When Chemical Intuition Meets Neural Networks: Hybrid AI Algorithm Predicts Metal–Ligand Coordination Structures

Hemoglobin, which gives blood its red color, the light-emitting materials in smartphone OLED displays, and catalysts used in manufacturing plastics and pharmaceuticals—all share a common feature that is they all involve metal–ligand coordination complexes. In these structures, a central metal ion is surrounded by organic molecules, known as ligands. Now, a novel artificial intelligence (AI) algorithm offers the promise of designing these complex materials more swiftly and accurately, significantly reducing trial-and-error efforts. Predicting how a metal ion will 'shake hands' with an organic molecule is an inherently complex task, often challenging even for seasoned scientists. This difficulty arises because organic ligands contain multiple potential binding sites, and the number and mode of coordination can vary widely depending on the specific metal and its oxidation state. As a result, dozens or even hundreds of potential structural configurations can exist. Led by Distinguished Professor Bartosz A. Grzybowski of the Department of Chemistry and Director of the Center for Algorithmic and Robotized Synthesis (CARS) within the Institute for Basic Science (IBS) at UNIST, the research team has addressed this challenge not through increased computational power, but by integrating 'chemical intuition' directly into AI. Published in Angewandte Chemie, their hybrid AI approach moves beyond simple pattern recognition by encoding fundamental chemical principles into neural networks. Figure 1. An AI-powered hybrid model predicts metal–ligand coordination modes by integrating chemical rules from crystal structures with machine learning, enabling accurate design of complex organometallic structures. "We have developed a 'hybrid' strategy that combines classical chemical knowledge with machine learning to accurately predict how metals and organic ligands bind," explains Professor Grzybowski. "By analyzing over 100,000 crystal structures from the Cambridge Structural Database, we trained a model that 'understands' the specific behavior of different metals and their oxidation states—something previous models could not achieve." This innovative technology is poised to become a key component in large-scale computational pipelines. Professor Grzybowski notes, “We expect it to be highly valuable in designing transition metal catalysts used across industries—from pharmaceuticals to advanced functional materials.” To make this cutting-edge tool accessible worldwide, the team has released RDMetallics, an open-source, Python-based software package, along with a dedicated web portal (coordinate.rdmetallics.net). Researchers anywhere in the world can now predict complex metal–ligand coordination modes directly through their web browsers—no coding skills or expensive hardware required. Supported by the Institute for Basic Science (IBS), this research was published online on Angewandte Chemie International Edition on February 21, 2026. Bartosz A. Grzybowski Distinguished Professor, Department of Chemistry, UNIST Director, IBS Center for Robotized and Algorithmic Synthesis (CARS) E: grzybor72@unist.ac.kr William I. Suh IBS Public Relations Team E: willisuh@ibs.re.kr Story Source Materials provided by the Institute of Basic Science. Journal Reference Galymzhan Moldagulov, Kisung Lee, Sanzhar Nurgaliyev, et al., "Hybrid Computational Strategy for Predicting Complex Ligand–Metal Architectures," Angew. Chem. Int. Ed. (2026). https://doi.org/10.1002/anie.202524655

2026.04.03

Targeting COL6A3-C5 with Nigericin to Reduce Endotrophin and Enhance Metabolic Health in Obesity

