Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “LABELLING”

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 1,171 records · Page 65

Heterologous expression and solution structure of defensin from lentil Lens culinaris

Highlights: • Lentil defensin Lc-def and its {sup 15}N-labeled analog were overexpressed in E. coli. • Lc-def is active against fungi, but does not inhibit growth of G+ and G− bacteria. • Lc-def spatial structure involves triple-stranded β-sheet and α-helix (CSαβ motif). • Lc-def is able to bind to anionic lipid vesicles under low-salt conditions. • NMR data revealed significant μs–ms mobility in the loops 1 and 3 of Lc-def. - Abstract: A new defensin Lc-def, isolated from germinated seeds of the lentil Lens culinaris, has molecular mass 5440.4 Da and consists of 47 amino acid residues. Lc-def and its {sup 15}N-labeled analog were overexpressed in Escherichia coli. Antimicrobial activity of the recombinant protein was examined, and its spatial structure, dynamics, and interaction with lipid vesicles were studied by NMR spectroscopy. It was shown that Lc-def is active against fungi, but does not inhibit the growth of Gram-positive and Gram-negative bacteria. The peptide is monomeric in aqueous solution and contains one α-helix and triple-stranded β-sheet, which form cysteine-stabilized αβ motif (CSαβ) previously found in other plant defensins. The sterically neighboring loop1 and loop3 protrude from the defensin core and demonstrate significant mobility on the μs–ms timescale. Lc-def does not bind to the zwitterionic lipid (POPC) vesicles but interacts with the partially anionic (POPC/DOPG, 7:3) membranes under low-salt conditions. The Lc-def antifungal activity might be mediated through electrostatic interaction with anionic lipid components of fungal membranes.

60 APPLIED LIFE SCIENCES↗

Comparison of the biological effects of {sup 18}F at different intracellular levels

Highlights: • We estimated the inductions of DNA DSB in cell treated with {sup 18}F-FDG. • We found that inductions of DNA DSB are dependent on accumulation of {sup 18}F in cell. • Accumulation of {sup 18}F in cell may be indispensable for risk estimation of PET. - Abstract: We herein examined the biological effects of cells treated with {sup 18}F labeled drugs for positron emission tomography (PET). The relationship between the intracellular distribution of {sup 18}F and levels of damaged DNA has yet to be clarified in detail. We used culture cells (Chinese Hamster Ovary cells) treated with two types of {sup 18}F labeled drugs, fluorodeoxyglucose (FDG) and fluorine ion (HF). FDG efficiently accumulated in cells, whereas HF did not. To examine the induction of DNA double strand breaks (DSB), we measured the number of foci for 53BP1 that formed at the site of DNA DSB. The results revealed that although radioactivity levels were the same, the induction of 53BP1 foci was stronger in cells treated with {sup 18}F-FDG than in those treated with {sup 18}F-HF. The clonogenic survival of cells was significantly lower with {sup 18}F-FDG than with {sup 18}F-HF. We concluded that the efficient accumulation of {sup 18}F in cells led to stronger biological effects due to more severe cellular lethality via the induction of DNA DSB.

60 APPLIED LIFE SCIENCES↗

Lateral diffusion contributes to FRET from lanthanide-tagged membrane proteins

Diffusion can enhance Förster resonance energy transfer (FRET) when donors or acceptors diffuse distances that are similar to the distances separating them during the donor's excited state lifetime. Lanthanide donors remain in the excited state for milliseconds, which makes them useful for time-resolved FRET applications but also allows time for diffusion to enhance energy transfer. Here we show that diffusion dramatically enhances FRET between membrane proteins labeled with lanthanide donors. This phenomenon complicates interpretation of experiments that use long-lived donors to infer association or proximity of mobile membrane proteins, but also offers a method of monitoring diffusion in membrane domains in real time in living cells. - Highlights: • Diffusion enhances TR-FRET from membrane proteins labeled with lanthanide donors. • Diffusion-dependent FRET can overshadow FRET due to oligomerization or clustering. • FRET studies using lanthanide-tagged membrane proteins should consider diffusion. • FRET from lanthanide donors can be used to monitor membrane protein diffusion.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Ca{sup 2+}/calmodulin-dependent protein kinase phosphatase (CaMKP/PPM1F) interacts with neurofilament L and inhibits its filament association

