Calculation Of Atomic Charges Research Articles

Coordinate-Free and Low-Order Scaling Machine Learning Model for Atomic Partial Charge Prediction for Any Size of Molecules.

The atomic partial charge is of great importance in many fields, such as chemistry and drug-target recognition. However, conventional quantum-based computing of atomic charges is relatively slow, limiting further applications of atomic charge analysis. With the help of machine learning methods, various kinds of models appear to speed up atomic charge calculations. However, there are still some concerning problems. Some models based on geometric coordinates require high-accuracy geometry optimization as a preprocess, while other models have a limitation on the size of input molecules that narrow the applications of the model. Here, we propose a machine learning atomic charge model based on a message-passing featurizer. This preprocessing featurizer can quickly extract atomic environment information from a molecule according to the connectivity inside the molecule. The resulting descriptor can be used with a neural network to quickly predict the atomic partial charge. The model is able to automatically adapt to any size of molecule while remaining efficient and achieves a root-mean-square error in the Hirshfeld charge prediction of 0.018e, with an overall time complexity of O(n2). Thus, this model could enlarge the range of applications of atomic partial charge to more fields and cases.

Exploring the Potential Mechanisms of Guanxinshutong Capsules in Treating Pathological Cardiac Hypertrophy based on Network Pharmacology, Computer-Aided Drug Design, and Animal Experiments.

Cardiovascular diseases (CVDs) are significant causes of morbidity and mortality worldwide, and pathological cardiac hypertrophy (PCH) is an essential predictor of many heart diseases. Guanxinshutong capsule (GXST) is a Chinese patent medicine widely used in the clinical treatment of CVD, In our previous research, we identified 111 compounds of GXST. In order to reveal the potential molecular mechanisms by which GXST treats PCH, this study employed network pharmacology methods to screen for the active ingredients of GXST in treating PCH and predicted the potential targets. The results identified 26 active ingredients of GXST and 110 potential targets for PCH. Through a protein-protein interaction (PPI) network, gene ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, we confirmed AKT1, MAPK1, and MAPK3 as the core proteins in GXST treatment of PCH, thus establishing the PI3K/AKT and MAPK signaling pathways as the significant mechanisms of GXST in treating PCH. The results of molecular docking (MD) demonstrate that flavonoid naringenin and diterpenoid tanshinone iia have the highest binding affinity with the core protein. Before performing molecular dynamics simulations (MDSs), the geometric structure of naringenin and tanshinone iia was optimized using density functional theory (DFT) at the B97-3c level, and RESP2 atomic charge calculations were carried out at the B3LYP-D3(BJ)/def2-TZVP level. Further MDS results demonstrated that in the human body environment, the complex of naringenin and tanshinone iii with core proteins exhibited high stability, flexibility, and low binding free energy. Additionally, naringenin and tanshinone iia showed favorable absorption, distribution, metabolism, excretion, and toxicity (ADMET) characteristics and passed the drug similarity (DS) assessment. Ultrasound cardiograms and cardiac morphometric measurements in animal experiments demonstrate that GXST can improve the PCH induced by isoproterenol (ISO). Protein immunoblotting results indicate that GXST increases the expression of P-eNOS and eNOS by activating the PI3K/AKT signaling pathway and the MAPK signaling pathway, further elucidating the mechanism of action of GXST in treating PCH. This study contributes to the elucidation of the key ingredients and molecular mechanisms of GXST in treating PCH.

Regulating the hydroxyl groups reactivity of cellulose by grafting the quaternary ammonium group to achieve a salt-free and low alkali dyeing process for reactive dye

In this work, a salt-free and low alkali dyeing process was developed through cationic modification of cotton fabric with a series of quaternary ammonium salts (QAS). The dyeing performance indicated that the cationic cotton fabric treated with 3-chloro-2-hydroxypropyldimethyloctane ammonium chloride (CT-8) achieved better K/S value (8.87) and dye fixation (90.47 %) compared to the conventional dyeing process. Notably, the CT-8 treated fabric performed exceptionally under salt-free conditions and with a Na2CO3 concentration of 5 g/L. The rationale behind the adoption of a salt-free and low-alkali dyeing process was attributed to the positive charge of quaternary ammonium groups, which had an augmenting impact on the hydroxyl reaction activity of cotton fabrics. The condensed Fukui function, atomic charge, and HOMO orbital calculations showed that the QAS structure could regulate the hydroxyl reactivity of cationic cotton fabric. Our salt-free and low alkali dyeing process not only achieved the aim of reducing chemical consumption and emissions, but also contributed to better understand the effect of hydroxyl reactivity of cationic cotton on the fixation reaction with reactive dye, and provided a new direction to achieve the require of sustainable development and clean production for a variety of industrial crops and products.