Abstract Endotrophin, a cleavage product of collagen VIα3 (COL6A3), contributes to fibroinflammation in adipose tissue and exacerbates systemic insulin resistance in obesity. Previously, we demonstrated that various hypoxia-induced matrix metalloproteinases (MMPs) are directly involved in the cleavage of COL6A3 to generate endotrophin in obese adipose tissue; thus, inhibition of endotrophin generation by blocking MMP access could be beneficial for treating obesity-related metabolic disease. Here we identified nigericin as an inhibitor of endotrophin generation, which improves fibroinflammation and insulin sensitivity in both hypoxic adipocytes in vitro and diet-induced obese mice in vivo. Mechanistically, nigericin directly binds to the COL6A3-C5 domain, competing with MMPs and thereby disrupting the interactions between the COL6A3-C5 domain and MMPs. This interference prevents the cleavage of endotrophin from the COL6A3 by MMPs, ultimately inhibiting its generation. Taken together, these results strongly suggest that pharmacological blockade of endotrophin cleavage, by using nigericin, effectively decreases endotrophin levels and improves endotrophin-mediated fibroinflammation and insulin resistance in obesity. Furthermore, this new therapeutic strategy could be applied to various metabolic diseases and solid tumors where endotrophin levels are pathologically elevated. Obesity is known to be a major risk factor that exacerbates metabolic disorders, such as type 2 diabetes. A key molecule involved in this process is endotrophin, a signaling protein that links excess fat accumulation to metabolic decline. Recent research from UNIST has identified a natural compound, capable of directly inhibiting endotrophin production, paving the way for innovative therapeutic interventions. Led by Professor Jiyoung Park of the Department of Biological Sciences, the research team demonstrated that nigericin, a natural substance derived from microorganisms, effectively suppresses endotrophin synthesis in obese adipose tissue. Their findings reveal that nigericin reduces fibrosis and inflammation within fat tissue while significantly improving insulin sensitivity. Obese adipose tissue often undergoes hypoxia, which triggers chronic inflammation and pathological extracellular matrix remodeling. This process leads to increased production of endotrophin, a fragment of collagen VIα3 (COL6A3), known to impair adipose function and worsen metabolic health. The team uncovered that nigericin binds specifically to a region of COL6A3, preventing matrix metalloproteinases (MMPs)—enzymes responsible for collagen cleavage—from accessing their substrate. By occupying this site, nigericin blocks endotrophin formation at the molecular level. Figure 1. Schematic representation of the proposed mechanism by which NGC regulates ETP generation and improves insulin sensitivity in obese AT. "This strategy is particularly noteworthy because it directly targets the initial step in endotrophin production, unlike traditional therapies that indirectly suppress inflammation or alter gene expression," explained the research team. "It offers a new paradigm for disrupting the pathological signaling pathways underlying metabolic diseases." In preclinical models, administration of nigericin to high-fat diet-fed mice resulted in marked reductions in adipose tissue fibrosis and inflammation. These effects occurred without adverse effects on liver or kidney function. Notably, fasting blood glucose levels decreased by approximately 30%, and insulin sensitivity was significantly improved. The researchers screened over 1,000 natural compounds and identified nigericin—produced by actinomycetes—as a promising candidate, owing to its ability to stably inhibit endotrophin synthesis without directly affecting proteolytic enzymes. Professor Park concluded, "This discovery reveals a novel molecular mechanism for controlling endotrophin levels and highlights the potential for developing targeted therapies for obesity, diabetes, and other conditions characterized by fibroinflammation, including certain cancers." Supported by the Ministry of Science and ICT (MSIT) and the National Research Foundation of Korea (NRF), the study was published online in Experimental & Molecular Medicine, on March 5, 2026. Journal Reference Chu-Sook Kim, Woobeen Jo, Jungsun Yoo, et al., "Targeting COL6A3-C5 with nigericin suppresses endotrophin formation and enhances insulin sensitivity in obesity," Exp. Mol. Med., (2026).

2026.04.01

Advanced Interface Engineering for High-Performance Solar Cells and Green Hydrogen

