Effect of graphene electrode functionalization on machine learning-aided single nucleotide classification.

Solid-state nanogap-based DNA sequencing with a quantum tunneling approach has emerged as a promising avenue due to its potential to deliver swift and precise sequencing outcomes. Nevertheless, despite significant progress, experimentally achieving single base resolution with a high signal-to-noise ratio remains a daunting challenge. In this work, we have utilized a machine learning (ML) framework coupled with the quantum transport method to assess and compare the nucleotide identification performance of graphene nanogaps functionalized with four different edge-saturating entities (C, H, N, and OH). The optimized ML model, especially the random forest classifier (RFC), demonstrates high accuracy (>90%) in classifying unlabeled nucleotides from their transmission readouts with the four functionalized graphene nanogaps. Additionally, the minor variance in the accuracy of nucleotide classification across the nanogaps highlights that RFC can capture the role of electrode-nucleotide coupling dynamics in their transmission function. Moreover, we have also conducted conductance sensitivity (%) and current-voltage (I-V) analyses of each functionalized nanogap. Among the edge-saturating entities, the nitrogen atom terminated graphene nanogap (NGN) is found to be the most sensitive for distinguishing DNA nucleotides. Our quantum transport combined ML study provides a useful perspective by conducting a comparative analysis of the role of edge-saturating entities in single-molecule DNA sequencing.

Integration of Artificial Intelligence and Quantum Transport toward Stereoselective Identification of Carbohydrate Isomers.

Detection of stereoisomers of carbohydrates with molecular resolution, a challenging goal analysts desire to achieve, is key to the full development of glycosciences. Despite the promise that analytical techniques made, including widely used nuclear magnetic resonance and mass spectrometry, high throughput de novo carbohydrate sequencing remains an unsolved issue. Notably, while next-generation sequencing technologies are readily available for DNA and proteins, they are conspicuously absent for carbohydrates due to the immense stereochemical and structural complexity inherent in these molecules. In this work, we report a novel computational technique that employs quantum tunneling coupled with artificial intelligence to detect complex carbohydrate anomers and stereoisomers with excellent sensitivity. The quantum tunneling footprints of carbohydrate isomers show high distinguishability with an in-depth analysis of underlying chemistry. Our findings open up a new route for carbohydrate sensing, which can be seamlessly integrated with next-generation sequencing technology for real-time analysis.

Milling-Induced "Turn-off" Luminescence in Copper Nanoclusters.

Atomically precise copper nanoclusters (NCs) attract research interest due to their intense photoluminescence, which enables their applications in photonics, optoelectronics, and sensing. Exploring these properties requires carefully designed clusters with atomic precision and a detailed understanding of their atom-specific luminescence properties. Here, we report two copper NCs, [Cu4(MNA)2(DPPE)2] and [Cu6(MNA-H)6], shortly Cu4 and Cu6, protected by 2-mercaptonicotinic acid (MNA-H2) and 1,2-bis(diphenylphosphino)ethane (DPPE), showing "turn-off" mechanoresponsive luminescence. Single-crystal X-ray diffraction reveals that in the Cu4 cluster, two Cu2 units are appended with two thiols, forming a flattened boat-shaped Cu4S2 kernel, while in the Cu6 cluster, two Cu3 units form an adamantane-like Cu6S6 kernel. High-resolution electrospray ionization mass spectrometry studies reveal the molecular nature of these clusters. Lifetime decay profiles of the two clusters show the average lifetimes of 0.84 and 1.64 µs, respectively. These thermally stable Cu NCs become nonluminescent upon mechanical milling but regain their emission upon exposure to solvent vapors. Spectroscopic data of the clusters match well with their computed electronic structures. This work expands the collection of thermally stable and mechanoresponsive luminescent coinage metal NCs, enriching the diversity and applications of such materials.

Integration of Density Functional Theory and Machine Learning for Electrolyte Optimization in High-Voltage Dual-Ion Battery Design.

Designing dual-ion batteries (DIBs) by using various electrolytes through experiments or computationally is highly time-consuming and needs high-cost sophisticated resources. Here, we have utilized the ultrafast screening capability of machine learning (ML) to search for suitable salt-electrolytes toward the design of DIBs, choosing voltage as the desirable descriptor. Considering 50 different salts and their suitable staging mechanisms, the XGBoost Regressor ML model has been found to perform with remarkable accuracy. This is further validated by density functional theory, cross-validation, and experimental findings. An interpretable ML technique has been employed for local and global feature analysis to interpret the ML predicted results, underscoring the importance of choosing input features. This ML assisted DIB design approach has the potential to explore unknown salt-electrolytes that have yet to be tested in DIBs. Finally, we introduce the predicted voltages for all of the salt-electrolyte combinations as well as their probable staging mechanism. We signify the absence of a general trend in the predicted voltages, as various combinations of cations and anions are found to deliver unique voltages. Our study can guide researchers toward tuning constituent salts as well as staging mechanisms for the design of efficient DIBs.

Multicolor photoluminescence of Cu14 clusters modulated using surface ligands.

Copper nanoclusters exhibit unique structural features and their molecular assembly results in diverse photoluminescence properties. In this study, we present ligand-dependent multicolor luminescence observed in a Cu14 cluster, primarily protected by ortho-carborane-9,12-dithiol (o-CBDT), featuring an octahedral Cu6 inner kernel enveloped by eight isolated copper atoms. The outer layer of the metal kernel consists of six bidentate o-CBDT ligands, in which carborane backbones are connected through µ3-sulphide linkages. The initially prepared Cu14 cluster, solely protected by six o-CBDT ligands, did not crystallize in its native form. However, in the presence of N,N-dimethylformamide (DMF), the cluster crystallized along with six DMF molecules. Single-crystal X-ray diffraction (SCXRD) revealed that the DMF molecules were directly coordinated to six of the eight capping Cu atoms, while oxygen atoms were bound to the two remaining Cu apices in antipodal positions. Efficient tailoring of the cluster surface with DMF shifted its luminescence from yellow to bright red. Luminescence decay profiles showed fluorescence emission for these clusters, originating from the singlet states. Additionally, we synthesized microcrystalline fibers with a one-dimensional assembly of DMF-appended Cu14 clusters and bidentate DPPE linkers. These fibers exhibited bright greenish-yellow phosphorescence emission, originating from the triplet state, indicating the drastic surface tailoring effect of secondary ligands. Theoretical calculations provided insights into the electronic energy levels and associated electronic transitions for these clusters. This work demonstrated dynamic tuning of the emissive excited states of copper nanoclusters through the efficient engineering of ligands.

Machine learning empowered next generation DNA sequencing: perspective and prospectus.

The pursuit of ultra-rapid, cost-effective, and accurate DNA sequencing is a highly sought after aspect of personalized medicine development. With recent advancements, mainstream machine learning (ML) algorithms hold immense promise for high throughput DNA sequencing at the single nucleotide level. While ML has revolutionized multiple domains of nanoscience and nanotechnology, its implementation in DNA sequencing is still in its preliminary stages. ML-aided DNA sequencing is especially appealing, as ML has the potential to decipher complex patterns and extract knowledge from complex datasets. Herein, we present a holistic framework of ML-aided next-generation DNA sequencing with domain knowledge to set directions toward the development of artificially intelligent DNA sequencers. This perspective focuses on the current state-of-the-art ML-aided DNA sequencing, exploring the opportunities as well as the future challenges in this field. In addition, we provide our personal viewpoints on the critical issues that require attention in the context of ML-aided DNA sequencing.

Lithiophilic Dibenzamide Linkages to Impart Lithium Storage Capacity in Porous Polybenzamides.

Polymer-based organic cathode materials have shown immense promise for lithium storage, owing to their structural diversity and functional group tunability. However, designing appropriate high-performance cathode materials with a high-rate capability and long cycle life remains a significant challenge. It is quintessential to design polymer-based electrodes with lithiophilic linkages. Herein, we design a bifurcated dibenzamide (DBA) linkage having lithiophilic functionalities. 1H NMR has been used as an experimental tool to understand the lithiophilic nature of the DBAs. Considering the strong Li+ affinity of DBAs, a series of polybenzamides have been designed as lithium storage systems. The design of porous polybenzamides consists of amides as only redox-active functionalities, and the rest are inactive phenyl units. Porous polybenzamides, when tested as cathodes against a Li-metal anode, displayed high capacity and rate performance, demonstrating their redox activity. The most efficient polybenzamide (TAm-TA) delivered a specific capacity of 248 mA h g-1 at 1C. TAm-TA retained 63% of its specific capacity at a very high rate of 10C (157 mA h g-1). Notably, polybenzamides displayed a capacity enhancement during long cycling, tending to achieve their theoretical capacity. Long cycling stability tests over 3000 cycles at a rate of 1.3C and over 6000 cycles at elevated rates (5C to 40C) demonstrate the electrochemical robustness of dibenzamide linkages. Finally, two full-cell experiments using TAm-TA as both cathode and anode were conducted, which delivered high capacity, demonstrating that TAm-TA is a promising candidate for Li+-ion batteries (LIBs). Furthermore, the ex situ Fourier transform infrared (FT-IR), X-ray photoemission spectroscopy (XPS), and density functional theory (DFT) studies revealed the stepwise lithiation/delithiation mechanism for polybenzamides.

Integrated Supervised and Unsupervised Machine Learning Approach to Map the Electrochemical Windows Over 4500 Solvents for Battery Applications.

The compatibility between solvent electrolytes and high-voltage electrode materials represents a significant impediment to the progress of rechargeable metal-ion batteries. Rapidly identifying suitable solvent electrolytes with optimized electrochemical windows (ECWs) within an extensive search space poses a formidable challenge. In this study, we introduce a combined supervised and unsupervised (clustering) machine learning (ML) approach to discern distinct clusters of solvent electrolytes exhibiting varying ECW ranges. Through supervised machine learning, we have accurately predicted optimal solvent electrolytes with desired ECWs from a vast pool of 4882 solvents. Our ML model boasts superior accuracy compared to previously reported data from density functional theory (DFT). Besides, the exploration of the vast solvent space through K-means clustering (unsupervised approach) yields 11 optimal clusters, each encompassing different solvents characterized by diverse ECW ranges and frequencies. The expedited reduction of solvent space achieved through clustering occurs within a very short time frame and with minimal resource expenditure. Consequently, this method is highly capable of streamlining the subsequent experimental investigations for battery applications, avoiding the need for a trial-and-error approach.

Luminescent Hydride-Free [Cu7(SC5H9)7(PPh3)3] Nanocluster: Facilitating Highly Selective C-C Bond Formation.

Amidst burgeoning interest, atomically precise copper nanoclusters (Cu NCs) have emerged as a remarkable class of nanomaterials distinguished by their unparalleled reactivity. Nonetheless, the synthesis of hydride-free Cu NCs and their role as stable catalysts remain infrequently explored. Here, we introduce a facile synthetic approach to fabricate a hydride-free [Cu7(SC5H9)7(PPh3)3] (Cu7) NC and delineate its photophysical properties intertwined with their structural configuration. Moreover, the utilization of its photophysical properties in a photoinduced C-C coupling reaction demonstrates remarkable specificity toward cross-coupling products with high yields. The combined experimental and theoretical investigation reveals a nonradical mechanistic pathway distinct from its counterparts, offering promising prospects for designing hydride-free Cu NC catalysts in the future and unveiling the selectivity of the hydride-free [Cu7(SC5H9)7(PPh3)3] NC in photoinduced Sonogashira C-C coupling through a polar reaction pathway.

MoS2/Mayenite Electride Hybrid as a Cathode Host for Suppressing Polysulfide Shuttling and Promoting Kinetics in Lithium-Sulfur Batteries.

The commercial viability of emerging lithium-sulfur batteries (LSBs) remains greatly hindered by short lifespans caused by electrically insulating sulfur, lithium polysulfides (Li2Sn; 1 ≤ n ≤ 8) shuttling, and sluggish sulfur reduction reactions (SRRs). This work proposes the utilization of a hybrid composed of sulfiphilic MoS2 and mayenite electride (C12A7:e-) as a cathode host to address these challenges. Specifically, abundant cement-based C12A7:e- is the most stable inorganic electride, possessing the ultimate electrical conductivity and low work function. Through density functional theory simulations, the key aspects of the MoS2/C12A7:e- hybrid including electronic properties, interfacial binding with Li2Sn, Li+ diffusion, and SRR have been unraveled. Our findings reveal the rational rules for MoS2 as an efficient cathode host by enhancing its mutual electrical conductivity and surface polarity via MoS2/C12A7:e-. The improved electrical conductivity of MoS2 is attributed to the electron donation from C12A7:e- to MoS2, yielding a semiconductor-to-metal transition. The resultant band positions of MoS2/C12A7:e- are well matched with those of conventional current-collecting materials (i.e., Cu and Ni), electrochemically enhancing the electronic transport. The accepted charge also intensifies MoS2 surface polarity for attracting polar Li2Sn by forming stronger bonds with Li2Sn via ionic Li-S bonds than electrolytes with Li2Sn, thereby preventing polysulfide shuttling. Importantly, MoS2/C12A7:e- not only promotes rapid reaction kinetics by reducing ionic diffusion barriers but also lowers the Gibbs free energies of the SRR for effective S8-to-Li2S conversion. Beyond the reported applications of C12A7:e-, this work highlights its functionality as an electrode material to boost the efficiency of LSBs.

