CEEM: a Chemically Explainable Deep Learning Platform for Identifying Compounds with Low Effective Mass.

The semiconductor industry occupies a crucial position in the fields of integrated circuits, energy, and communication systems. Effective mass (mE ), which is closely related to electron transition, thermal excitation, and carrier mobility, is a key performance indicator of semiconductor. However, the highly neglected mE is onerous to measure experimentally, which seriously hinders the evaluation of semiconductor properties and the understanding of the carrier migration mechanisms. Here, a chemically explainable effective mass predictive platform (CEEM) is constructed by deep learning, to identify n-type and p-type semiconductors with low mE . Based on the graph network, a versatile explainable network is innovatively designed that enables CEEM to efficiently predict the mE of any structure, with the area under the curve of 0.904 for n-type semiconductors and 0.896 for p-type semiconductors, and derive the most relevant chemical factors. Using CEEM, the currently largest mE database is built that contains 126 335 entries and screens out 466 semiconductors with low mE for transparent conductive materials, photovoltaic materials, and water-splitting materials. Moreover, a user-friendly and interactive CEEM web is provided that supports query, prediction, and explanation of mE . CEEM's high efficiency, accuracy, flexibility, and explainability open up new avenues for the discovery and design of high-performance semiconductors.

RotNet: A Rotationally Invariant Graph Neural Network for Quantum Mechanical Calculations.

Deep learning has proven promising in biological and chemical applications, aiding in accurate predictions of properties such as atomic forces, energies, and material band gaps. Traditional methods with rotational invariance, one of the most crucial physical laws for predictions made by machine learning, have relied on Fourier transforms or specialized convolution filters, leading to complex model design and reduced accuracy and efficiency. However, models without rotational invariance exhibit poor generalization ability across datasets. Addressing this contradiction, this work proposes a rotationally invariant graph neural network, named RotNet, for accurate and accelerated quantum mechanical calculations that can overcome the generalization deficiency caused by rotations of molecules. RotNet ensures rotational invariance through an effective transformation and learns distance and angular information from atomic coordinates. Benchmark experiments on three datasets (protein fragments, electronic materials, and QM9) demonstrate that the proposed RotNet framework outperforms popular baselines and generalizes well to spatial data with varying rotations. The high accuracy, efficiency, and fast convergence of RotNet suggest that it has tremendous potential to significantly facilitate studies of protein dynamics simulation and materials engineering while maintaining physical plausibility.

MatGPT: A Vane of Materials Informatics from Past, Present, to Future.

Combining materials science, artificial intelligence (AI), physical chemistry, and other disciplines, materials informatics is continuously accelerating the vigorous development of new materials. The emergence of "GPT (Generative Pre-trained Transformer) AI" shows that the scientific research field has entered the era of intelligent civilization with "data" as the basic factor and "algorithm + computing power" as the core productivity. The continuous innovation of AI will impact the cognitive laws and scientific methods, and reconstruct the knowledge and wisdom system. This leads to think more about materials informatics. Here, a comprehensive discussion of AI models and materials infrastructures is provided, and the advances in the discovery and design of new materials are reviewed. With the rise of new research paradigms triggered by "AI for Science", the vane of materials informatics: "MatGPT", is proposed and the technical path planning from the aspects of data, descriptors, generative models, pretraining models, directed design models, collaborative training, experimental robots, as well as the efforts and preparations needed to develop a new generation of materials informatics, is carried out. Finally, the challenges and constraints faced by materials informatics are discussed, in order to achieve a more digital, intelligent, and automated construction of materials informatics with the joint efforts of more interdisciplinary scientists.

A Data-Driven Platform for Two-Dimensional Hybrid Lead-Halide Perovskites.

