Quantum Mechanical Calculations Research Articles

A Quantum Mechanical Approach to The Mechanism of Asymmetric Synthesis of Chiral Amine by Imine Reductase from Stackebrandtia Nassauensis.

The asymmetric synthesis of tetrahydroisoquinolines (THIQs) has gained importance in recent years due to their significant potential in drug development studies. In this study, the conversion of 1-methyl-3,4-dihydroisoquinoline substrate to a chiral amine, 1-methyl-1,2,3,4-tetrahydroisoquinoline, under the catalysis of the stereoselective imine reductase enzyme from Stackebrandtia nassauensis (SnIR) was investigated in detail to elucidate the mechanism and explain the experimental enantioselectivity. The results were found to be in agreement with the experimental data. To elucidate the reaction mechanism, quantum mechanical calculations were performed by considering a large cluster of the active site of the enzyme. In this regard, possible reaction pathways leading to both R- and S-products with the corresponding intermediates and the transition states for the hydride transfer from the cofactor to the substrate were considered by density functional theory (DFT) calculations, and the factors contributing to the observed stereoselectivity were sought. The calculations supported a stepwise mechanism rather than the concerted protonation and the hydride transfer steps. The stereoselectivity in the hydride transfer was found to be due not only to the stability of the enzyme-subtrate complex but also to the corresponding reaction barriers. The calculations were performed at the wB97XD/6-311+G(2df,2p)//B3LYP/6-31G(d,p) level of theory using the PCM approach.

Engineering the Substrate Specificity of UDP-Glycosyltransferases for Synthesizing Triterpenoid Glycosides with a Linear Trisaccharide as Aided by Ancestral Sequence Reconstruction.

Triterpenoids have wide applications in the pharmaceutical and agricultural industries. The glycosylation of triterpenoids catalyzed by UDP-glycosyltransferases (UGTs) is a crucial method for producing valuable derivatives with enhanced functions. However, only a few UDP-glucosyltransferases have been reported to synthesize the rare triterpenoids with linear-chain trisaccharide at C3-OH. This study revealed that the UGT91H subfamily primarily contributed to the 2"-O-glycosylation of triterpenoids with high regioselectivity, then the substrate scope was further expanded by ancestral sequence reconstruction (ASR). With ancestral enzyme UGT91H_A1 as a model, the sequence-structure-function relationship was explored. A RTAS loop (R212/T213/A214/S215) was identified to affect the substrate specificity of UGT91H_A1. Transferring this RTAS loop to the corresponding position of UGT91H enzymes successfully expanded their substrate spectra. The functional role of RTAS loop was further elucidated by molecular dynamics simulation and quantum mechanical computation. UGT91H_A1 was applied to the low-cost synthesis of terpenoid rhamnosides with a linear trisaccharide in combining with a self-sufficient UDP-rhamnose regeneration system. Finally, we developed a phylogeny-based platform to efficiently mining new UGT91Hs from plant genomic data. This study provided robust biocatalysts for synthesizing various triterpenoid glycosides with a linear trisaccharide and demonstrated ASR as an efficient tool in engineering the function of UDP-glycosyltransferases.

Quantum mechanical analysis of newly synthesized HIV-1 protease inhibitors: evaluation of wild-type and resistant strain binding interactions.

Inhibition of HIV-1 protease is a cornerstone of antiretroviral therapy. However, the notorious ability of HIV-1 to develop resistance to protease inhibitors (PIs), particularly darunavir (DRV), poses a major challenge. Using quantum chemistry and computer simulations, this study aims to investigate the interactions between two novel PIs, GRL-004 and GRL-063, as well as a wild-type (WT) HIV strain and a DRV-resistant mutant strain. To do this, we used molecular docking, molecular dynamics simulations, and quantum mechanical calculations to check how well GRL-004 and GRL-063 bound to both WT and DRV-resistant proteases. The results show that GRL-004 and GRL-063 bind very well to ASP29 in the WT strain. ASP29 is an important amino acid in the HIV protease dimer. Remarkably, amino acids such as ILE50 in the WT strains showed substantial binding energies to both drugs. Quantum energy calculations showed a slight reduction in the energy affinity of the interaction between the MUT strain and the ligand GRL-063, compared to the WT strain. GRL-004 showed similar interaction energy for both strains, suggesting that it has greater plasticity than GRL-063 despite its lower interaction affinity. Furthermore, GLY49B demonstrated strong binding energies regardless of mutations. Other relevant residues with strong binding energies include GLY49B, PHE82A, PRO81A, ASP29A, ASP25A and ALA28B. This study improves our understanding of receptor-ligand dynamics and the adaptability of new protease inhibitors (PIs), which has profound implications for the innovation of future antiretroviral drugs.