Abstract Self-assembled monolayer (SAM)-based hole-selective layers (HSLs) offer a promising route to defect-passivated and energy-aligned interfaces in perovskite organic tandem solar cells (POTSCs). However, their practical implementation remains hindered by weak anchoring to transparent conductive oxides (TCOs), leading to desorption during perovskite deposition and poor interfacial durability under polar solvent exposure. Here, we present a chemical interfacial stabilization strategy in which potassium carbonate (K2CO3) mediates the controlled deprotonation of [2-(9H-carbazol-9-yl)ethyl]phosphonic acid (2PACz), forming mixed mono- and di-deprotonated species (2PACz-K) that bind strongly to indium tin oxide (ITO). The resulting SAM exhibits superior solvent resistance, improved energy-level alignment, and enhanced buried interface quality. POTSCs incorporating 2PACz-K achieve 25.10% power conversion efficiency (PCE) with a high open-circuit voltage (VOC) of 2.230 V, while retaining 80% of their initial PCE after 220 h of maximum power point (MPP) tracking under simulated 1-sun illumination. Beyond photovoltaics, the robust 2PACz-K interface is further integrated into a perovskite/organic tandem photocathode (POT-PEC), representing the first transparent, metal-free tandem PEC architecture capable of stable operation in aqueous electrolyte, delivering a photovoltage (Vph) of 2.16 V and achieving a solar-to-hydrogen (STH) conversion efficiency of 7.7%. This work establishes a versatile interfacial design paradigm that bridges photovoltaic and photoelectrochemical energy conversion. Researchers at UNIST have unveiled a novel interface engineering technique that significantly improves both the performance and durability of perovskite/organic tandem solar cells (POTSCs). Published in the February 2026 issue of Energy & Environmental Science, their study demonstrates how precise control of self-assembled monolayers (SAMs) at the molecular level can lead to more stable, high-efficiency solar devices and open new pathways for solar-driven hydrogen production. POTSCs are among the most promising photovoltaic technologies, combining different light-absorbing materials to maximize energy conversion. However, instability at the interface between the transparent electrode and the perovskite layer has long hampered their long-term reliability, especially under operational conditions. Led by Professors Jin-Young Kim and Dong-Seok Kim from the Graduate School of Carbon Neutrality, along with Professor Seung-Jae Shin from the School of Energy and Chemical Engineering, the team developed a method to chemically stabilize this critical interface. They focused on a self-assembled monolayer (SAM) known as 2PACz, which facilitates hole extraction in solar cells. Figure 1. Schematic illustration of the device structure of POTSCs with 2PACz-K and their performance. By introducing potassium carbonate (K₂CO₃), the team induced a controlled deprotonation of 2PACz molecules—partially removing hydrogen ions from their phosphonic acid groups. This process creates a mixture of mono- and di-deprotonated species (referred to as 2PACz-K), which acquire a negative charge and form stronger, more stable bonds with indium tin oxide (ITO) electrodes. This enhanced bonding results in an interface that is more resistant to solvents used during device fabrication, maintaining uniformity and stability. Devices fabricated with this chemically tailored interface demonstrated remarkable performance: perovskite solar cells achieved a power conversion efficiency (PCE) of 25.1% and an open-circuit voltage (VOC) of 2.23 V. Moreover, these cells retained over 80% of their initial efficiency after 220 hours of continuous operation under simulated sunlight, indicating substantial improvements in operational stability. The team also applied this interface engineering approach to photoelectrochemical (PEC) cells designed for water splitting. The resulting tandem PEC device exhibited a high photovoltage of 2.16 V and achieved a solar-to-hydrogen (STH) conversion efficiency of 7.7% without the need for external bias—marking a significant step toward practical, solar-driven hydrogen generation. Professor Jin-Young Kim stated, "Controlling the chemical state of interfaces at the molecular level allows us to dramatically improve both the efficiency and long-term stability of solar energy devices. This strategy offers a promising avenue for developing integrated systems that convert sunlight directly into electricity and hydrogen, supporting a sustainable energy future." The study has been supported by the National Research Foundation of Korea (NRF) and the Ministry of Science and ICT (MSIT). Journal Reference Jung Geon Son, Ha-eun Koo, Woojin Lee, et al., "Deprotonated self-assembled molecules as robust hole-selective layers for perovskite/organic tandem solar cells and photocathodes," Energy Environ. Sci., (2026).

