Wiley Interdisciplinary Reviews: Computational Molecular Science最新文献

The design and discovery of new and improved catalysts are driving forces for accelerating scientific and technological innovations in the fields of energy conversion, environmental remediation, and chemical industry. Recently, the use of machine learning (ML) in combination with experimental and/or theoretical data has emerged as a powerful tool for identifying optimal catalysts for various applications. This review focuses on how ML algorithms can be used in computational catalysis and materials science to gain a deeper understanding of the relationships between materials properties and their stability, activity, and selectivity. The development of scientific data repositories, data mining techniques, and ML tools that can navigate structural optimization problems are highlighted, leading to the discovery of highly efficient catalysts for a sustainable future. Several data-driven ML models commonly used in catalysis research and their diverse applications in reaction prediction are discussed. The key challenges and limitations of using ML in catalysis research are presented, which arise from the catalyst's intrinsic complex nature. Finally, we conclude by summarizing the potential future directions in the area of ML-guided catalyst development.

This article is categorized under:

Graph theory has a long history in chemistry. Yet as the breadth and variety of chemical data is rapidly changing, so too do graph encoding methods and analyses that yield qualitative and quantitative insights. Using illustrative cases within a basic mathematical framework, we showcase modern chemical graph theory's utility in Chemists' analysis and model development toolkit. The encoding of both experimental and simulation data is discussed at various levels of granularity of information. This is followed by a discussion of the two major classes of graph theoretical analyses: identifying connectivity patterns and partitioning methods. Measures, metrics, descriptors, and topological indices are then introduced with an emphasis upon enhancing interpretability and incorporation into physical models. Challenging data cases are described that include strategies for studying time dependence. Throughout, we incorporate recent advancements in computer science and applied mathematics that are propelling chemical graph theory into new domains of chemical study.

This article is categorized under:

Understanding the function of eukaryotic genomes, including the human genome, is undoubtedly one of the major scientific challenges of the 21st century. The cornerstone of eukaryotic genome organization is nucleosomes—elementary building blocks of chromatin about 10 nm in size that wrap DNA around an octamer of histone proteins. Nucleosomes are integral players in all genomic processes, including transcription, DNA replication and repair. They mediate genome regulation at the epigenetic level, bridging the discrete nature of the genetic information encoded in DNA with the analog physical nature of the intermolecular interactions required to access that information. Due to their relatively large size and dynamic nature, nucleosomes are difficult objects for experimental characterization. Molecular dynamics (MD) simulations have emerged over the years as a useful tool to complement experimental studies. Particularly in recent years, advances in computing power, refinement of MD force fields and codes have opened up new frontiers in terms of simulation timescales and quality for nucleosomes and related systems. It has become possible to elucidate in atomistic detail their functional dynamics modes such as DNA unwrapping and sliding, to characterize the effects of epigenetic modifications, DNA and protein sequence variation on nucleosome structure and stability, to describe the mechanisms governing nucleosome interactions with chromatin-associated proteins and the formation of supranucleosome structures. In this review, we systematically analyzed all-atom MD simulation studies of nucleosomes and related structures published since 2018 and discussed their relevance in the context of older studies, experimental data, and related coarse-grained and multiscale studies.

This article is categorized under:

A transformer is the foundational architecture behind large language models designed to handle sequential data by using mechanisms of self-attention to weigh the importance of different elements, enabling efficient processing and understanding of complex patterns. Recently, transformer-based models have become some of the most popular and powerful deep learning (DL) algorithms in molecular science, owing to their distinctive architectural characteristics and proficiency in handling intricate data. These models leverage the capacity of transformer architectures to capture complex hierarchical dependencies within sequential data. As the applications of transformers in molecular science are very widespread, in this review, we only focus on the technical aspects of transformer technology in molecule domain. Specifically, we will provide an in-depth investigation into the algorithms of transformer-based machine learning techniques in molecular science. The models under consideration include generative pre-trained transformer (GPT), bidirectional and auto-regressive transformers (BART), bidirectional encoder representations from transformers (BERT), graph transformer, transformer-XL, text-to-text transfer transformer, vision transformers (ViT), detection transformer (DETR), conformer, contrastive language-image pre-training (CLIP), sparse transformers, and mobile and efficient transformers. By examining the inner workings of these models, we aim to elucidate how their architectural innovations contribute to their effectiveness in processing complex molecular data. We will also discuss promising trends in transformer models within the context of molecular science, emphasizing their technical capabilities and potential for interdisciplinary research. This review seeks to provide a comprehensive understanding of the transformer-based machine learning techniques that are driving advancements in molecular science.

This article is categorized under:

Polarizable continuum solvation models are popular in both, quantum chemistry and in biophysics, though typically with different requirements for the numerical methods. However, the recent trend of multiscale modeling can be expected to blur field-specific differences. In this regard, numerical methods based on domain decomposition (dd) have been demonstrated to be sufficiently flexible to be applied all across these levels of theory while remaining systematically accurate and efficient. In this contribution, we present ddX, an open-source implementation of dd-methods for various solvation models, which features a uniform interface with classical as well as quantum descriptions of the solute, or any hybrid versions thereof. We explain the key concepts of the library design and its application program interface, and demonstrate the use of ddX for integrating into standard chemistry packages. Numerical tests illustrate the performance of ddX and its interfaces.