"Visualizing" the partially reversible conversion of gold nanoclusters via the Au23(S-c-C6H11)17 intermediate.

Transformation chemistry of atomically precise metal nanoclusters has emerged as a novel strategy for fundamental research on the structure-property correlations of nanomaterials. However, a thorough understanding of the transformation mechanism is indeed necessary to understand the structural growth patterns and corresponding property evolutions in nanoclusters. Herein, we present the ligand-exchange-induced transformation of the [Au23(SR)16]- (8e-) nanocluster to the [Au25(SR')18]- (8e-) nanocluster, through the Au23(SR)17 (6e-) intermediate species. Identification of this key intermediate through a partially reversible transformation helped in a detailed investigation into the transformation mechanism with atomic precision. Moreover, photophysical studies carried out on this Au23(SR)17 species, which only differs by a single ligand from that of the [Au23(SR)16]- nanocluster reveal the property evolutions at the slightest change in the nanocluster structure.

Comparison of lattice thermal conductivity usingab-initioDFT, machine learning interatomic potentials, and temperature dependent effective potential: a case study of hexagonal BN and BP bilayer.

Discovering high thermal conductivity materials is essential for various practical applications, particularly in electronic cooling. The significance of two-dimensional (2D) materials lies in their unique properties that emerge due to their reduced dimensionality, making them highly promising for a wide range of applications. Hexagonal boron nitride (BN), both monolayer and bilayer forms, has garnered attention for its fascinating properties. In this work, we focus on bilayer boron phosphide (BP), which is isostructural to its BN analogue. The lattice thermal conductivity of both bilayer BN and BP have been calculated usingab-initiodensity functional theory, machine learning with the moment tensor potential method, and the temperature-dependent effective-potential method (TDEP). The TDEP approach gives more accurate results for both BN and BP materials. The lattice thermal conductivity of bilayer BP is lower than that of bilayer BN at room temperature, attributed to increased phonon anharmonicity. This study highlights the importance of understanding phonon scattering mechanisms in determining the thermal conductivity of 2D materials, contributing to the broader understanding and potential applications of these materials in future technologies.

Machine Learning-Assisted Direct RNA Sequencing with Epigenetic RNA Modification Detection via Quantum Tunneling.

RNA sequence information holds immense potential as a drug target for diagnosing various RNA-mediated diseases and viral/bacterial infections. Massively parallel complementary DNA (c-DNA) sequencing helps but results in a loss of valuable information about RNA modifications, which are often associated with cancer evolution. Herein, we report machine learning (ML)-assisted high throughput RNA sequencing with inherent RNA modification detection using a single-molecule, long-read, and label-free quantum tunneling approach. The ML tools achieve classification accuracy as high as 100% in decoding RNA and 98% for decoding both RNA and RNA modifications simultaneously. The relationships between input features and target values have been well examined through Shapley additive explanations. Furthermore, transmission and sensitivity readouts enable the recognition of RNA and its modifications with good selectivity and sensitivity. Our approach represents a starting point for ML-assisted direct RNA sequencing that can potentially decode RNA and its epigenetic modifications at the molecular level.

Advancement of Next-Generation DNA Sequencing through Ionic Blockade and Transverse Tunneling Current Methods.

DNA sequencing is transforming the field of medical diagnostics and personalized medicine development by providing a pool of genetic information. Recent advancements have propelled solid-state material-based sequencing into the forefront as a promising next-generation sequencing (NGS) technology, offering amplification-free, cost-effective, and high-throughput DNA analysis. Consequently, a comprehensive framework for diverse sequencing methodologies and a cross-sectional understanding with meticulous documentation of the latest advancements is of timely need. This review explores a broad spectrum of progress and accomplishments in the field of DNA sequencing, focusing mainly on electrical detection methods. The review delves deep into both the theoretical and experimental demonstrations of the ionic blockade and transverse tunneling current methods across a broad range of device architectures, nanopore, nanogap, nanochannel, and hybrid/heterostructures. Additionally, various aspects of each architecture are explored along with their strengths and weaknesses, scrutinizing their potential applications for ultrafast DNA sequencing. Finally, an overview of existing challenges and future directions is provided to expedite the emergence of high-precision and ultrafast DNA sequencing with ionic and transverse current approaches.