2026.03.27

New Study Unveils Adaptive AI Breakthrough for Reflection Separation in the Wild

Abstract Single Image Reflection Separation (SIRS) aims to reconstruct both the transmitted and reflected images from a single image that contains a superimposition of both, captured through a glass-like reflective surface. Recent learning-based methods of SIRS have significantly improved performance on typical images with mild reflection artifacts; however, they often struggle with diverse images containing challenging reflections captured in the wild. In this paper, we propose a universal SIRS framework based on a flexible dual-stream architecture, capable of handling diverse reflection artifacts. Specifically, we incorporate a Mixture-of-Experts mechanism that dynamically assigns specialized experts to image patches based on spatially heterogeneous reflection characteristics. The assigned experts then cooperate to extract complementary features between the transmission and reflection streams in an adaptive manner. In addition, we leverage the multi-head attention mechanism of Transformers to simultaneously exploit both high and low cross-correlations, which are then complementarily used to facilitate adaptive inter-stream feature interactions. Experimental results evaluated on diverse real-world datasets demonstrate that the proposed method significantly outperforms existing state-of-the-art methods qualitatively and quantitatively. Capturing a picturesque scene through reflective materials, such as glass, often results in an unintended superimposition—showing both the transmitted scene and undesired reflected scene. While traditional reflection removal techniques have made progress, they frequently struggle with complex reflection patterns and varying lighting conditions, leaving residual artifacts that diminish image quality. Recent advances in artificial intelligence are now providing powerful new solutions. Led by Professor Jae-Young Sim from the Graduate School of Artificial Intelligence at UNIST, researchers have developed an innovative AI model that effectively separates reflections from the transmitted scene, enabling clearer and more authentic views beyond the reflective surface. Recognizing the limitations of existing methods, which often falter in complex and spatially heterogeneous reflection scenarios, the team designed an approach that intelligently segments the superimposed image for targeted analysis. This segmentation allows for precise removal of reflections while maintaining the integrity of the transmitted scene. The core of this breakthrough lies in two techniques, which are Complementary Mixture-of-Experts (CoME) and Complementary Cross-Attention (CoCA). CoME employs a mixture-of-experts (MoE) architecture that dynamically assigns specialized neural networks—referred to as 'experts'—to different regions within an image based on local reflection characteristics. These experts collaboratively analyze both the transmitted and reflected layers, exchanging relevant information to improve separation accuracy, especially in regions with diverse reflection patterns. CoCA enhances the reconstruction process by considering both strongly correlated and weakly correlated regions. Unlike traditional attention mechanisms that focus solely on highly related areas, CoCA recognizes that meaningful reflection details can also exist in less correlated regions, enabling a more comprehensive and effective separation. Extensive evaluations on diverse real-world datasets demonstrate that this approach surpasses existing state-of-the-art methods in both visual quality and quantitative performance. It remains robust even in challenging scenarios with intricate reflection distortions—areas where prior models often underperform. Professor Sim remarked, "Reflections in natural scenes are inherently complex and vary widely. Traditional neural networks often struggle to handle this variability. Our approach, with its adaptive expert allocation and dual attention mechanisms, offers a more flexible and effective solution. We believe this technology has significant potential across a range of imaging applications, from photography to autonomous systems." Supported by the Institute for Information & Communications Technology Planning & Evaluation (IITP) and the National Research Foundation of Korea (NRF), this research has been published in the IEEE Transactions on Image Processing. Journal Reference Jonghyuk Park and Jae-Young Sim, "Complementary Mixture-of-Experts and Complementary Cross-Attention for Single Image Reflection Separation in the Wild," IEEE TIP, (2026).

2026.03.26

New Study Unveils Atomic Disorder Strategy to Enhance Stability and Efficiency of High-Capacity Batteries

Abstract Anionic redox in lithium-rich layered oxides (LRLOs) offers a breakthrough to higher energy density but is limited by voltage hysteresis arising from irreversible structural disorder. While enhancing transition metal–oxygen (TM-O) covalency through π-type interaction improves the reversibility of anionic processes, inevitable structural disorder during the first cycle still deteriorates TM-O hybridization. Here, we propose a counterintuitive strategy that embraces pre-synthetic cation disorder to preserve TM-O π-redox. The in-plane disordered arrangement modulates the first-cycle phase evolution, suppressing O3–O1 slab gliding and relaxing localized cationic oxidation at high voltage. This structural control maintains robust TM-O coordination and stabilized oxygen states even under high-voltage operation, yielding markedly reduced voltage hysteresis (0.31 vs 0.62 V) and exceptional long-term stability with minimal voltage decay (−0.04 mV cycle–1) and 98.0% energy retention after 160 cycles. This work establishes structural-disorder-driven phase evolution control as a practical design principle for stabilizing π-redox chemistry, achieving high-energy, structurally resilient LRLOs. Researchers at UNIST, in collaboration with the Pohang Accelerator Laboratory (PAL) and KAIST, have introduced a novel approach to stabilizing high-capacity battery materials. By intentionally inducing atomic-level disorder within lithium-rich layered oxide (LRLO) cathodes, the team has effectively minimized structural degradation and energy losses, paving the way for next-generation batteries with higher energy density and longer lifespan. Lithium-rich layered oxide (LRLO) are among the most promising cathode materials for future energy storage solutions due to their exceptional capacity, which involves not only metal ions but also oxygen participating in electrochemical reactions. However, their practical application has been hindered by structural instability during repeated charge and discharge cycles, leading to capacity fade and voltage degradation. To address this challenge, Professor Hyun-Wook Lee from the School of Energy and Chemical Engineering at UNIST, along with Dr. Young Hwa Jung from PAL and Professor Dong-Hwa Seo from KAIST, employed an innovative strategy, deliberately designing the atomic structure to be disordered. This controlled atomic disorder prevents the initial phase transition—known as slab gliding—that typically causes irreversible structural damage. Consequently, the material preserves its integrity and electrochemical performance over extended cycling. Figure 1. (Top) Schematic illustration comparing structural evolution pathways and associated TM-O covalency in ordered and pre-disordered Li/TM arrangements. (Bottom) First-principles calculations showing relative O3–O1 stacking preference during Mn migration (left) and schematic illustration of the suppressed O3→O1 slab gliding in the pre-disordered structure (right). Using advanced computational modeling based on density functional theory (DFT) and synchrotron radiation analysis at PAL, the team confirmed that the disordered structure stabilizes the bonds between transition metals and oxygen. Experimental evaluations demonstrated that these disordered cathodes exhibit a voltage difference of only 0.31V between charge and discharge during the first cycle—less than half the 0.62V observed in conventional, ordered materials—and an initial energy loss of merely 0.6%. In comparison, traditional cathodes experienced twice the voltage gap and a 25.8% energy loss. Remarkably, the disordered cathodes maintained their energy capacity with minimal voltage decay—only 0.04 mV per cycle—and retained 98% of their initial energy after 160 cycles. This exceptional stability highlights the potential of atomic-level disorder as a universal design principle for high-performance cathode materials. "By transforming what was once considered a defect—the atomic disarray—into a strategic advantage, we have opened a new pathway for enhancing battery stability," said Myeongjun Choi from UNIST, the study's first author. "This approach is versatile and can be applied to a wide range of lithium-rich layered oxides, beyond specific compositions or structures." Professor Lee added, "While lithium-rich layered oxides possess significant energy potential, their commercialization has been limited by structural issues. Our findings offer a promising route toward more durable, lightweight batteries capable of storing more energy efficiently, which could significantly impact future energy storage technologies." The findings of this research have been published online in ACS Energy Letters on February 3, 2026. This study has been supported by funding from the National Research Foundation of Korea (NRF), through projects focused on nanomaterials and advanced energy materials. Journal Reference Myeongjun Choi, Jeongwoo Seo, Min-Ho Kim, et al., "Pre-Disordering for Preserving Transition Metal–Oxygen Covalency in Lithium-Rich Layered Oxide Cathodes," ACS Energy Lett., (2026).