This article is categorized under:

Many-body open quantum systems (OQSs) have a profound impact on various subdisciplines of physics, chemistry, and biology. Thus, the development of a computer program capable of accurately, efficiently, and versatilely simulating many-body OQSs is highly desirable. In recent years, we have focused on the advancement of numerical algorithms based on the fermionic hierarchical equations of motion (HEOM) theory. Being in-principle exact, this approach allows for the precise characterization of many-body correlations, non-Markovian memory, and non-equilibrium thermodynamic conditions. These efforts now lead to the establishment of a new computer program, HEOM for QUantum Impurity with a Correlated Kernel, version 2 (HEOM-QUICK2), which, to the best of our knowledge, is currently the only general-purpose simulator for fermionic many-body OQSs. Compared with version 1, the HEOM-QUICK2 program features more efficient solvers for stationary states, more accurate treatment of non-Markovian memory, and improved numerical stability for long-time dissipative dynamics. Integrated with quantum chemistry software, HEOM-QUICK2 has become a valuable theoretical tool for the precise simulation of realistic many-body OQSs, particularly the single atomic or molecular junctions. Furthermore, the unprecedented precision achieved by HEOM-QUICK2 enables accurate simulation of low-energy spin excitations and coherent spin relaxation. The unique usefulness of HEOM-QUICK2 is demonstrated through several examples of strongly correlated quantum impurity systems under non-equilibrium conditions. Thus, the new HEOM-QUICK2 program offers a powerful and comprehensive tool for studying many-body OQSs with exotic quantum phenomena and exploring applications in various disciplines.

This article is categorized under:

Reinforcement learning (RL) is one important branch of artificial intelligence (AI), which intuitively imitates the learning style of human beings. It is commonly derived from solving game playing problems and is extensively used for decision-making, control and optimization problems. It has been extensively applied for solving complicated problems with the property of Markov decision-making processes. With data accumulation and comprehensive analysis, researchers are not only satisfied with predicting the results for experimental systems but also hope to design or control them for the sake of obtaining the desired properties or functions. RL is potentially facilitated to solve a large number of complicated biological and chemical problems, because they could be decomposed into multi-step decision-making process. In practice, substantial progress has been made in the application of RL to the field of biomedicine. In this paper, we will first give a brief description about RL, including its definition, basic theory and different type of methods. Then we will review some detailed applications in various domains, for example, molecular design, reaction planning, molecular simulation and etc. In the end, we will summarize the essentialities of RL approaches to solve more diverse problems compared with other machine learning methods and also outlook the possible trends to overcome their limitations in the future.

This article is categorized under:

Orbital-free density functional theory (OFDFT) stands out as a many-body electronic structure approach with a low computational cost that scales linearly with system size, making it well suitable for large-scale simulations. The past decades have witnessed impressive progress in OFDFT, which opens a new avenue to capture the complexity of realistic systems (e.g., solids, liquids, and warm dense matters) and provide a complete description of some complicated physical phenomena under realistic conditions (e.g., dislocation mobility, ductile processes, and vacancy diffusion). In this review, we first present a concise summary of the major methodological advances in OFDFT, placing particular emphasis on kinetic energy density functional and the schemes to evaluate the electron–ion interaction energy. We then give a brief overview of the current status of OFDFT developments in finite-temperature and time-dependent regimes, as well as our developed OFDFT-based software package, named by ATLAS. Finally, we highlight perspectives for further development in this fascinating field, including the major outstanding issues to be solved and forthcoming opportunities to explore large-scale materials.

This article is categorized under:

Structure-based drug design is a widely applied approach in the discovery of new lead compounds for known therapeutic targets. In most structure-based drug design applications, the docking procedure is considered the crucial step. Here, a potential ligand is fitted into the binding site, and a scoring function assesses its binding capability. With the rise of modern machine-learning in drug discovery, novel scoring functions using machine-learning techniques achieved significant performance gains in virtual screening and ligand optimization tasks on retrospective data. However, real-world applications of these methods are still limited. Missing success stories in prospective applications are one reason for this. Additionally, the fast-evolving nature of the field makes it challenging to assess the advantages of each individual method. This review will highlight recent strides toward improved real world applicability of machine-learning based scoring, enabling a better understanding of the potential benefits and pitfalls of these functions on a project. Furthermore, a systematic way of classifying machine-learning based scoring that facilitates comparisons will be presented.

This article is categorized under:

For dealing with complex reaction (CR) systems that show typical chemical phenomena in molecular aggregation states, the Red Moon (RM) approach is introduced based on a new efficient and systematic RM methodology. First, the theoretical background with my motivation to develop the RM approach is presented from the recent necessity to perform ‘atomistic’ molecular simulation of large-scale and long-term phenomena of (i) complex chemical reactions, (ii) stereospecificity, and (iii) aggregation structures. The RM methodology uses both the molecular dynamics (MD) method for molecular motions (translation, rotation, and vibration of molecules) that frequently occur on a short-time scale and the Monte Carlo (MC) method for rare events such as chemical reactions that hardly do on that time scale. Then, under the transition rate using both the potential energy difference before and after a rare event trial and its chemical kinetic probability, it is tested and judged by the MC method whether the trial is possible (Metropolis method). Next, typical applications of the RM approach are reviewed in two main research fields, (i) polymerization and (ii) storage battery (rechargeable battery or secondary cell), with various examples of our successful studies. Finally, we conclude that the RM approach using the RM methodology should become an efficient new-generation approach as one promising computational molecular strategy (CMT). We believe it will play an essential role in surveying, at the multilevel resolution, various specificities of CR systems in molecular aggregation states.

This article is categorized under: