Search for: All records

Abstract Adhesive bonding of composite materials has become increasingly crucial for advanced engineering applications, offering unique advantages for lightweight and high-performance designs. This study presents a novel framework, physics-informed failure mode proportion prediction (PIFMP) model, for predicting failure mode proportions in composite adhesive joints, addressing critical gaps in understanding mixed-mode failure behaviors. In contrast to conventional approaches that focus solely on force or stress prediction, this research integrates important parameters from multistage manufacturing processes (MMPs) and simulation data into a physics-informed machine learning (PIML) framework, enabling proactive failure prediction and design optimization. The proposed framework unifies data-driven machine learning models with features derived from finite element analysis (FEA), incorporating cohesive zone modeling (CZM) to capture the physical dynamics of adhesive behavior under lap shearing. By embedding FEA-based physics features into the machine learning process and leveraging a time-series transformer model to analyze the temporal progression of interfacial damage and separation, the framework ensures predictive accuracy and physics-informed consistency, enabling precise analysis of failure mechanisms. The empirical study validates the effectiveness and the reliability of the framework, demonstrating enhanced predictive performance through cross-validation. The work establishes a foundational approach for failure analysis and provides a robust basis for future advancements. 

Metalloproteins tune the electronic properties of metal active sites through a combination of primary and secondary coordination sphere effects (PCS and SCS) to efficiently perform an array of redox chemistry, including electron transfer (ET) and catalysis. A major influence of these effects is the anisotropic spatial distribution of redox-active molecular orbitals (RAMOs), which in turn dictates redox chemistry of the metalloproteins. While much progress has been made in understanding the SCS effects on RAMOs in individual native metalloproteins, it has been difficult to experimentally examine the influence of the same SCS effects on RAMOs with different spatial distributions. Taking advantage of our recent studies of SCS effect on blue copper azurin from Pseudomonas aeruginosa (Blue CuAz) and its M121H/H46E variant that closely mimic the red copper protein (Red CuAz), in which their RAMOs are dominated by either Cu-S or Cu-S interactions, respectively, we herein compare and contrast how the same SCS modifications impact the electronic and geometric structures of blue and red Cu center in the same protein scaffold. Specifically, we expand our understanding of unpaired electron distribution at the Cu-binding site of Red CuAz and its SCS N47S, F114P, and F114N mutations using 1H and 14N electron-nuclear double resonance (ENDOR) spectroscopy, and then further combine these data sets with recent studies and DFT calculations to provide insight into how these mutations differentially (or similarly) impact electronic structure in Red vs. Blue CuAz. We find that electrostatics produce similar effects in both Red and Blue CuAz, where the introduction of dipole moments in the vicinity of Cu and S produce changes in spin density distribution and of the same sign and comparable magnitude. However, disruption of H-bonding with S through the F114P mutation leads to opposing effects in Red vs. Blue CuAz, which we propose arise from differences in the conformation of Cys112 sidechain adapted in the absence the stabilizing SC112•••H-N backbone interaction. 

Type 1 copper (T1Cu) centers are crucial in biological electron transfer (ET) processes, exhibiting a wide range of reduction potentials (E°′T1Cu) to match their redox partners and optimize ET rates. While tuning E°′T1Cu in mononuclear T1Cu proteins like azurin has been successful, it is more difficult for multicopper oxidases. Specifically, while replacing axial methionine to leucine in azurin increased its E°′T1Cu by ~100 mV, the corresponding M298L mutation in small laccase from Streptomyces coelicolor (SLAC) unexpectedly decreased its E°′T1Cu by 12 mV. X-ray crystallography revealed two axial water molecules in M298L-SLAC, leading to the decrease of E°′T1Cu due to decreased hydrophobicity. Structural alignment and molecular dynamics simulation indicated a key difference in T1Cu axial loop position, leading to the different outcome upon methionine to leucine mutation. Based on structural analyses, we introduced additional F195L and I200F mutations, leading to partial removal of axial waters, a 122-mV increase in E°′T1Cu, and a 7-fold increase in kcat/KM from M298L-SLAC. These findings highlight the complexity of tuning E°′T1Cu in multicopper oxidases and provide valuable insights into how structure-based protein engineering can contribute to the broader understanding of T1Cu center, E°′T1Cu and reactivity tuning for applications in solar energy transfer, fuel cells, and bioremediation. 