2026.03.23

Innovative Recycling Method to Convert Waste PET into High-Quality Raw Materials and Clean Hydrogen

Abstract Polyethylene terephthalate (PET) is a primary target for chemical plastic recycling due to its widespread use and relatively weak ester bonds in its structure. However, conventional PET depolymerization methods—such as alkaline hydrolysis, glycolysis, and methanolysis—are energy-intensive and require complex separation steps, which increase both costs and environmental impact. This study introduces a polyoxometalate-based recycling process to address these limitations. Under mild conditions (100 °C and low pressure in aqueous solution), polyoxometalates catalyze the depolymerization of PET via acid hydrolysis, producing high-purity terephthalic acid (TPA) and ethylene glycol (EG) as solid and liquid products, respectively. EG is further oxidized by polyoxometalates to yield valuable compounds such as glycolic acid and formic acid, while simultaneously storing electrons. Under optimized conditions, EG oxidation achieves high selectivity (∼85%) toward formic acid. The stored electrons can be utilized for low-energy hydrogen production (125 mA cm−2 at 1.2 V) or electricity generation (12.5 mW cm−2 at 0.05 V). Crucially, our techno-economic analysis reveals that this approach, which combines revenue from high-purity TPA and valorized co-products, is cost-competitive and has the potential to supply TPA at a price lower than that of the virgin material. This work presents a technically robust and economically viable pathway toward a circular economy for plastic waste. Despite polyethylene terephthalate (PET) being one of the most widely recycled plastics, only about 20% of used PET bottles are actually recovered as high-quality raw materials. The majority are transformed into lower-grade fibers or fillers before eventually being discarded. Addressing this gap, researchers at UNIST have developed a novel chemical recycling process that not only restores PET to its original high-grade form but also produces valuable chemicals and clean hydrogen, all under mild conditions. Led by Professors Jungki Ryu and Tae Hoon Oh from the School of Energy and Chemical Engineering, the team introduced a multifunctional catalyst that enables PET depolymerization at just 100°C—significantly milder than conventional methods that operate above 200°C. The process involves mixing shredded PET with water, a solvent (DMSO), and the catalyst, then heating the mixture. This causes the plastic to break down into terephthalic acid (TPA), a high-quality raw material for PET production, and ethylene glycol (EG), a valuable chemical. A key advantage of this approach is its simplicity: after filtration, high-purity TPA can be recovered for reuse. Additionally, the process produces formic acid (FA), a high-value chemical, while the catalyst can be recycled multiple times. Interestingly, the catalyst also functions as an electron reservoir, storing electrons during PET breakdown. These stored electrons can be harnessed to produce hydrogen efficiently at low voltage or generate electricity through fuel cells, offering an integrated solution for energy recovery. Experimental results demonstrated that hydrogen could be generated at a voltage 25% lower than typical water electrolysis—at just 1.2 volts—and a fuel cell system produced 12.5 milliwatts per square centimeter. Economic analysis further revealed that the minimum selling price of the recycled TPA is estimated at just $0.81 per kilogram, making it up to 46% cheaper than traditional chemical recycling methods and more affordable than virgin TPA derived from crude oil. Professor Ryu stated, "Our process shows that high-quality recycled PET and clean hydrogen can be produced cost-effectively under mild conditions. This not only supports a circular economy for plastics but also offers a sustainable pathway for hydrogen energy production." Supported by the Ministry of Science and ICT (MSIT) and the InnoCORE program of hydro*studio at UNIST, this research has been featured as the back cover of the January 2026 issue of Green Chemistry. Additional support was provided by the Basic Science Research Program, the Regional Leading Research Center (RLRC), and the Engineering Research Center of Excellence Program, all funded by the National Research Foundation (NRF) of Korea. The project also received backing from the Regional Innovation System & Education (RISE) through the Ulsan RISE Center, funded by the Ministry of Education (MOE) and Ulsan Metropolitan City. Journal Reference Hyeonmyeong Oh, Ye Chan Lee, Inhui Lee, et al., "Mild hydrolysis of PET and electrochemical energy recovery via multifunctional polyoxometalate catalysts," Green Chem., (2026).

2026.03.20