Ca{sup 2+}/calmodulin-dependent protein kinase phosphatase (CaMKP/PPM1F) is a Ser/Thr phosphatase that belongs to the PPM family. Growing evidence suggests that PPM phosphatases including CaMKP act as a complex with other proteins to regulate cellular functions. In this study, using the two-dimensional far-western blotting technique with digoxigenin-labeled CaMKP as a probe, in conjunction with peptide mass fingerprinting analysis, we identified neurofilament L (NFL) as a CaMKP-binding protein in a Triton-insoluble fraction of rat brain. We confirmed binding of fluorescein-labeled CaMKP (F-CaMKP) to NFL in solution by fluorescence polarization. The analysis showed that the dissociation constant of F-CaMKP for NFL is 73 ± 17 nM (n = 3). Co-immunoprecipitation assay using a cytosolic fraction of NGF-differentiated PC12 cells showed that endogenous CaMKP and NFL form a complex in cells. Furthermore, the effect of CaMKP on self-assembly of NFL was examined. Electron microscopy revealed that CaMKP markedly prevented NFL from forming large filamentous aggregates, suggesting that CaMKP-binding to NFL inhibits its filament association. These findings may provide new insights into a novel mechanism for regulating network formation of neurofilaments during neuronal differentiation. - Highlights: • NFL was identified as a CaMKP-binding protein in an insoluble fraction of rat brain. • CaMKP bound to NFL in solution with a K{sub d} value of 73 ± 17 nM. • A CaMKP-NFL complex was found in NGF-differentiated PC12 cells. • CaMKP-binding to NFL inhibited its filament association. • CaMKP may regulate network formation of neurofilaments in neurons.

60 APPLIED LIFE SCIENCES↗

Fluorescent 6-amino-6-deoxyglycoconjugates for glucose transporter mediated bioimaging

Two novel fluorescent bioprobes, namely, 6N-Gly-Cy3 and 6N-Gly-Cy5, were designed and synthesized for real-time glucose transport imaging as well as potentially useful tracer for galactokinase metabolism. The structure of the bioprobes was fully characterized by {sup 1}H NMR, {sup 13}C NMR, IR, and HRMS. The fluorescence properties, glucose transporter (GLUT) specificity, and the quenching and safety profiles were studied. The cellular uptake of both bioprobes was competitively diminished by D-glucose, 2-deoxy-D-glucose and GLUT specific inhibitor in a dose-dependent manner in human colon cancer cells (HT29). Comparison study results revealed that the 6N-derived bioprobes are more useful for real-time imaging of cell-based glucose uptake than the structurally similar fluorescent tracer 6-NBDG which was not applicable under physiological conditions. The up to 96 h long-lasting quenching property of 6N-Gly-Cy5 in HT29 suggested the potential applcability of the probe for cell labeling in xenograft transplantation as well as in vivo animal imaging studies. - Highlights: • Cy-3 and Cy-5 derived fluorescent 6-amino-6-deoxyglycoconjugates were prepared for glucose transporter mediated bioimaging. • The cellular uptake of the probes was inhibited by natural GLUT substrates and inhibitor. • The probes are useful for real-time imaging of cell-based glucose uptake under physiological conditions. • The probes showed up to 96 h long-lasting quenching profile in labeled cancer cells.