Genetically encoded fluorescent protein and fluorogenic RNA sensors are indispensable tools for imaging biomolecules in cells. To expand the toolboxes and improve the generalizability and stability of this type of sensor, we report herein a genetically encoded fluorogenic DNA aptamer (GEFDA) sensor by linking a fluorogenic DNA aptamer for dimethylindole red with an ATP aptamer. The design enhances red fluorescence by 4-fold at 650 nm in the presence of ATP. Additionally, upon dimerization, it improves the signal-to-noise ratio by 2–3 folds. We further integrated the design into a plasmid to create a GEFDA sensor for sensing ATP in live bacterial and mammalian cells. This work expanded genetically encoded sensors by employing fluorogenic DNA aptamers, which offer enhanced stability over fluorogenic proteins and RNAs, providing a novel tool for real-time monitoring of an even broader range of small molecular metabolites in biological systems. 

Low-latency and low-power edge AI is crucial for Virtual Reality and Augmented Reality applications. Recent advances demonstrate that hybrid models, combining convolution layers (CNN) and transformers (ViT), often achieve a superior accuracy/performance tradeoff on various computer vision and machine learning (ML) tasks. However, hybrid ML models can present system challenges for latency and energy efficiency due to their diverse nature in dataflow and memory access patterns. In this work, we leverage architecture heterogeneity from Neural Processing Units (NPU) and Compute-In-Memory (CIM) and explore diverse execution schemas to efficiently execute these hybrid models. We introduce H4H-NAS, a two-stage Neural Architecture Search (NAS) framework to automate the design of efficient hybrid CNN/ViT models for heterogeneous edge systems featuring both NPU and CIM. We propose a two-phase incremental supernet training in our NAS framework to resolve gradient conflicts between sampled subnets caused by different types of blocks in a hybrid model search space. Our H4H-NAS approach is also powered by a performance estimator built with NPU performance results measured on real silicon, and CIM performance based on industry IPs. H4H-NAS searches hybrid CNN-ViT models with fine granularity and achieves significant (up to 1.34%) top-1 accuracy improvement on ImageNet. Moreover, results from our algorithm/hardware co-design reveal up to 56.08% overall latency and 41.72% energy improvements by introducing heterogeneous computing over baseline solutions. Overall, our framework guides the design of hybrid network architectures and system architectures for NPU+CIM heterogeneous systems. 

Abstract The sensitivity and responsiveness of living cells to environmental changes are enabled by dynamic protein structures, inspiring efforts to construct artificial supramolecular protein assemblies. However, despite their sophisticated structures, designed protein assemblies have yet to be incorporated into macroscale devices for real-life applications. We report a 2D crystalline protein assembly of^C98/E57/E66L-rhamnulose-1-phosphate aldolase (^CEERhuA) that selectively blocks or passes molecular species when exposed to a chemical trigger.^CEERhuA crystals are engineered via cobalt(II) coordination bonds to undergo a coherent conformational change from a closed state (pore dimensions <1 nm) to an ajar state (pore dimensions ~4 nm) when exposed to an HCN(g) trigger. When layered onto a mesoporous silicon (pSi) photonic crystal optical sensor configured to detect HCN_(g), the 2D^CEERhuA crystal layer effectively blocks interferents that would otherwise result in a false positive signal. The 2D^CEERhuA crystal layer opens in selective response to low-ppm levels of HCN_(g), allowing analyte penetration into the pSi sensor layer for detection. These findings illustrate that designed protein assemblies can function as dynamic components of solid-state devices in non-aqueous environments. 

The “Histidine-brace” (His-brace) copper-binding site, composed of Cu(His)₂with a backbone amine, is found in metalloproteins with diverse functions. A primary example is lytic polysaccharide monooxygenase (LPMO), a class of enzymes that catalyze the oxidative depolymerization of polysaccharides, providing not only an energy source for native microorganisms but also a route to more effective industrial biomass conversion. Despite its importance, how the Cu His-brace site performs this unique and challenging oxidative depolymerization reaction remains to be understood. To answer this question, we have designed a biosynthetic model of LPMO by incorporating the Cu His-brace motif into azurin, an electron transfer protein. Spectroscopic studies, including ultraviolet-visible (UV–Vis) absorption and electron paramagnetic resonance, confirm copper binding at the designed His-brace site. Moreover, the designed protein is catalytically active towards both cellulose and starch, the native substrates of LPMO, generating degraded oligosaccharides with multiturnovers by C1 oxidation. It also performs oxidative cleavage of the model substrate 4-nitrophenyl-D-glucopyranoside, achieving a turnover number ~9% of that of a native LPMO assayed under identical conditions. This work presents a rationally designed artificial metalloenzyme that acts as a structural and functional mimic of LPMO, which provides a promising system for understanding the role of the Cu His-brace site in LPMO activity and potential application in polysaccharide degradation.