The exceptional properties of two-dimensional hybrid organic-inorganic lead-halide perovskites (2D HOIPs) have led to a rapid increase in the number of low-dimensional materials for optoelectronic engineering and solar energy conversion. The flexibility and controllability of 2D HOIPs create a vast structural space, which presents an urgent issue to effectively explore 2D HOIPs with better performance for practical applications. However, the traditional RP-DJ classification method falls short in describing the influence of structure on the electronic properties of 2D HOIPs. To overcome this limitation, we employed inorganic structure factors (SF) as a classification descriptor, which considers the influence of inorganic layer distortion of 2D HOIPs. And we investigated the relationship between SF, other physicochemical features, and band gaps of 2D HOIPs. By using this structural descriptor as a feature for a machine learning model, a database of 304920 2D HOIPs and their structural and electronic properties was generated. A large number of previously neglected 2D HOIPs were discovered. With the establishment of this database, experimental data and machine learning methods were combined to develop a 2D HOIPs exploration platform. This platform integrates searching, download, analysis, and online prediction, providing a useful tool for the further discovery of 2D HOIPs.

Binding Affinity and Mechanisms of Potential Antidepressants Targeting Human NMDA Receptors.

Depression, a mental disorder that plagues the world, is a burden on many families. There is a great need for new, fast-acting antidepressants to be developed. N-methyl-D-aspartic acid (NMDA) is an ionotropic glutamate receptor that plays an important role in learning and memory processes and its TMD region is considered as a potential target to treat depression. However, due to the unclear binding sites and pathways, the mechanism of drug binding lacks basic explanation, which brings great complexity to the development of new drugs. In this study, we investigated the binding affinity and mechanisms of an FDA-approved antidepressant (S-ketamine) and seven potential antidepressants (R-ketamine, memantine, lanicemine, dextromethorphan, Ro 25-6981, ifenprodil, and traxoprodil) targeting the NMDA receptor by ligand-protein docking and molecular dynamics simulations. The results indicated that Ro 25-6981 has the strongest binding affinity to the TMD region of the NMDA receptor among the eight selected drugs, suggesting its potential effective inhibitory effect. We also calculated the critical binding-site residues at the active site and found that residues Leu124 and Met63 contributed the most to the binding energy by decomposing the free energy contributions on a per-residue basis. We further compared S-ketamine and its chiral molecule, R-ketamine, and found that R-ketamine had a stronger binding capacity to the NMDA receptor. This study provides a computational reference for the treatment of depression targeting NMDA receptors, and the proposed results will provide potential strategies for further antidepressant development and is a useful resource for the future discovery of fast-acting antidepressant candidates.

Machine-learning-assisted rational design of 2D doped tellurene for fin field-effect transistor devices.

Fin field-effect transistors (FinFETs) have been widely used in electronic devices on account of their excellent performance, but this new type of device is facing many challenges because of size constraints. Two-dimensional (2D) materials with a layer structure can meet the required thickness of FinFETs and provide ideal carrier transport performance. In this work, we used 2D tellurene as the parent material and modified it with doping techniques to improve electronic device performance. High-performance FinFET devices were prepared with 23 systems screened from 385 doping systems by a combination of first-principle calculations and a machine-learning (ML) model. Moreover, theoretical calculations demonstrated that 1S1@Te and 2S2@Te have high carrier mobility and stability with an electron mobility and a hole mobility of 6.211 × 104 cm2 V-1 S-1 and 1.349 × 104 cm2 V-1 S-1, respectively. This work can provide a reference for subsequent experiments and advance the development of functional materials by using an ML-assisted design paradigm.

Single atom modified two-dimensional bismuthenes for toxic gas detection.

Accurate detection of toxic gases at low concentrations is often difficult because they are colorless, odorless, flammable and denser than air. Therefore, it is urgent to develop highly stable and sensitive toxic gas detectors. However, most gas sensors operate at high temperatures, making the detection of toxic gases more challenging. Two-dimensional materials with high specific surface area and abundant modulation methods of properties provide new inspirations for the development of new toxic gas sensing materials. Here, bismuthene, a single element two-dimensional material with high carrier mobility and excellent stability, was used as a substrate material to investigate the effects of anchoring and doping on its gas detection performance by density functional theory (DFT) calculations. It is revealed that the surface structure altered by single metal atoms (Ba, Be, Ca, K, Li, Mg, Na, and Sr) can promote the improvement of gas detection sensitivity. Buckled honeycomb bismuthene (bBi) with the Be atom anchored (A-Be-Bi) show superior sensitivity to H2S, while D-Ca-Bi, D-Li-Bi, D-Mg-Bi and D-Sr-Bi also have relatively high toxic gas detection sensitivity. We further discussed the recovery times of these modified bBis at various temperatures to determine the potential for applications. The ultra-fast recovery time of less than 0.5 seconds demonstrates the potential of these systems at room temperature and can be applied to the manufacture of toxic gas sensors used under practical sensing conditions.