Hydration sphere structure of architectural molecules: Polyethylene glycol and polyoxymethylene oligomers

Non-toxic, chemically inert, organic polymers as polyethylene glycol (PEG) and polyoxymethylene (POM) have versatile applications in basic research, industry and pharmacy. In this work, we aim to characterize the hydration structure of PEG and POM oligomers by exploring how the solute disturbs the water structure compared to the bulk solvent and how the solute chain interacts with the solvent. We explore the effect of (i) the C–C–O (PEG) versus C-O (POM) constitution of the chain and (ii) chain length. To this end, MD simulations followed by clustering and topological analysis of the hydration network, as well as quantum mechanical calculations of atomic charges are used.We show that the hydration varies with chain conformation and length. The degree of folding of the chain impacts its degree of solvation, which is measurable by different parameters as for example the number of water molecules in the first solvation shell and the solvent accessible surface. Atomic charges calculated on the oligomers in gas phase are stable throughout conformation and chain length and seem not to determine solvation. Hydration however induces charge transfer from the solute molecule to the solvent, which depends on the degree of hydration.

αCharges: partial atomic charges for AlphaFold structures in high quality.

The AlphaFold2 prediction algorithm opened up the possibility of exploring proteins' structural space at an unprecedented scale. Currently, >200 million protein structures predicted by this approach are deposited in AlphaFoldDB, covering entire proteomes of multiple organisms, including humans. Predicted structures are, however, stored without detailed functional annotations describing their chemical behaviour. Partial atomic charges, which map electron distribution over a molecule and provide a clue to its chemical reactivity, are an important example of such data. We introduce the web application αCharges: a tool for the quick calculation of partial atomic charges for protein structures from AlphaFoldDB. The charges are calculated by the recent empirical method SQE+qp, parameterised for this class of molecules using robust quantum mechanics charges (B3LYP/6-31G*/NPA) on PROPKA3 protonated structures. The computed partial atomic charges can be downloaded in common data formats or visualised via the powerful Mol* viewer.The αCharges application is freely available at https://alphacharges.ncbr.muni.cz with no login requirement.

K2Fe(C2O4)2: An Oxalate Cathode for Li/Na-Ion Batteries Exhibiting a Combination of Multielectron Cation and Anion Redox.

The development of multielectron redox-active cathode materials is a top priority for achieving high energy density with long cycle life in the next-generation secondary battery applications. Triggering anion redox activity is regarded as a promising strategy to enhance the energy density of polyanionic cathodes for Li/Na-ion batteries. Herein, K2Fe(C2O4)2 is shown to be a promising new cathode material that combines metal redox activity with oxalate anion (C2O4 2-) redox. This compound reveals specific discharge capacities of 116 and 60 mAh g-1 for sodium-ion batterie (NIB) and lithium-ion batterie (LIB) cathode applications, respectively, at a rate of 10 mA g-1, with excellent cycling stability. The experimental results are complemented by density functional theory (DFT) calculations of the average atomic charges.

Structures of LaH10, EuH9, and UH8 superhydrides rationalized by electron counting and Jahn-Teller distortions in a covalent cluster model.

The superconducting hydrides LaH10, EuH9 and UH8 are studied using chemically intuitive bonding analysis of periodic and molecular models. We find trends in the crystallographic and electronic structures of the materials by focusing on chemically meaningful building blocks in the predicted H sublattices. Atomic charge calculations, using two complementary techniques, allow us to assign oxidation states to the metals and divide the H sublattice into neutral and anionic components. Cubic [H8]q- clusters are an important structural motif, and molecular orbital analysis of this cluster in isolation shows the crystal structures to be consistent with our oxidation state assignments. Crystal orbital Hamilton population analysis confirms the applicability of the cluster model to the periodic electronic structure. A Jahn-Teller distortion predicted by MO analysis rationalises the distortion observed in a prior study of EuH9. The impact of this distortion on superconductivity is determined, and implications for crystal structure prediction in other metal-hydrogen systems are discussed. Additionally, the performance of electronic structure analysis methods at high pressures are tested and recommendations for future studies are given. These results demonstrate the value of simple bonding models in rationalizing chemical structures under extreme conditions.