Parameterization of a Fluctuating Charge Model for Complexes Containing 3d Transition Metals.

Metalloproteins widely exist in biology, playing pivotal roles in diverse life processes. Meanwhile, molecular dynamics (MD) simulations based on classical force fields has emerged as an important tool in scientific research. Partial charges are critical parameters within classical force fields and usually derived from quantum mechanical (QM) calculations. However, QM calculations are often time-consuming and prone to basis set dependence. Alternatively, fluctuating charge (FQ) models offer another avenue for partial charge derivation, which has significant speed advantages and can be used for large-scale screening. Building upon our previous work, which introduced an FQ model for zinc-containing complexes, herein we extend this model to include additional 3d transition metals which are important to the life sciences, namely chromium, manganese, iron, cobalt, and nickel. Employing CM5 charges as target for parametrization, our FQ model accurately reproduces partial charges for 3d metal complexes featuring biologically relevant ligands. Furthermore, by using atomic charges derived by our FQ model, MD simulations have been performed. These charges showed excellent performance in simulating proteomic metal sites housing multiple metal ions, specifically, a metalloprotein containing an iron-sulfur cluster and another containing a dimanganese metal site, showcasing comparable performance to those of RESP charges. We anticipate that our study can accelerate the parametrization of atomic charges for metalloproteins featuring 3d transition metals, thereby facilitating simulations of relevant systems.

Pyrene Based Schiff Base for Cascade Fluorescent "On-Off" Detection of Aluminium(III) and Fluoride ions and its Applications in Live Cells Imaging.

An easy-to-prepare pyrene-based Schiff base PNZ was synthesized by condensing equimolar amount of 1-pyrenebutyric hydrazide with 2-hydroxy-naphthaldehyde, and employed as a fluorescent chemosensor for in-situ cascade detection of aluminium (Al3+) and fluoride (F-)ions. In DMSO:H2O (1:1, v/v), the weakly emissive PNZ showed a significant fluorescence enhancement at 455 nm selectively upon the addition of Al3+ due to the complexation-induced formation of a pyrene excimer. Schiff base PNZ and Al3+ formed a complex in 2:1 binding ratio with the estimated binding constants of K1:1 = 9826.01 M-1 and K2:1 = 3188.49 M-1. The sensing mechanism was explored by performing quantum mechanical calculations and 1H NMR titration of PNZ with Al3+. The in-situ formed PNZ-Al3+ complex species enabled the fluorescent turn-off detection of F-. Using PNZ and PNZ-Al3+, the concentrations of Al3+ and F- ions can be detected down to 2.89×10-7 M and 1.88×10-7 M, respectively. The cytotoxicity of the PNZ and its ability to bioimage Al3+ and F- ions was examined in the human cervical cancer cell line. Finally, the receptor PNZ was applied for the quantification of Al3+ and F- ions in various real samples, such as tap water, river water, rainwater, mouthwash, and toothpaste.

Physics-Informed Machine Learning with Data-Driven Equations for Predicting Organic Solar Cell Performance.

Organic solar cells (OSCs) have emerged as a promising solution in pursuing sustainable energy. This study presents a comprehensive approach to advancing OSC development by integrating data-driven equations from quantum mechanical (QM) descriptors with physics-informed machine learning (PIML) models. We circumvent traditional experimental limitations through high-throughput QM calculations, prioritizing transparent and interpretable models. Using the SISSO++ method, we identified key descriptors that effectively map the relationships between input variables and photovoltaic performance metrics. Our innovative predictive models, derived from SISSO outputs, excel in forecasting critical OSC parameters such as short-circuit current (JSC), open-circuit voltage (VOC), fill factor (FF), and power conversion efficiency (PCEmax), achieving high accuracy even with limited data sets. To validate our models' practical utility, we applied the PIML framework to a newly compiled data set of OSC devices, demonstrating their versatility and capability in pinpointing high-performance materials. This research underscores the strong predictive power of our models, bridging the gap between experimental results and theoretical predictions and making significant contributions to the advancement of sustainable energy technologies.