60 APPLIED LIFE SCIENCES↗

Lead discovery for mammalian elongation of long chain fatty acids family 6 using a combination of high-throughput fluorescent-based assay and RapidFire mass spectrometry assay

A high-throughput RapidFire mass spectrometry assay is described for elongation of very long-chain fatty acids family 6 (Elovl6). Elovl6 is a microsomal enzyme that regulates the elongation of C12-16 saturated and monounsaturated fatty acids. Elovl6 may be a new therapeutic target for fat metabolism disorders such as obesity, type 2 diabetes, and nonalcoholic steatohepatitis. To identify new Elovl6 inhibitors, we developed a high-throughput fluorescence screening assay in 1536-well format. However, a number of false positives caused by fluorescent interference have been identified. To pick up the real active compounds among the primary hits from the fluorescence assay, we developed a RapidFire mass spectrometry assay and a conventional radioisotope assay. These assays have the advantage of detecting the main products directly without using fluorescent-labeled substrates. As a result, 276 compounds (30%) of the primary hits (921 compounds) in a fluorescence ultra-high-throughput screening method were identified as common active compounds in these two assays. It is concluded that both methods are very effective to eliminate false positives. Compared with the radioisotope method using an expensive {sup 14}C-labeled substrate, the RapidFire mass spectrometry method using unlabeled substrates is a high-accuracy, high-throughput method. In addition, some of the hit compounds selected from the screening inhibited cellular fatty acid elongation in HEK293 cells expressing Elovl6 transiently. This result suggests that these compounds may be promising lead candidates for therapeutic drugs. Ultra-high-throughput fluorescence screening followed by a RapidFire mass spectrometry assay was a suitable strategy for lead discovery against Elovl6. - Highlights: • A novel assay for elongation of very-long-chain fatty acids 6 (Elovl6) is proposed. • RapidFire mass spectrometry (RF-MS) assay is useful to select real screening hits. • RF-MS assay is proved to be beneficial because of its high-throughput and accuracy. • A combination of fluorescent and RF-MS assays is effective for Elovl6 inhibitors.

60 APPLIED LIFE SCIENCES↗

Hydroxyl radical generation by dissociation of water molecules during 1.65 MHz frequency ultrasound irradiation under aerobic conditions

The dissociation of water molecules by ultrasound irradiation under aerobic conditions was demonstrated experimentally. To be able to detect the dissociation of water molecules, we performed the ultrasound irradiation of {sup 17}O-labelled water (H{sub 2}{sup 17}O) under aerobic conditions. The hydroxyl and hydrogen radicals generated during the ultrasound irradiation process were trapped with 5,5-dimethyl-1-pyrroline-N-oxide (DMPO), and electron spin resonance (ESR) spectroscopy was performed on the DMPO spin adducts. In the ESR spectrum, a 15-line signal attributable to the trapping of the hydroxyl radicals containing {sup 17}O ({sup 17}OH radicals) by DMPO together with a 4-line signal attributable to the trapping of the hydroxyl radicals containing {sup 16}O ({sup 16}OH radicals) by DMPO were observed. The generation of {sup 17}OH radicals indicated that H{sub 2}{sup 17}O was dissociated by the sonolysis process under aerobic conditions. On the other hand, the ESR signal attributable to the trapping of hydrogen radicals by DMPO was not observed, suggesting that hydrogen radicals were not generated during the dissociation of water molecules. - Highlights: • OH radicals was observed in air-saturated water under 1.65 MHz ultrasound. • The OH radicals were originated from both H{sub 2}O and O{sub 2.} • The H{sub 2}O sonolysis was detected by using {sup 17}O-labelled water.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Pharmacological characterization of synthetic serine palmitoyltransferase inhibitors by biochemical and cellular analyses