A deep transfer learning-based protocol accelerates full quantum mechanics calculation of protein.

Effective full quantum mechanics (FQM) calculation of protein remains a grand challenge and of great interest in computational biology with substantial applications in drug discovery, protein dynamic simulation and protein folding. However, the huge computational complexity of the existing QM methods impends their applications in large systems. Here, we design a transfer-learning-based deep learning (TDL) protocol for effective FQM calculations (TDL-FQM) on proteins. By incorporating a transfer-learning algorithm into deep neural network (DNN), the TDL-FQM protocol is capable of performing calculations at any given accuracy using models trained from small datasets with high-precision and knowledge learned from large amount of low-level calculations. The high-level double-hybrid DFT functional and high-level quality of basis set is used in this work as a case study to evaluate the performance of TDL-FQM, where the selected 15 proteins are predicted to have a mean absolute error of 0.01 kcal/mol/atom for potential energy and an average root mean square error of 1.47 kcal/mol/$ {\rm A^{^{ \!\!\!o}}} $ for atomic forces. The proposed TDL-FQM approach accelerates the FQM calculation more than thirty thousand times faster in average and presents more significant benefits in efficiency as the size of protein increases. The ability to learn knowledge from one task to solve related problems demonstrates that the proposed TDL-FQM overcomes the limitation of standard DNN and has a strong power to predict proteins with high precision, which solves the challenge of high precision prediction in large chemical and biological systems.

Phase Transitions and Stabilities among Three Phases of Di-p-tolyl Disulfides.

Di-p-tolyl disulfides (p-Tol2S2) are employed as load-carrying additives because of their anti-wear and extreme load-bearing qualities. External pressure triggers conformational up-conversion (leads to phase transition) in the molecules of p-Tol2S2, by compensating for the stress and absorbing its energy. These features make p-Tol2S2 a potential candidate for next-generation energy storage devices. Upon lithiation, MoS2 expands up to 103% which causes stress and affects battery stability and performance. Therefore, it is essential to study these materials under different physical conditions. In this work, we used density functional theory (DFT) at ωB97XD/6-31G* functional level, to calculate lattice parameters, Gibbs free energies, and vibrational spectra of three phases (i.e., α, ß, and γ) of p-Tol2S2 under different pressure and temperature conditions. The phase transition between phases α and ß occurred at a pressure and temperature of 0.65 GPa and 463 K, respectively. Furthermore, phase transition between phases α and γ was found at a pressure and temperature of 0.35 GPa and 400 K, respectively. Moreover, no phase transition was observed between phases ß and γ under the pressure range studied (0 GPa to 5.5 GPa). We also computed and compared the FT-IR spectra of the three phases. These results can guide scientists and chemists in designing more stable battery materials.

Clustered tree regression to learn protein energy change with mutated amino acid.

Accurate and effective prediction of mutation-induced protein energy change remains a great challenge and of great interest in computational biology. However, high resource consumption and insufficient structural information of proteins severely limit the experimental techniques and structure-based prediction methods. Here, we design a structure-independent protocol to accurately and effectively predict the mutation-induced protein folding free energy change with only sequence, physicochemical and evolutionary features. The proposed clustered tree regression protocol is capable of effectively exploiting the inherent data patterns by integrating unsupervised feature clustering by K-means and supervised tree regression using XGBoost, and thus enabling fast and accurate protein predictions with different mutations, with an average Pearson correlation coefficient of 0.83 and an average root-mean-square error of 0.94kcal/mol. The proposed sequence-based method not only eliminates the dependence on protein structures, but also has potential applications in protein predictions with rare structural information.