Computational studies of caffeoylquinic acids from Brazilian green propolis and its anti-viral potential against SARS-CoV-2 Mpro

Purpose: The present study evaluated the following caffeoylquinic acids, Neochlorogenic acid (3-CQA), Cryptochlorogenic acid (4-CQA), Chlorogenic Acid (5-CQA), Isochlorogenic acid A (3,5-DCQA) Isochlorogenic acid B (3,4-DCQA) and Isochlorogenic acid C (4,5-DCQA) found in Brazilian green propolis, regarding its physicochemical, pharmacokinetic and conformational characteristics. Additionally, the compounds were explored on the efficacy of inhibiting the pharmacological target SARS-CoV-2 Mpro, a protein involved in SARS-Cov-2 infection. Methods: The physicochemical and pharmacokinetic proprieties were obtained from the projected 2D structures of the respective compounds. The conformational characteristics were obtained from the three-dimensional models were subjected to geometric optimization by molecular mechanics under MMFF94 force field, subsequently, the calculation of partial atomic charges by employing AM1 semi-empirical methodology was performed. Molecular docking was performed, and the protein was loaded in PDB Data Bank (PDB ID 6LU7). Results: The physicochemical and pharmacokinetic results indicate that these phytochemicals have from medium to low potential for gastrointestinal tract absorption, principally in relation to the LogP values, and violated some druglikeness’ criteria. However molecular docking showed that the compounds 3-CQA, 4-CQA, 5-CQA, 3,5-DCQA, 3,4-DCQA and 4,5-DCQA potentially binds with the active site of the SARS-CoV-2 Mpro, through stables complexes, with a docking score of -6.44, -6.11, -6.48, -6.26, -7.01 and -7.40 kcal/mol, respectively. Conclusion: The physicochemical and pharmacokinetic parameters, as well as Hydrogen-bond formation, energy landscape, indicate that the compound with the greatest therapeutic potential are 5-CQA, 3,4-DCQA and 4,5-DCQA however this study necessitates further in vitro and in vivo experimental validation.

Is an Inductive Effect Explicit Account Required for Atomic Charges Aimed at Use within the Force Fields?

Polarization and inductive effects are the concepts that have been widely used in qualitative and even quantitative descriptions of experimentally observed properties in chemistry. The polarization effect has proven to be important in cases of biomolecular modeling though still the vast majority of molecular simulations use the classical non-polarizable force fields. In the last few decades, a lot of effort has been put into promoting the polarization effect and incorporating it into modern force fields and charge calculation methods. In contrast, the inductive effect has not attracted such attention and is effectively absent in both classic and modern force fields. Thus, a question is whether this difference corresponds to the difference in the physical significance of the effects and their explicit account, or is an artifact that should be corrected in the next generation of force fields. The significance of the electronic effects is studied in this paper through the prism of performance of specific models for atomic charge calculation that take into explicit account a nested set of effects: the formal charge, the nearest neighbors, the inductive effect, and finally the model, which takes into account all effects, which are possible to account for using atomic charges. The specific choice for the methods is the following: formal charges, MMFF94 bond charge increments, Dynamic Electronegativity Relaxation (DENR), and RESP. We propose a special scheme for the separate estimation of each particular effect contribution. By pairwise comparing the residual molecular electrostatic potential (MEP) errors of those charge models (aimed at best reproducing the quantum chemical reference MEP), we sequentially revealed how the account of each effect contributes to the better-quality MEP reproduction. The following relative importance of effects was estimated; thus, the natural hierarchy of the effects was established. First, the account of formal charges is of primordial importance. Second, the nearest neighbors account is the next in significance. Third, the explicit account of inductive effect in empirical charge calculation schemes was shown to significantly─both qualitatively and quantitatively─improve the quality of MEP reproduction. Fourth, the contribution of polarization is indirectly assessed. Surprisingly, it is of the order of magnitude of the inductive effect even for the molecular systems, for which it is anticipated to be more significant. Finally, the relative importance of anisotropic effects in neutral molecules was additionally reviewed.