Highlights: • Tetrahydropyrazolopyridine derivatives were characterized as SPT inhibitors. • The SPT inhibitors exhibited dual inhibitory activities against SPT2 and SPT3. • A SPT inhibitor bound to the SPT2 or SPT3 subunit in SPT enzyme complexes. • One of the SPT inhibitors exhibited a significantly slow dissociation from SPT. • The SPT inhibitors suppressed ceramide content in a non-small-cell lung cancer cell line. Human serine palmitoyltransferase (SPT) is a PLP-dependent enzyme residing in the endoplasmic reticulum. It catalyzes the synthesis of 3-ketodihydrosphingosine (3-KDS) from the substrates palmitoyl-CoA and l-serine. It is a rate-limiting enzyme for sphingolipid synthesis in cells. In the present study, we characterized and pharmacologically profiled a series of tetrahydropyrazolopyridine derivatives that potently inhibit human SPT enzymatic activity, including two cell-active derivatives and one fluorescent-labelled derivative. These SPT inhibitors exhibited dual inhibitory activities against SPT2 and SPT3. We used a fluorescent-labelled probe to molecularly assess the inhibitory mechanism and revealed its binding to the SPT2 or SPT3 subunit in the small subunit (ss) SPTa/SPT1/SPT2/or ssSPTa/SPT1/SPT3 functional complexes. One of the SPT inhibitors exhibited a significantly slow dissociation from the SPT complex. We confirmed that our SPT inhibitors suppressed ceramide content in non-small-cell lung cancer cell line, HCC4006, by performing a target engagement analysis. The potency of ceramide reduction correlated to that observed in a recombinant SPT2 enzyme assay. We thus elucidated and provided a fundamental understanding of the molecular mode of action of SPT inhibitors and developed potent, cell-active SPT inhibitors that can be used to clarify the biological function of SPT.

60 APPLIED LIFE SCIENCES↗

Probing the light scattering properties of insulin secretory granules in single live cells

Light scattering was recently demonstrated to serve as an intrinsic indicator for pancreatic islet cell mass and secretion. The insulin secretory granule (ISG), in particular, was proposed to be a reasonable candidate as the main intracellular source of scattered light due to the densely-packed insulin semi-crystal in the granule lumen. This scenario, if confirmed, would in principle open new perspectives for label-free single-granule imaging, tracking, and analysis. Contrary to such expectations, here we demonstrate that ISGs are not a primary source of scattering in primary human β-cells, as well as in immortalized β-like cells, quantitatively not superior to other intracellular organelles/structures, such as lysosomes and internal membranes. This result is achieved through multi-channel imaging of scattered light along with fluorescence arising from selectively-labelled ISGs. Co-localization and spatiotemporal cross-correlation analysis is performed on these signals, and compared among different cell lines. Obtained results suggest a careful re-thinking of the possibility to exploit intrinsic optical properties originating from ISGs for single-granule imaging purposes.

60 APPLIED LIFE SCIENCES↗

Host analysis-guided selection and targeted engineering (HASTE) of Lipomyces tetrasporus for the conversion of CO2-derived feedstocks

Efficient and cost-competitive bioproduction calls for utilizing CO2-derived feedstocks, such as products from electro-reduction of CO2 and hydrolysate from lignocellulosic biomass. However, efficiently using all their carbon components, including acetate, glucose, and xylose, remains a challenge. Here, we characterize Lipomyces tetrasporus, a novel, robust yeast strain capable of effectively assimilating these carbon sources. We used an integrated systems biology approach combining ¹³C metabolic flux analysis, dynamic labeling experiments, and RNA sequencing. We conducted the first metabolic flux analysis for glucose, xylose, and acetate catabolism in this species. Dynamic labeling revealed a highly active TCA cycle during acetate metabolism, evidenced by rapid citrate and malate accumulation. The strain demonstrated strong NADH/NADPH production and acetyl-CoA synthase activity. Using insights and gene targets from this analysis, we engineered L. tetrasporus for malate production. The engineered strain produced 7.5 g/L malic acid (0.25 g/g yield) in shake flasks with glucose-acetate media and 28.8 g/L malic acid at a yield of 0.20 g/g in fed-batch mode with corn-stover hydrolysate. Together, these insights and rational strain engineering establish L. tetrasporus as a versatile, Crabtree-negative platform that is an energy-CO2-bioproduction nexus for channeling CO2 carbon into value-added bioproducts.