An inductive transfer learning force field (ITLFF) protocol builds protein force fields in seconds.

Accurate simulation of protein folding is a unique challenge in understanding the physical process of protein folding, with important implications for protein design and drug discovery. Molecular dynamics simulation strongly requires advanced force fields with high accuracy to achieve correct folding. However, the current force fields are inaccurate, inapplicable and inefficient. We propose a machine learning protocol, the inductive transfer learning force field (ITLFF), to construct protein force fields in seconds with any level of accuracy from a small dataset. This process is achieved by incorporating an inductive transfer learning algorithm into deep neural networks, which learn knowledge of any high-level calculations from a large dataset of low-level method. Here, we use a double-hybrid density functional theory (DFT) as a case functional, but ITLFF is suitable for any high-precision functional. The performance of the selected 18 proteins indicates that compared with the fragment-based double-hybrid DFT algorithm, the force field constructed by ITLFF achieves considerable accuracy with a mean absolute error of 0.0039 kcal/mol/atom for energy and a root mean square error of 2.57 $\mathrm{kcal}/\mathrm{mol}/{\AA}$ for force, and it is more than 30 000 times faster and obtains more significant efficiency benefits as the system increases. The outstanding performance of ITLFF provides promising prospects for accurate and efficient protein dynamic simulations and makes an important step toward protein folding simulation. Due to the ability of ITLFF to utilize the knowledge acquired in one task to solve related problems, it is also applicable for various problems in biology, chemistry and material science.

An Ensemble Learning Platform for the Large-Scale Exploration of New Double Perovskites.

Lead-free double perovskites are regarded as stable and green optoelectronic alternatives to single perovskites, but may exhibit indirect band gaps and high effective masses, thus limiting their maximum photovoltaic efficiency. Considering that the trial-and-error experimental and computational approaches cannot quickly identify ideal candidates, we propose an ensemble learning workflow to screen all suitable double perovskites from the periodic table, with a high predictive accuracy of 92% and a computed speed that is ∼108 faster than ab initio calculations. From ∼23 314 unexplored double perovskites, we successfully identify six candidates that exhibit suitable band gaps (1.0-2.0 eV), where two have direct band gaps and low effective masses. They all show good thermal stabilities that are hopefully able to be synthesized. The proposed ML workflow immensely shortens the screening cycle for double perovskites, which will greatly promote the development and application of photovoltaic devices.

Quantum Mechanical-Based Stability Evaluation of Crystal Structures for HIV-Targeted Drug Cabotegravir.

The long-acting parenteral formulation of the HIV integrase inhibitor cabotegravir (GSK744) is currently being developed to prevent HIV infections, benefiting from infrequent dosing and high efficacy. The crystal structure can affect the bioavailability and efficacy of cabotegravir. However, the stability determination of crystal structures of GSK744 have remained a challenge. Here, we introduced an ab initio protocol to determine the stability of the crystal structures of pharmaceutical molecules, which were obtained from crystal structure prediction process starting from the molecular diagram. Using GSK744 as a case study, the ab initio predicted that Gibbs free energy provides reliable further refinement of the predicted crystal structures and presents its capability for becoming a crystal stability determination approach in the future. The proposed work can assist in the comprehensive screening of pharmaceutical design and can provide structural predictions and stability evaluation for pharmaceutical crystals.

Binding affinity and mechanisms of SARS-CoV-2 variants.

During the rapid worldwide spread of SARS-CoV-2, the viral genome has been undergoing numerous mutations, especially in the spike (S) glycoprotein gene that encode a type-I fusion protein, which plays an important role in the infectivity and transmissibility of the virus into the host cell. In this work, we studied the effect of S glycoprotein residue mutations on the binding affinity and mechanisms of SARS-CoV-2 using molecular dynamics simulations and sequence analysis. We quantitatively determined the degrees of binding affinity caused by different S glycoprotein mutations, and the result indicated that the 501Y.V1 variant yielded the highest enhancements in binding affinity (increased by 36.8%), followed by the N439K variant (increased by 29.5%) and the 501Y.V2 variant (increased by 19.6%). We further studied the structures, chemical bonds, binding free energies (enthalpy and entropy), and residue contribution decompositions of these variants to provide physical explanations for the changes in SARS-CoV-2 binding affinity caused by these residue mutations. This research identified the binding affinity differences of the SARS-CoV-2 variants and provides a basis for further surveillance, diagnosis, and evaluation of mutated viruses.

Machine learning builds full-QM precision protein force fields in seconds.

Full-quantum mechanics (QM) calculations are extraordinarily precise but difficult to apply to large systems, such as biomolecules. Motivated by the massive demand for efficient calculations for large systems at the full-QM level and by the significant advances in machine learning, we have designed a neural network-based two-body molecular fractionation with conjugate caps (NN-TMFCC) approach to accelerate the energy and atomic force calculations of proteins. The results show very high precision for the proposed NN potential energy surface models of residue-based fragments, with energy root-mean-squared errors (RMSEs) less than 1.0 kcal/mol and force RMSEs less than 1.3 kcal/mol/Å for both training and testing sets. The proposed NN-TMFCC method calculates the energies and atomic forces of 15 representative proteins with full-QM precision in 10-100 s, which is thousands of times faster than the full-QM calculations. The computational complexity of the NN-TMFCC method is independent of the protein size and only depends on the number of residue species, which makes this method particularly suitable for rapid prediction of large systems with tens of thousands or even hundreds of thousands of times acceleration. This highly precise and efficient NN-TMFCC approach exhibits considerable potential for performing energy and force calculations, structure predictions and molecular dynamics simulations of proteins with full-QM precision.

Ab initio determination of crystal stability of di-p-tolyl disulfide.

With the rapid growth of energy demand and the depletion of existing energy resources, the new materials with superior performances, low costs and environmental friendliness for energy production and storage are explored. Di-p-tolyl disulfide (p-Tol2S2) is a typical lubricating material, which has been applied in the field of energy storage. The conformational properties and phase transformations of p-Tol2S2 have been studied by pioneers, but their polymorphs and the polymorphism induced crystal structure changes require further analysis. In this study, we perform the crystal structural screening, prediction and optimization of p-Tol2S2 crystal with quantum mechanical calculations, i.e., density functional theory (DFT) and second-order Møller-Plesset perturbation (MP2) methods. A series of crystal structures with different molecular arrangements are generated based on the crystal structure screening. As compared to long-established lattice energy calculation, we take an advantage of using more accurate technique, which is Gibbs free energy calculation. It considers the effects of entropy and temperature to predict the crystal structures and energy landscape. By comparing the Gibbs free energies between predicted and experimental structures, we found that phase α is the most stable structure for p-Tol2S2 crystal at ambient temperature and standard atmospheric pressure. Furthermore, we provide an efficient method to discriminate different polymorphs that are otherwise difficult to be identified based on the Raman/IR spectra. The proposed work enable us to evaluate the quality of various crystal polymorphs rapidly.

Neural Networks Accelerate the Ab Initio Prediction of Solid-Solid Phase Transitions at High Pressures.

High-level ab initio chemical calculations, such as second-order Møller-Plesset perturbation (MP2), are highly accurate but time-consuming, making it inefficient to apply to macromolecular systems. Here, we propose a newly efficient approach based on the neural network and fragment method to predict the Gibbs free energy, structural characteristics, and thus phase transition of solid crystal structures. The proposed approach has the same prediction accuracy as the MP2 calculation but is hundreds of times faster than the MP2. The predicted structures and phase transitions of two selected ice phases (IX and XV) under extreme conditions are in excellent agreement with the MP2 calculations and experimental results but with an extremely low computational cost. It not only predicts the high-pressure structures and phase diagrams of solid systems accurately and efficiently but also solves the problem of extreme calculation cost during a high-precision theoretical study on high-pressure molecular crystals with potentially essential applications.

Potential inhibitors for the novel coronavirus (SARS-CoV-2).