Evaluation of OH Radical Reaction Positions in 3-Ring PAHs Using Transition State Energy and Atomic Charge Calculations

In this study, transition state energy and atomic charge were calculated using the Gaussian 09 program with focus on three-ring PAHs, such as acenaphthylene and anthracene, which are most likely found in contaminated sites. The calculation results were then compared with the radical reaction positions reported in the existing literature. Because the energy difference between the reactant and the transition state according to the reaction position was very small, no distinct correlation was obtained when results were compared with those of the OH radical test findings reported in the literature. It was also found that the charge calculation makes it possible to accurately predict the radical reaction position of the target material. In addition, MK and HLY charges were found to be more accurate than CHelpG charges in predicting the radical reaction positions. The charge calculation can also be applied in predicting radical reaction positions for hazardous materials with different molecular structures.

Atomic polar tensor charge calculation of beta asarone in solvation models

The Atomic Polar Tensor charges of Beta Asarone was computed in gas state and also in solvation. The solutions used for simulations are water and ethanol. These charges are compared to the already discussed Mulliken’s charge analysis. These would throw light on the biological activity of Beta Asarone which is a major component of Acorus Calamus – a medicinal plant.

Influence of glucose, sucrose, and dextran coatings on the stability and toxicity of silver nanoparticles

Aqueous colloids, consisting of 15–30 nm-sized silver nanoparticles (Ag NPs), were prepared using the reducing and stabilizing abilities of glucose, sucrose, and dextran. The long-term stability of coated Ag NPs increases from glucose over sucrose to dextran, i.e., with the increase of the molecular weight of carbohydrate molecules. The density functional theory (DFT) calculations of the partial atomic (Mulliken) charges and adsorption energies are applied to explain the enhanced stability of coated Ag NPs. All coated Ag NPs have a significantly broader concentration range of nontoxic behavior toward pre-osteoblast cells than bare Ag NPs prepared using sodium borohydride. The carbohydrate-coated Ag NPs display the same level of toxic ability against Gram-negative (Escherichia coli) and Gram-positive (Staphylococcus aureus) bacteria as bare Ag NPs. The differences in toxicity mechanism of the coated and bare Ag NPs are a consequence of the absence and presence of co-occurring Ag+ ions in examined dispersion, respectively.

Structural characterization, DFT studies and luminescent properties of dinuclear copper(I)-diimine complexes with the S-shape configurations

Three dinuclear Cu(I) complexes, [Cu2(pnd)(PPh3)4](BF4)2(1), [Cu2(pnd)(dppp)2] (BF4)2(2), and [Cu2(pnd)(DPEphos)2](BF4)2(3) (PPh3 = triphenylphosphine, dppp = 1, 3-bis (diphenylphosphino)propane, DPEphos = bis(2-diphenylphosphinophenyl)ether, pnd = N, N’-bis (pyridin-2-ylmethylene)naphthyl-1, 5-diimine), have been prepared and characterized by IR, 1H NMR, 31P NMR, 13C NMR, XRD, elemental analysis and X-ray crystal structure analysis. The structural analysis reveals that the flexible pnd in 1–3 acts in a trans bridging coordination mode to connect two Cu+ ions, forming a S-shape conformation, while each of Cu (I) ion is four-coordinate N2P2, adopting a distorted-tetrahedral geometry. DFT calculations show that the HOMOs of complexes 1–3 are primarily located on the copper d-orbital and the phosphine ligand, together with a specific contribution of pnd, while LUMOs are mostly composed of pnd with admixed Cu(I) d-orbital character. Corresponding calculations of atomic charges indicate that the charge of the copper atom in complexes 1–3 is below + 1, while the total charges of the coordinated atoms are 0.743–0.786. Moreover, the solid-state emission spectra of complexes 1–3 show the existence of broad emission bands at 350–550 nm.