Xiao, Zhengyang↗

CdSe/ZnS quantum dots based electrochemical immunoassay for the detection of phosphorylated bovine serum albumin

A CdSe/ZnS quantum dot (QD) based electrochemical immunoassay of phosphorylated bovine serum albumin as a protein biomarker is presented. The QDs were used as labels and were conjugated with the secondary anti-phosphoserine antibody in a heterogeneous sandwich immunoassay. First, the primary BSA antibody was immobilized on polystyrene microwells, followed by the addition of BSA-OP. After that, the QD-labeled anti-phosphoserine antibody was added into microwells for immunorecognition. Finally, the bound QD was dissolved in an acid-dissolution step and was detected by electrochemical stripping analysis. The measured current responses were proportional to the concentration of BSA-OP. Under optimal conditions, the voltammetric response was linear over the range of 0.5 - 500 ng mL-1 of BSA-OP, with a detection limit of 0.5 ng mL-1 at a deposition potential of -1.2 V for 120 s. It also shows good reproducibility with a relative standard deviation of 8.6% of six times determination of 25 ng mL-1 of BSA-OP. This QD-based electrochemical immunoassay offers great promise for simple and cost-effective analysis of protein biomarkers.

59 BASIC BIOLOGICAL SCIENCES↗

[11C]PR04.MZ, a promising DAT ligand for low concentration imaging: synthesis, efficient 11C-0-methylation and initial small animal PET studies

PR04.MZ was designed as a highly selective dopamine transporter inhibitor, derived from natural cocaine. Its binding profile indicates that [{sup 11}C]PR04.MZ may be suited as a PET radioligand for the non-invasive exploration of striatal and extrastriatal DAT populations. As a key feature, its structural design facilitates both, labelling with fluorine-18 at its terminally fluorinated butynyl moiety and carbon-11 at its methyl ester function. The present report concerns the efficient [{sup 11}C]MeI mediated synthesis of [{sup 11}C]PR04.MZ from an O-desmethyl precursor trifluoroacetic acid salt with Rb{sub 2}CO{sub 3} in DMF in up to 95 {+-} 5% labelling yield. A preliminary {mu}PET-experiment demonstrates the reversible, highly specific binding of [{sup 11}C]PR04.MZ in the brain of a male Sprague-Dawley rat.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Effective properties of a fly ash geopolymer: Synergistic application of X-ray synchrotron tomography, nanoindentation, and homogenization models

Microstructural and micromechanical investigation of a fly ash-based geopolymer using: (i) synchrotron X-ray tomography (XRT) to determine the volume fraction and tortuosity of pores that are influential in fluid transport, (ii) mercury intrusion porosimetry (MIP) to capture the volume fraction of smaller pores, (iii) scanning electron microscopy (SEM) combined with multi-label thresholding to identify and characterize the solid phases in the microstructure, and (iv) nanoindentation to determine the component phase elastic properties using statistical deconvolution, is reported in this paper. The phase volume fractions and elastic properties are used in multi-step mean field homogenization (Mori–Tanaka and double inclusion) models to determine the homogenized macroscale elastic modulus of the composite. The homogenized elastic moduli are in good agreement with the flexural elastic modulus determined on macroscale paste beams. The combined use of microstructural and micromechanical characterization tools at multiple scales provides valuable information towards the material design of fly ash geopolymers.

36 MATERIALS SCIENCE↗

Changes in lipid bilayer structure caused by the helix-to-sheet transition of an HIV-1 gp41 fusion peptide derivative

HIV-1, like other enveloped viruses, undergoes fusion with the cell membrane to infect it. Viral coat proteins are thought to bind the virus to the membrane and actively fuse the viral and cellular membranes together. The actual molecular mechanism of fusion is challenging to visualize, resulting in the use of model systems. In this paper, the bilayer curvature modifying properties of a synthetic variant of the HIV-1 gp41 fusion peptide with lipid bilayer vesicles composed of a mixture of dimyristoyl phosphatidylcholine (DMPC) and dimyristoyl phosphatidylserine (DMPS) were studied. In 7:3 DMPC:DMPS vesicles made with deuterium-labeled DMPC, the peptide was observed to undergo a concentration-dependent conformational transition between an α-helix and an antiparallel β-sheet. Through the use of small-angle neutron scattering (SANS) and selective deuterium labeling, it was revealed that conformational transition of the peptide is also accompanied by a transition in the structure of the lipid bilayer. In addition to changes in the distribution of the lipid between the leaflets of the vesicle, the SANS data are consistent with two regions having different thicknesses. Finally, of the two different bilayer structures, the one corresponding to the smaller area fraction, being ~8% of the vesicle area, is much thicker than the remainder of the vesicle, which suggests that there are regions of localized negative curvature similar to what takes place at the point of contact between two membranes immediately preceding fusion.

59 BASIC BIOLOGICAL SCIENCES↗

Optimized anion exchange column isolation of zirconium-89 (89Zr) from yttrium cyclotron target: Method development and implementation on an automated fluidic platform

Zirconium-89 (89Zr), produced by the (p,n) reaction from naturally monoisotopic yttrium (natY), is a promising positron emitting isotope for immunoPET imaging. Its long half-life of 78.4 h is sufficient for evaluating slow physiological processes. A prototype automated fluidic system, coupled to on-line and in-line detectors, has been constructed to facilitate development of new 89Zr purification methodologies. The highly reproducible reagent delivery platform and near-real time monitoring of column effluents allows for efficient method optimization. The separation of Zr from dissolved Y metal targets was evaluated using several anion exchange resins. Each resin was evaluated against its ability to quantitatively capture Zr from a load solution that is high in dissolved Y. The most appropriate anion exchange resin for this application was identified, and the separation method was optimized. The method is capable of a high Y decontamination factor (>105) and has been shown to separate Fe, an abundant contaminant in Y foils, from the 89Zr elution fraction. Finally, the performance of the method was evaluated using cyclotron bombarded Y foil targets. The separation method was shown to achieve >95% recovery of the 89Zr present in the foils. The 89Zr eluent, however, was in a chemical matrix not immediately conducive to labeling onto proteins. The main intent of this study was to develop a tandem column 89Zr purification process, wherein the anion exchange column method described here is the first separation in a dual-column purification process.

Zirconium-89, ImmunoPET, medical isotope, automati↗

Separable physics-informed DeepONet: Breaking the curse of dimensionality in physics-informed machine learning

The deep operator network (DeepONet) has shown remarkable potential in solving partial differential equations (PDEs) by mapping between infinite-dimensional function spaces using labeled datasets. However, in scenarios lacking labeled data, the physics-informed DeepONet (PI-DeepONet) approach, which utilizes the residual loss of the governing PDE to optimize the network parameters, faces significant computational challenges, particularly due to the curse of dimensionality. This limitation has hindered its application to high-dimensional problems, making even standard 3D spatial with 1D temporal problems computationally prohibitive. Additionally, the computational requirement increases exponentially with the discretization density of the domain. Here, to address these challenges and enhance scalability for high-dimensional PDEs, we introduce the Separable physics-informed DeepONet (Sep-PI-DeepONet). This framework employs a factorization technique, utilizing sub-networks for individual one-dimensional coordinates, thereby reducing the number of forward passes and the size of the Jacobian matrix required for gradient computations. By incorporating forward-mode automatic differentiation (AD), we further optimize computational efficiency, achieving linear scaling of computational cost with discretization density and dimensionality, making our approach highly suitable for high-dimensional PDEs. We demonstrate the effectiveness of Sep-PI-DeepONet through three benchmark PDE models: the viscous Burgers’ equation, Biot’s consolidation theory, and a parameterized heat equation. Our framework maintains accuracy comparable to the conventional PI-DeepONet while reducing training time by two orders of magnitude. Notably, for the heat equation solved as a 4D problem, the conventional PI-DeepONet was computationally infeasible (estimated 289.35 h), while the Sep-PI-DeepONet completed training in just 2.5 h. These results underscore the potential of Sep-PI-DeepONet in efficiently solving complex, high-dimensional PDEs, marking a significant advancement in physics-informed machine learning.