Precision Basecalling of Single DNA Nucleotide from Overlapped Transmission Readouts with Machine Learning Aided Solid-State Nanogap.

DNA sequencing with the quantum tunneling technique heralds a paradigm shift in genetic analysis, promising rapid and accurate identification for diverging applications ranging from personalized medicine to security issues. However, the widespread distribution of molecular conductance, conduction orbital alignment for resonant transport, and decoding crisscrossing conductance signals of isomorphic nucleotides have been persistent experimental hurdles for swift and precise identification. Herein, we have reported a machine learning (ML)-driven quantum tunneling study with solid-state model nanogap to determine nucleotides at single-base resolution. The optimized ML basecaller has demonstrated a high predictive basecalling accuracy of all four nucleotides from seven distinct data pools, each containing complex transmission readouts of their different dynamic conformations. ML classification of quaternary, ternary, and binary nucleotide combinations is also performed with high precision, sensitivity, and F1 score. ML explainability unravels the evidence of how extracted normalized features within overlapped nucleotide signals contribute to classification improvement. Moreover, electronic fingerprints, conductance sensitivity, and current readout analysis of nucleotides have promised practical applicability with significant sensitivity and distinguishability. Through this ML approach, our study pushes the boundaries of quantum sequencing by highlighting the effectiveness of single nucleotide basecalling with promising implications for advancing genomics and molecular diagnostics.

Localized charge-induced ORR/OER activity in doped fullerenes for Li-air battery applications.

Non-aqueous Li-air batteries have garnered significant interest in recent years. The key challenge lies in the development of efficient catalysts to overcome the sluggish kinetics associated with the oxygen reduction reaction (ORR) during discharge and the oxygen evolution reaction (OER) during charging at the cathode. In this work, we conducted a comprehensive study on B/N-doped and BN co-doped fullerenes using first-principles analysis. Our results show significant changes in the geometries, electronic properties, and catalytic behaviors of doped and co-doped fullerenes. The coexistence of boron and nitrogen boosts the formation energy, enhancing stability compared to pristine and single-doped structures. C179B exhibits minimal overpotentials (0.98 V), implying superior catalyst performance for ORR and OER in LABs and significantly better performance than Pt (111) (3.48 V) and standard graphene (3.51 V). The electron-deficient nature of the B atom makes it provide its vacant 2pz orbital for conjugation with the p-electrons of nearby carbon atoms. Consequently, boron serves as a highly active site due to the localization of positive charge, which improves the adsorption of intermediates through oxygen atoms. Moreover, the higher activity of B-doped systems than N-doped systems in lithium-rich environments is opposite to the observed trend in the reported PEM fuel cells. This work introduces doped and co-doped fullerenes as LAB catalysts, offering insights into their tunable ORR/OER activity via doping with various heteroatoms and fullerene size modulation.

Insights of the efficient hydrogen evolution reaction performance in bimetallic Au4Cu2 nanoclusters.

The design of efficient electrocatalysts for improving hydrogen evolution reaction (HER) performance using atomically precise metal nanoclusters (NCs) is an emerging area of research. Here, we have studied the HER electrocatalytic performance of monometallic Cu6 and Au6 nanoclusters and bimetallic Au4Cu2 nanoclusters. A bimetallic Au4Cu2/MoS2 composite exhibits excellent HER catalytic activity with an overpotential (η10) of 155 mV vs. reversible hydrogen electrode observed at 10 mA cm-2 current density. The improved HER performance in Au4Cu2 is due to the increased electrochemically active surface area (ECSA), and Au4Cu2 NCs exhibits better stability than Cu6 and Au6 systems and bare MoS2. This augmentation offers a greater number of active sites for the favorable adsorption of reaction intermediates. Furthermore, by employing X-ray photoelectron spectroscopy (XPS) and Raman analysis, the kinetics of HER in the Au4Cu2/MoS2 composite were elucidated, attributing the favorable performance to better electronic interactions occurring at the interface between Au4Cu2 NCs and the MoS2 substrate. Theoretical analysis reveals that the inherent catalytic enhancement in Au4Cu2/MoS2 is due to favorable H atom adsorption over it and the smallest ΔGH* value. The downshift in the d-band of the Au4Cu2/MoS2 composite influences the binding energy of intermediate catalytic species. This new catalyst sheds light on the structure-property relationship for improving electrocatalytic performance at the atomic level.