Study on the selective separation of methanol and methyl ethyl ketone from the azeotropic system using ionic liquids and their separation mechanism

In this work, two ionic liquids (ILs) 1-decyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide ([DMIM][NTf2]) and tributylmethylammonium bis(trifluoromethanesulfonyl)imide ([N4441][NTf2]) were synthesized and used to separate the methanol (MeOH) and methyl ethyl ketone (MEK) azeotropic system. The VLE data of the binary system (MeOH + MEK) and the ternary system (MeOH + MEK + IL) were measured at 101.3 kPa. The binary VLE data were tested by Wisniak’s L-W method and van Ness method, the ternary VLE data were tested by van Ness method, and it was found all the binary and ternary VLE data were thermodynamically consistent. The NRTL and UNIFAC-Lei activity coefficient models were used to fit the experimental data and it was found that both models could be applied to the system, but the NRTL model was more suitable. Through the analysis of the experimental data, it was found that both ILs could increase the relative volatility of MeOH to MEK, the effect became obvious with the increase of the IL content, and the azeotropic phenomenon could be eliminated as the content of ILs achieved a certain value. The ability of the two ILs to break azeotrope was in the order of [DMIM][NTf2] > [N4441][NTf2]. The minimum mole fraction of [DMIM][NTf2] and [N4441][NTf2] breaking the azeotrope of MeOH and MEK is 0.035 and 0.039, respectively, which is lower than that of the reported best ionic liquid. Finally, the separation mechanism was explored through excess enthalpy analysis, σ-profile analysis, and atomic charge calculation to promote the understanding of the process.

A Reduced Generalized Force Field for Biological Halogen Bonds.

The halogen bond (or X-bond) is a noncovalent interaction that is increasingly recognized as an important design tool for engineering protein-ligand interactions and controlling the structures of proteins and nucleic acids. In the past decade, there have been significant efforts to characterize the structure-energy relationships of this interaction in macromolecules. Progress in the computational modeling of X-bonds in biological molecules, however, has lagged behind these experimental studies, with most molecular mechanics/dynamics-based simulation methods not properly treating the properties of the X-bond. We had previously derived a force field for biological X-bonds (ffBXB) based on a set of potential energy functions that describe the anisotropic electrostatic and shape properties of halogens participating in X-bonds. Although fairly accurate for reproducing the energies within biomolecular systems, including X-bonds engineered into a DNA junction, the ffBXB with its seven variable parameters was considered to be too unwieldy for general applications. In the current study, we have generalized the ffBXB by reducing the number of variables to just one for each halogen type and show that this remaining electrostatic variable can be estimated for any new halogenated molecule through a standard restricted electrostatic potential calculation of atomic charges. In addition, we have generalized the ffBXB for both nucleic acids and proteins. As a proof of principle, we have parameterized this reduced and more general ffBXB against the AMBER force field. The resulting parameter set was shown to accurately recapitulate the quantum mechanical landscape and experimental interaction energies of X-bonds incorporated into DNA junction and T4 lysozyme model systems. Thus, this reduced and generalized ffBXB is more readily adaptable for incorporation into classical molecular mechanics/dynamics algorithms, including those commonly used to design inhibitors against therapeutic targets in medicinal chemistry and materials in biomolecular engineering.

Synthesis, structures, DFT studies and luminescent properties of copper(I)-diimine complexes and application in yellow-green light-emitting diode

A series of copper(I)-diimine complexes, [Cu(bdpp)(bpy)]BF4 (1), [Cu(bdpp) (pen)]BF4 (2), [Cu(bdpp)(damp)]BF4 (3), [Cu(bdpp)(pen)]BF4 (4) (bdpp = 4,6-bis-(biphenyl phosphino) phenoxazine, bpy = 2,2′-bipyridine, phen = 1,10-Phenanthroline, dmp = 2,9-dimethyl-1,10-phenanthroline, bphen = 4,7-diphenyl-1,10-phenanthroline), have been synthesized and characterized by IR, 1H NMR, 31P NMR, XRD and X-ray crystal structure analysis. Structural analysis reveals that each Cu+ in 1–4 is four-coordinated by two N atoms and two P atoms and adopts a distorted tetrahedral geometry. DFT calculations show that the HOMOs in 1–4 are mainly comprised of bdpp and copper d–orbitals, while the LUMOs are mainly dominated by diimine ligands. Calculation of atomic charges indicates that the charges of the central Cu atoms in 1–4 are less than + 1, and the diimine ligands may be the key to the metal charge transfer to the coordinated atom. In 1–4, the maximum phosphorescent lifetimes and quantum yields are up to 349 μs and 22%, respectively. Moreover, complex 3 is used for fabricating the LED device, showing a decent performance with the maximum EQE of 11%.