Neural operator↗

Overcoming small minirhizotron datasets using transfer learning

Minirhizotron technology is widely used to study root growth and development. Yet, standard approaches for tracing roots in minirhiztron imagery is extremely tedious and time consuming. Machine learning approaches can help to automate this task. However, lack of enough annotated training data is a major limitation for the application of machine learning methods. Transfer learning is a useful technique to help with training when available datasets are limited. In this paper, we investigated the effect of pre-trained features from the massives-cale, irrelevant ImageNet dataset and a relatively moderate-scale, but relevant peanut root dataset on switchgrass root imagery segmentation applications. We compiled two minirhizotron image datasets to accomplish this study: one with 17,550 peanut root images and another with 28 switchgrass root images. Both datasets were paired with manually labeled ground truth masks. Deep neural networks based on the U-net architecture were used with different pre-trained features as initialization for automated, precise pixel-wise root segmentation in minirhizotron imagery. We observed that features pre-trained on a closely related but relatively moderate size dataset like our peanut dataset were more effective than features pre-trained on the large but unrelated ImageNet dataset. Here, we achieved high quality segmentation on peanut root dataset with 99.04% accuracy at the pixel-level and overcame errors in human-labeled ground truth masks. By applying transfer learning technique on limited switchgrass dataset with features pre-trained on peanut dataset, we obtained 99% segmentation accuracy in switchgrass imagery using only 21 images for training (fine tuning). Furthermore, the peanut pre-trained features can help the model converge faster and have much more stable performance.

59 BASIC BIOLOGICAL SCIENCES↗

Generative learning for slow manifolds and bifurcation diagrams

In dynamical systems characterized by separation of time scales, the approximation of so called “slow manifolds”, on which the long term dynamics lie, is a useful step for model reduction. Initializing on such slow manifolds is a useful step in modeling, since it circumvents fast transients, and is crucial in multiscale algorithms (like the equation-free approach) alternating between fine scale (fast) and coarser scale (slow) simulations. In a similar spirit, when one studies the infinite time dynamics of systems depending on parameters, the system attractors (e.g., its steady states) lie on bifurcation diagrams (curves for one-parameter continuation, and more generally, on manifolds in state parameter space. Sampling these manifolds gives us representative attractors (here, steady states of ODEs or PDEs) at different parameter values. Algorithms for the systematic construction of these manifolds (slow manifolds, bifurcation diagrams) are required parts of the “traditional” numerical nonlinear dynamics toolkit. In more recent years, as the field of Machine Learning develops, conditional score-based generative models (cSGMs) have been demonstrated to exhibit remarkable capabilities in generating plausible data from target distributions that are conditioned on some given label. It is tempting to exploit such generative models to produce samples of data distributions (points on a slow manifold, steady states on a bifurcation surface) conditioned on (consistent with) some quantity of interest (QoI, observable). In this work, we present a framework for using cSGMs to quickly (a) initialize on a low-dimensional (reduced-order) slow manifold of a multi-time-scale system consistent with desired value(s) of a QoI (a “label”) on the manifold, and (b) approximate steady states in a bifurcation diagram consistent with a (new, out-of-sample) parameter value. This conditional sampling can help uncover the geometry of the reduced slow-manifold and/or approximately “fill in” missing segments of steady states in a bifurcation diagram. Finally, the quantity of interest, which determines how the sampling is conditioned, is either known a priori or identified using manifold learning-based dimensionality reduction techniques applied to the training data.

Dynamical systems↗