Molecular architecture of a novel indoline-fused chromenylium-cyanine probe carrying methionine biomolecule for ultrasensitive analyzing Hg2+ ion in real samples

In this study, by decorating the chromenylium-cyanine structure with an indoline ring, a new sensor platform CSI with a highly rigid structure and improved photophysical properties was developed for the first time. The CSI platform was modified with glyoxal and L-methionine methyl ester to produce a new NIR, ratiometric, and switch-on probe (CIME). This probe is biocompatible, water soluble, and has a specific binding mode for Hg2+ ion selectivity. The crystal structure of CSI was determined by x-ray crystallography, and its electronic characteristics, including HOMO-LUMO energy levels were determined by quantum mechanical calculations. CIME is the first instance of an indoline-fused chromenylium-cyanine used for quantification of trace level of Hg2+ in both the living cells and the environment. The new probe has a high level of selectivity and sensitivity for Hg2+ in aqueous media, with broad linear ranges (0–300 μM for UV-Vis and 0–225 μM for fluorescence). It has a rapid response time of 30 seconds, and UV-Vis and fluorescence detection limits of 7.4 ng/mL and 0.13 ng/mL, respectively. Probe quantitatively detected Hg2+ in real samples using UV-Vis, fluorescence spectrophotometry and a smartphone. The interaction between CIME and Hg2+ was verified by FTIR, HRMS and DFT calculations. Furthermore, the ability of CIME to sense Hg2+ in living cells was confirmed in both MCF-7 cells and 3T3 cells.

A novel application of hydrophilic interaction liquid chromatography for the identification of compounds with intramolecular hydrogen bonds

The incorporation of intramolecular hydrogen bonds (IMHB) into small molecules constitutes an interesting optimization strategy to afford potential drug candidates with enhanced solubility as well as permeability and consequently improved bioavailability (if metabolic stability is high). Common methods to assess IMHB rely on spectroscopic or diffraction techniques, which, however, have limited throughput when screening for hit compounds in early phases of drug discovery. Inspired by literature findings using supercritical fluid chromatography (SFC) as an indirect method for IMHB identification in a screening context, we aimed at developing a secondary chromatographic methodology taking advantage of commonly used HPLC-MS instrumentation. In this work, we explored hydrophilic interaction liquid chromatography (HILIC) and developed a method for discriminating compounds based on their hydrogen bonding features. By quantifying retention of different matched molecular pairs (MMP) and using information about their low energy conformations from quantum-mechanical calculations, we defined a hydrogen bonding-driven adsorption (kads) chromatographic parameter to assess a compound’s propensity to forming IMHB. In addition to the MMP analysis, we found that the kads parameter allows for the differentiation of analytes forming IMHB regardless of the comparison with control compounds.

From `crystallographic accuracy' to `thermodynamic accuracy': a redetermination of the crystal structure of calcium atorvastatin trihydrate (Lipitor®).

With ever-improving quantum-mechanical computational methods, the accuracy requirements for experimental crystal structures increase. The crystal structure of calcium atorvastatin trihydrate, which has 56 degrees of freedom when determined with a real-space algorithm, was determined from powder diffraction data by Hodge et al. [Powder Diffr. (2020), 35, 136-143]. The crystal structure was a good fit to the experimental data, indicating that the electron density had been captured essentially correctly, but two independent quantum-mechanical calculations disagreed with the experimental structure and with each other. Using the same experimental data, the crystal structure was redetermined from scratch and it was shown that it can be reproduced within a root-mean-square Cartesian displacement of 0.1 Å by two independent quantum-mechanical calculations. The consequences for the calculated energies and solubilities are described.

Reaction-driven restructuring of defective PtSe2 into ultrastable catalyst for the oxygen reduction reaction.

PtM (M = S, Se, Te) dichalcogenides are promising two-dimensional materials for electronics, optoelectronics and gas sensors due to their high air stability, tunable bandgap and high carrier mobility. However, their potential as electrocatalysts for the oxygen reduction reaction (ORR) is often underestimated due to their semiconducting properties and limited surface area from van der Waals stacking. Here we show an approach for synthesizing a highly efficient and stable ORR catalyst by restructuring defective platinum diselenide (DEF-PtSe2) through electrochemical cycling in an O2-saturated electrolyte. After 42,000 cycles, DEF-PtSe2 exhibited 1.3 times higher specific activity and 2.6 times higher mass activity compared with a commercial Pt/C electrocatalyst. Even after 126,000 cycles, it maintained superior ORR performance with minimal decay. Quantum mechanical calculations using hybrid density functional theory reveal that the improved performance is due to the synergistic contributions from Pt nanoparticles and the apical active sites on the DEF-PtSe2 surface. This work highlights the potential of DEF-PtSe2 as a durable electrocatalyst for ORR, offering insights into PtM dichalcogenide electrochemistry and the design of advanced catalysts.

Distinct Adjacent Substrate Binding Pocket Regulates the Activity of a Decameric Feruloyl Esterase from Bacteroides thetaiotaomicron.

Understanding how the human gut microbiota contribute to the metabolism of dietary carbohydrates is of great interest, particularly those with ferulic acid (FA) decorations that have manifold health benefits. Here, we report the crystal structure of a decameric feruloyl esterase (BtFae) from Bacteroides thetaiotaomicron in complex with methyl ferulate (MFA), revealing that MFA is situated in a noncatalytic substrate binding pocket adjacent to the catalytic pocket. Molecular docking and mutagenesis studies further demonstrated that the adjacent pocket affects substrate binding in the active site and negatively regulates the BtFae activity on both synthetic and natural xylan substrates. Additionally, quantum mechanics (QM) calculations were employed to investigate the catalytic process of BtFae from substrate binding to product release, and identified TS_2 in the acylation step is rate-limiting. Collectively, this study unmasks a novel regulatory mechanism of FAE activity, which may contribute to further investigation of FA-conjugated polysaccharides metabolism in the human gut.

8018 Improving Anti-Breast Cancer Activities by Optimizing Torsion Angle Energies

Abstract Disclosure: D.R. Zak: None. E.C. Fink: None. G.R. Hancock: None. K.S. Young: None. S.W. Fanning: None. Overexpression of estrogen receptor alpha (ER-alpha) is found in 70% of breast cancer tumors. Normally, ER-alpha remains inactive and sequestered in the cytoplasm until estrogen binds to the ligand-binding domain (LBD) to enable nuclear translocation, coregulator recruitment, and subsequent activation of proliferative ER-alpha-dependent transcriptomic profiles. The majority of breast cancers are treated successfully with ER-targeted endocrine therapies, but hotspot ER-alpha LBD mutants are selected for, namely Y537S and D538G causing therapeutic resistance, relapse, and aggressive metastatic disease. The selective estrogen receptor modulator (SERM) lasofoxifene (laso) and selective estrogen receptor degrader (SERD) elacestrant (RAD1901) are chemically distinctive and show great promise in the ER-alpha mutant setting with potent anti-tumor effects and improved side effect profiles compared to the standard-of-care SERD fulvestrant (ICI). To elucidate the structural-transcriptional relationships of these ligands, we created a laso/RAD1901 chemical hybrid T6I-29 that demonstrated both canonical and unexpected therapeutic mechanisms of action. However, it possesses reduced anti-proliferative potency compared to ICI. We hypothesized that minor chemical changes to T6I-29 could improve potency while maintaining its favorable pharmaceutical profiles. To address this hypothesis, we performed quantum mechanical calculations of a series of T6I-derivatives. The addition of a methyl linker within the scaffold core significantly improved the torsion angle energies of the ligand binding pose. Multiple derivatives were synthesized based on these computations. The new compound T6I-B2 exerted profound anti-proliferative effects on breast cancer cells and adopted a unique ligand binding pose within the LBD. This approach shows that T6I-SERM potency can be improved by optimizing ligand torsion angle energies. Presentation: 6/3/2024

High-performance metal-organic frameworks for efficient adsorption, controlled release, and membrane separation of organophosphate pesticides

Metal-organic frameworks (MOFs) are considered good choices for the absorption, separation, and release of various substances. The rapid increase in the number of MOFs being produced makes it impractical to conduct experimental research to identify the most promising candidates in these areas. Therefore, computational screening has emerged as a robust approach. We employed this approach to identify the most efficient MOFs for the removal, separation, and controlled release of organophosphorus pesticides, including diazinon, chlorpyrifos, and chlorpyrifos-methyl. From a database of approximately 14,000 MOFs, 584 MOFs with the largest cavity diameters were used in the screening process to evaluate promising candidates for each pesticide. Monte Carlo simulations were used to estimate the pesticide uptake capacities and isosteric heats of the selected MOFs. The structural properties of MOFs were correlated with their adsorption capacities and isosteric heats. Computational screening identified 325, 330, and 450 MOFs as potential absorbents and carriers for diazinon, chlorpyrifos-methyl, and chlorpyrifos, respectively. Additionally, it proposed 235, 254, and 135 MOFs as membranes for the effective separation of these organophosphorus pesticides. The reactivity of the pesticides was explored using quantum mechanics calculations. Molecular dynamics simulations of the top three MOFs with high uptake capacities for each pesticide were performed and analyzed to understand the interactions between the MOF, pesticide, and water molecules. The solvation-free energies of the pesticides were calculated for various solvents to identify the most effective solvent for desorbing pesticides from the MOFs, thereby enabling MOF regeneration.

High Affinity Inhibitors of the Macrophage Infectivity Potentiator Protein from Trypanosoma cruzi, Burkholderia pseudomallei, and Legionella pneumophila─A Comparison.

Since Chagas disease, melioidosis, and Legionnaires' disease are all potentially life-threatening infections, there is an urgent need for new treatment strategies. All causative agents, Trypanosoma cruzi, Burkholderia pseudomallei, and Legionella pneumophila, express a virulence factor, the macrophage infectivity potentiator (MIP) protein, emerging as a promising new therapeutic target. Inhibition of MIP proteins having a peptidyl-prolyl isomerase activity leads to reduced viability, proliferation, and cell invasion. The affinity of a series of pipecolic acid-type MIP inhibitors was evaluated against all MIPs using a fluorescence polarization assay. The analysis of structure-activity relationships led to highly active inhibitors of MIPs of all pathogens, characterized by a one-digit nanomolar affinity for the MIPs and a very effective inhibition of their peptidyl-prolyl isomerase activity. Docking studies, molecular dynamics simulations, and quantum mechanical calculations suggest an extended σ-hole of the meta-halogenated phenyl sulfonamide to be responsible for the high affinity.

Red-emitting carbon dots for fluorescent and smartphone-assisted dual-mode detection of Cu(II), Hg(II), and Fe(III) and investigation of the sensing mechanism

Red-emitting carbon dots (R-CDs) have triggered ever-increasing interest in biochemical sensing due to their deep tissue penetration and minimal biological auto-fluorescence interference. Herein, dual-emitting intrinsic R-CDs are fabricated and well-characterized. R-CDs display ratiometric and colorimetric sensing properties for heavy metal ions (HMIs, Cu2+, Hg2+, and Fe3+), with the corresponding detection limits of 24.5, 74.2, and 170 nM, respectively. Additionally, smartphone imaging colorimetry based on red-green-blue (RGB) value variations is employed to achieve real-time and on-site detection of these HMIs. Crucially, discrimination of these HMIs is facilitated by principal component analysis. Furthermore, R-CDs demonstrate effective optical sensing of these HMIs in tap/lake water and human urine with satisfactory recoveries (95.82–104.8 %). The underlying mechanism of optical sensing, which involves metal ion-triggered aggregation-induced quenching, is elucidated by spectroscopic, morphological analysis, and quantum mechanical calculations. This work presents a feasible real-time approach for the fluorescent and smartphone-assisted dual-mode detection of HMIs.

Orbital Hall Effect Accompanying Quantum Hall Effect: Landau Levels Cause Orbital Polarized Edge Currents.

The quantum Hall effect emerges when two-dimensional samples are subjected to strong magnetic fields at low temperatures: Topologically protected edge states cause a quantized Hall conductivity in multiples of e^{2}/h. Here we show that the quantum Hall effect is accompanied by an orbital Hall effect. Our quantum mechanical calculations fit well the semiclassical interpretation in terms of "skipping orbits." The chiral edge states of a quantum Hall system are orbital polarized akin to a hypothetical orbital version of the quantum anomalous Hall effect in magnetic systems. The orbital Hall resistivity scales quadratically with the magnetic field, making it the dominant effect at high fields.

ANI/EFP: Modeling Long-Range Interactions in ANI Neural Network with Effective Fragment Potentials.