ICMMS-2: Metal−organic Frameworks for Hydrogen Storage: Theoretical Prospective

The adsorption of H2 in dehydrated and hydrated Materials of Institute Lavoisier (MIL-101) was investigated theoretically. The effect of terminal water molecules on adsorption as one of the more vital MIL-n trivalent chromium-based porous carboxylates in metal-organic frameworks application in the renewable energy field was also theoretically studied. The MIL-101 structures were optimized for geometry and energy minimization was performed. The calculations were carried out using density functional theory approach with B3LYP functional and mixed basis set of Lanl2DZ and 6-31G(d, p) for Cr and light atoms (C, H, O, F), respectively, as implemented in the Gaussian 09 program package. The spin and atomic charges distribution on the Cr metal atoms, adsorbate, and water molecules are calculated using natural bond orbital (NBO). The density of states (DOS) for the clusters was obtained using Gaussian smearing of Kohn–Sham orbital energies. The natural bond orbital (NBO) for molecular orbital analysis and atomic charge calculations were utilized. For the dehydrated MIL-101, more adsorbate molecules were found near the exposed Cr2 sites than the fluorine saturated Cr1 sites. Furthermore, terminal water molecules in the hydrated MIL-101 made more interaction sites and enhanced adsorption.

Structural and Electronic Factors Governing the Action of Anticonvulsants

A selection of eighteen anticonvulsant drug molecules has been subjected to both geometrical analysis and atomic charge calculation to search for a link between structure and activity. A significant consequence of this study is the discovery of a fused 5,6,7 chemical ring system which suggests that these drugs may interact with a receptor site which is its topological complement. Molecular orbital calculations support the observation that anticonvulsant activity appears to be related to the net atomic charge on a specific carbon atom which is central to the fused ring system.
 Keywords: Anticonvulsant, lamotrigine, epilepsy, charge calculation.

Contributions to the Interpretation of Mass Spectra of the TEOS Methoxy-Transesters. Linked Scans, Accurate Mass Measurements and M+1, M+2 Isotopic Effects

The aim of this article is the study of the fragmentation reactions of the methoxy-transesters of tetraethoxysilane (TEOS), with structural formula (CH3O)x(C2H5O)ySi, initiated by electronic impact in the ionization chamber of a double focusing mass spectrometer. The initiation centers of fragmentation reactions for the neutral and ionized TEOS methoxy-transesters were established by quantum calculations of net atomic charges. The ions of the mass spectra of the TEOS methoxy-transesters were obtained experimentally by the B/E and B/E(1-E)1/2 linked scans. Thus, there can be written the fragmentation pathways for the primary events by B/E linked scan (cleavage of methoxy and ethoxy groups) and eliminations of neutral molecules (acetaldehyde, ethylene, formaldehyde, molecular hydrogen) by B/E(1-E)1/2 linked scan. Accurate mass measurements at high resolution (5000), double bond equivalent (DBE) for the molecular ions and fragmentation ions of the TEOS methoxy-transesters and M+1, M+2 isotopic effects were used to confirm their identification. The reaction pathways and accurate mass in mass spectrum of dimethoxydiethoxysilane (TR2M) are detailed.

DeepAtomicCharge: a new graph convolutional network-based architecture for accurate prediction of atomic charges.

Atomic charges play a very important role in drug-target recognition. However, computation of atomic charges with high-level quantum mechanics (QM) calculations is very time-consuming. A number of machine learning (ML)-based atomic charge prediction methods have been proposed to speed up the calculation of high-accuracy atomic charges in recent years. However, most of them used a set of predefined molecular properties, such as molecular fingerprints, for model construction, which is knowledge-dependent and may lead to biased predictions due to the representation preference of different molecular properties used for training. To solve the problem, we present a new architecture based on graph convolutional network (GCN) and develop a high-accuracy atomic charge prediction model named DeepAtomicCharge. The new GCN architecture is designed with only the atomic properties and the connection information between the atoms in molecules and can dynamically learn and convert molecules into appropriate atomic features without any prior knowledge of the molecules. Using the designed GCN architecture, substantial improvement is achieved for the prediction accuracy of atomic charges. The average root-mean-square error (RMSE) of DeepAtomicCharge is 0.0121 e, which is obviously more accurate than that (0.0180 e) reported by the previous benchmark study on the same two external test sets. Moreover, the new GCN architecture needs much lower storage space compared with other methods, and the predicted DDEC atomic charges can be efficiently used in large-scale structure-based drug design, thus opening a new avenue for high-performance atomic charge prediction and application.