natural product ligand discovery and for structure-based screening of small molecule binders to macromolecular targets, promising utility for drug discovery.

MicroED↗
DOE OSTI2025

Data imbalance in drug response prediction: multi-objective optimization approach in deep learning setting

Abstract Drug response prediction (DRP) methods tackle the complex task of associating the effectiveness of small molecules with the specific genetic makeup of the patient. Anti-cancer DRP is a particularly challenging task requiring costly experiments as underlying pathogenic mechanisms are broad and associated with multiple genomic pathways. The scientific community has exerted significant efforts to generate public drug screening datasets, giving a path to various machine learning models that attempt to reason over complex data space of small compounds and biological characteristics of tumors. However, the data depth is still lacking compared to application domains like computer vision or natural language processing domains, limiting current learning capabilities. To combat this issue and improves the generalizability of the DRP models, we are exploring strategies that explicitly address the imbalance in the DRP datasets. We reframe the problem as a multi-objective optimization across multiple drugs to maximize deep learning model performance. We implement this approach by constructing Multi-Objective Optimization Regularized by Loss Entropy loss function and plugging it into a Deep Learning model. We demonstrate the utility of proposed drug discovery methods and make suggestions for further potential application of the work to achieve desirable outcomes in the healthcare field.

Biochemistry & Molecular Biology↗
DOE OSTI2025

A machine learning approach to quantify degradation of nuclear fuels and the effects of fission products

Nuclear fuel performance is critically dependent on understanding the evolution of fuel properties under operational conditions, a complex challenge driven by chemical changes and substantial radiation damage during fission. Traditionally, property evolution has been determined via empirical data collected following irradiation. However, these empirical correlations are limited in their applicability beyond the specific conditions in which they were obtained. This study explores a novel approach to address this challenge by applying materials informatics to develop a machine learning random forest (ML-RF) model that captures the effects of fission products on fuel compounds. The model predicts formation enthalpy (ΔH f ) by leveraging extensive quantum materials property data and correlating it with material descriptors such as composition, atomic and site features, and crystal lattice properties. This ML-RF model enables rapid interpolation across the compositional and structural spaces covered by the training data, thus supporting high-throughput screening and energetic ranking of candidate phases. The model demonstrates the ability to predict ΔH f with a mean absolute error (MAE) of approximately 0.1 to 0.2 eV/atom across a wide range of compounds, including key nuclear fuel systems (U-O, U-N, U-C, U-Si, and U-Mo). For example, it was used to assess shifts in stoichiometry for UO 2 (O/M) and UN (N/M) fuels, revealing their distinct tendencies in chemical potential variation and enabling preliminary convex hull analyses. Furthermore, the model provides insights into how individual fission products affect fuel properties. Results indicate that larger fission products (e.g., Nd, Pu, Ce) have a more pronounced impact on UO 2 , while lighter ones (e.g., Zr) strongly influence UN. Here, the model developed in this work can be used to support the Accelerated Fuel Qualification approach by facilitating preliminary evaluations prior to extensive materials modeling and experimentation. To this end, the trained model has been made available to the fuel community to support ongoing fuel development efforts.

Accelerated fuel qualification↗
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↗