Deep learning Neural Networks (NN) have been developed in the field of molecular modeling for the purpose of circumventing the high computational cost of quantum-mechanical calculations while rivaling their accuracies. Although these networks have found great success, they generally lack the ability to accurately describe long-range interactions, which makes them unusable for extended molecular systems. Herein, we provide a method for partially retraining the deep learning general-use neural network ANI, in which the long-range interactions are represented via atomic electrostatic potentials. The electrostatic potentials, generated with polarizable effective fragment potentials (EFP), are used as an additional input feature for the network. This new ANI/EFP network can predict solute-solvent interaction energies on a trained data set with a kcal/mol accuracy. It also shows promise in predicting the interaction energies of a solute in solvent environments that have not been included in a training data set. The proposed protocol can be taken as an example and further developed, leading to highly accurate and transferable neural network potentials capable of handling long-range interactions and extended molecular systems.

Artificial Intelligence Modeling of Materials’ Bulk Chemical and Physical Properties

Energies of the atomic and molecular orbitals belonging to one and two atom systems from the fourth and fifth periods of the periodic table have been calculated using ab initio quantum mechanical calculations. The energies of selected occupied and unoccupied orbitals surrounding the highest occupied and lowest unoccupied orbitals (HOMOs and LUMOs) of each system were selected and used as input for our artificial intelligence (AI) software. Using the AI software, correlations between orbital parameters and selected chemical and physical properties of bulk materials composed of these elements were established. Using these correlations, the materials’ bulk properties were predicted. The Q2 correlation for the single-atom predictions of first ionization potential, melting point, and boiling point were 0.3589, 0.4599, and 0.1798 respectively. The corresponding Q2 correlations using orbital parameters describing two-atom systems increased the capability to predict the experimental properties to the respective values of 0.8551, 0.8207, and 0.7877. The accuracy in predicting materials’ bulk properties was increased up to four-fold by using two atoms instead of one. We also present results of the prediction of molecules for materials relevant to energy systems.

Quantum refinement in real and reciprocal space using the Phenix and ORCA software.

X-ray and neutron crystallography, as well as cryogenic electron microscopy (cryo-EM), are the most common methods to obtain atomic structures of biological macromolecules. A feature they all have in common is that, at typical resolutions, the experimental data need to be supplemented by empirical restraints, ensuring that the final structure is chemically reasonable. The restraints are accurate for amino acids and nucleic acids, but often less accurate for substrates, inhibitors, small-molecule ligands and metal sites, for which experimental data are scarce or empirical potentials are harder to formulate. This can be solved using quantum mechanical calculations for a small but interesting part of the structure. Such an approach, called quantum refinement, has been shown to improve structures locally, allow the determination of the protonation and oxidation states of ligands and metals, and discriminate between different interpretations of the structure. Here, we present a new implementation of quantum refinement interfacing the widely used structure-refinement software Phenix and the freely available quantum mechanical software ORCA. Through application to manganese superoxide dismutase and V- and Fe-nitrogenase, we show that the approach works effectively for X-ray and neutron crystal structures, that old results can be reproduced and structural discrimination can be performed. We discuss how the weight factor between the experimental data and the empirical restraints should be selected and how quantum mechanical quality measures such as strain energies should be calculated. We also present an application of quantum refinement to cryo-EM data for particulate methane monooxygenase and show that this may be the method of choice for metal sites in such structures because no accurate empirical restraints are currently available for metals.

Design principles for selective nitrogen adsorption via chelation of metals in metal organic frameworks

Designing porous materials with N2 selectivity performance is highly attractive for the efficient removing N2 from natural gas. However, selectively capturing N2 over CH4 is challenging due to their similar physical properties. In this work, we proposed achieving N2 selectivity adsorption over CH4 by chelating Fe2+ sites on the ligands of MOFs. The significant enhancement of N2/CH4 separation by Fe chelation was demonstrated using multiscale simulations, including quantum mechanics calculations, force field development, and GCMC simulations. Specifically, Fe chelated MOFs can achieve a record-high N2 adsorption capacity while maintaining N2/CH4 selectivity comparable to that of current benchmark N2 selective capture MOFs. Moreover, based on simulation data from 331 Fe chelated MOFs with various topologies, we quantified the structure-properties relationships for N2/CH4 separation using Fe chelated MOFs through interpretable machine learning, which provides precise and intuitive guidance for designing novel adsorbents for selective N2 adsorption.