Functionality Modulation Toward Thianthrene-based Metal-Free Electrocatalysts for Water Splitting.

The development of metal-free bifunctional electrocatalysts for hydrogen and oxygen evolution reactions (HER and OER) is significant but rarely demonstrated. Porous organic polymers (POPs) with well-defined electroactive functionalities show superior performance in hydrogen evolution reaction (HER) and oxygen evolution reaction (OER). Precise control of the active sites' local environment requires careful modulation of linkers through the judicious selection of building units. Here, a systematic strategy is introduced for modulating functionality to design and synthesize a series of thianthrene-based bifunctional sp2 C═C bonded POPs with hollow spherical morphologies exhibiting superior electrocatalytic activity. This precise structural tuning allowed to gain insight into the effects of heteroatom incorporation, hydrophilicity, and variations in linker length on electrocatalytic activity. The most efficient bifunctional electrocatalyst THT-PyDAN achieves a current density of 10 mA cm─2 at an overpotential (η10) of ≈65 mV (in 0.5 m H2SO4) and ≈283 mV (in 1 m KOH) for HER and OER, respectively. THT-PyDAN exhibits superior activity to all previously reported metal-free bifunctional electrocatalysts in the literature. Furthermore, these investigations demonstrate that THT-PyDAN maintains its performance even after 36 h of chronoamperometry and 1000 CV cycling. Post-catalytic characterization using FT-IR, XPS, and microscopic imaging techniques underscores the long-term durability of THT-PyDAN.

A thiolated copper-hydride nanocluster with chloride bridging as a catalyst for carbonylative C-N coupling of aryl amines under mild conditions: a combined experimental and theoretical study.

Atomically precise copper nanoclusters (Cu NCs), an emerging class of nanomaterials, have garnered significant attention owing to their versatile core-shell architecture and their potential applications in catalytic reactions. In this study, we present a straightforward synthesis strategy for [Cu29(StBu)12(PPh3)4Cl6H10][BF4] (Cu29) NCs and explore their catalytic activity in the carbonylative C-N coupling reaction involving aromatic amines and N-heteroarenes with dialkyl azodicarboxylates. Through a combination of experimental investigations and density functional theory studies, we elucidate the radical mechanisms at play. The crucial step in the catalytic process is identified as the decomposition of diisopropyl azodicarboxylates on the surface of Cu29 NCs, leading to the generation of oxyacyl radicals and the liberation of nitrogen gas. Subsequently, an oxyacyl radical abstracts a hydrogen atom from aniline, initiating the formation of an aminyl radical. Finally, the aminyl radical reacts with another oxyacyl radical, culminating in the synthesis of the desired carbamate product. This detailed analysis provides insights into the intricate catalytic pathways of Cu29 NCs, shedding light on their potential for catalyzing carbonylative C-N coupling reactions.

Unravelling CO2 Reduction Reaction Intermediates on High Entropy Alloy Catalysts: An Interpretable Machine Learning Approach to Establish Scaling Relations.

Establishment of a scaling relation among the reaction intermediates is highly important but very much challenging on complex surfaces, such as surfaces of high entropy alloys (HEAs). Herein, we designed an interpretable machine learning (ML) approach to establish a scaling relation among CO2 reduction reaction (CO2 RR) intermediates adsorbed at the same adsorption site. Local Interpretable Model-Agnostic Explanations (LIME), Accumulated Local Effects (ALE), and Permutation Feature Importance (PFI) are used for the global and local interpretation of the utilized black box models. These methods were successfully applied through an iterative way and validated on CuCoNiZnMg and CuCoNiZnSnbased HEAs data. Finally, we successfully predicted adsorption energies of *H2 CO (MAE: 0.24 eV) and *H3 CO (MAE: 0.23 eV) by using the *HCO training data. Similarly, adsorption energy of *O (MAE: 0.32 eV) is also predicted from *H training data. We believe that our proposed method can shift the paradigm of state-of-the-art ML in catalysis towards better interpretability.