The lack of a vaccine or any effective treatment for the aggressive novel coronavirus disease (COVID-19) has created a sense of urgency for the discovery of effective drugs. Several repurposing pharmaceutical candidates have been reported or envisaged to inhibit the emerging infections of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), but their binding sites, binding affinities and inhibitory mechanisms are still unavailable. In this study, we use the ligand-protein docking program and molecular dynamic simulation to ab initio investigate the binding mechanism and inhibitory ability of seven clinically approved drugs (Chloroquine, Hydroxychloroquine, Remdesivir, Ritonavir, Beclabuvir, Indinavir and Favipiravir) and a recently designed α-ketoamide inhibitor (13b) at the molecular level. The results suggest that Chloroquine has the strongest binding affinity with 3CL hydrolase (Mpro) among clinically approved drugs, indicating its effective inhibitory ability for SARS-CoV-2. However, the newly designed inhibitor 13b shows potentially improved inhibition efficiency with larger binding energy compared with Chloroquine. We further calculate the important binding site residues at the active site and demonstrate that the MET 165 and HIE 163 contribute the most for 13b, while the MET 165 and GLN 189 for Chloroquine, based on residual energy decomposition analysis. The proposed work offers a higher research priority for 13b to treat the infection of SARS-CoV-2 and provides theoretical basis for further design of effective drug molecules with stronger inhibition.

Ab Initio Prediction of the Phase Transition for Solid Ammonia at High Pressures.

Ammonia is one of the most basic components on the planet and its high-pressure characteristics play an important role in planetary science. Solid ammonia crystals frequently adopt multiple distinct polymorphs exhibiting different properties. Predicting the crystal structure of these polymorphs and under what thermodynamic conditions these polymorphs are stable would be of great value to environmental industry and other fields. Theoretical calculations based on the classical force fields and density-functional theory (DFT) are versatile methods but lack of accurate description of weak intermolecular interactions for molecular crystals. In this study, we employ an ab initio computational study on the solid ammonia at high pressures, using the second-order Møller-Plesset perturbation (MP2) theory and the coupled cluster singles, doubles, and perturbative triples (CCSD(T)) theory along with the embedded fragmentation method. The proposed algorithm is capable of performing large-scale calculations using high-level wavefunction theories, and accurately describing covalent, ionic, hydrogen bonding, and dispersion interactions within molecular crystals, and therefore can predict the crystal structures, Raman spectra and phase transition of solid ammonia phases I and IV accurately. We confirm the crystal structures of solid ammonia phases I and IV that have been controversial for a long time and predict their phase transition that occurs at 1.17 GPa and 210 K with small temperature dependence, which is in line with experiment.

Combining the Fragmentation Approach and Neural Network Potential Energy Surfaces of Fragments for Accurate Calculation of Protein Energy.

Accurate and efficient all-atom quantum mechanical (QM) calculations for biomolecules still present a challenge to computational physicists and chemists. In this study, an extensible generalized molecular fractionation with a conjugate caps method combined with neural networks (NN-GMFCC) is developed for efficient QM calculation of protein energy. In the NN-GMFCC scheme, the total energy of a given protein is calculated by taking a proper combination of the high-precision neural network potential energies of all capped residues and overlapping conjugate caps. In addition, the two-body interaction energies of residue pairs are calculated by molecular mechanics (MM). With reference to the GMFCC/MM calculation at the ωB97XD/6-31G* level, the overall mean unsigned errors of the energy deviations and atomic force root-mean-squared errors calculated by NN-GMFCC are only 2.01 kcal/mol and 0.68 kcal/mol/Å, respectively, for 14 proteins (containing up to 13,728 atoms). Meanwhile, the NN-GMFCC approach is about 4 orders of magnitude faster than the GMFCC/MM method. The NN-GMFCC method could be systematically improved by inclusion of two-body QM interaction and multibody electronic polarization effect. Moreover, the NN-GMFCC approach can also be applied to other macromolecular systems such as DNA/RNA, and it is capable of providing a powerful and efficient approach for exploration of structures and functions of proteins with QM accuracy.