Mutations that cause structural changes in proteins can sometimes reduce drug efficacy dramatically, a phenomenon known as mutation-induced drug resistance. For example, emerging drug-resistant mutations in the SARS-CoV-2 main protease $({\rm M}^{pro})$ threaten the long-term efficacy of nirmatrelvir, the active component of Paxlovid. Various methods have been developed to predict the impact of such mutations, with differing levels of reliability. In this study, comparative binding free energy calculations using Molecular Mechanics/Poisson–Boltzmann Surface Area (MM/PBSA) and Alchemical Transformation (also known as Free Energy Perturbation, or FEP) were performed to assess five naturally occurring Mpro mutations (SER144ALA, MET165ALA, GLU166ALA, HIE172ALA, and GLN192ALA) at the nirmatrelvir binding site. The results reveal a weak correlation $(R_{Pearson}=0.18)$ between MM/PBSA predictions and experimental data. In contrast, FEP calculations using either the Multistate Bennett Acceptance Ratio (MBAR) or Thermodynamic Integration (TI) yield stronger linear correlations $(R_{Pearson} = 0.56$ and $0.57,$ respectively). This study highlights the superior reliability of FEP in quantifying binding affinity losses due to drug resistance and underscores its potential for the proactive surveillance of clinical resistance mutations. Moreover, such insights are crucial for advancing antiviral drug development and guiding the design of inhibitors with a reduced risk of resistance evolution.

Organic optoelectronic materials, owing to their exceptional photoelectronic properties, have extensive applications across diverse fields, such as lighting and display, photovoltaic devices, and bioimaging. Machine learning (ML) provides new opportunities for advancing research on organic optoelectronic materials. ML leverages existing datasets to establish robust input-output correlations for predicting material properties, thereby substantially reducing computational costs and enhancing efficiency. This review comprehensively explores recent progress on ML applications for organic optoelectronic material. We focused on three key aspects. First, we review applications ML in predicting photophysical properties of organic dyes, including absorption/emission wavelengths, quantum yields, and aggregation-induced emission/aggregation-caused quenching effects. Second, we examine ML applications in predicting subcellular targeting of fluorescent probes. Third, we discuss the role of ML in screening key descriptors for organic photovoltaics material. The advances in data science position ML as a pivotal tool for elucidating intricate structure-property correlations in molecular systems, driving the accelerated innovation of optoelectronic devices.

This study presents a detailed examination of enzyme-catalyzed pericyclic reactions, specifically focusing on the synthesis of spirotetronate (decatromicin), catalyzed by the Plol4 enzyme. Utilizing the QM/MM method, we have discovered that the Plol4 enzyme significantly reduces the energy barrier in the Diels-Alder (DA) reaction, demonstrating its effectiveness in facilitating complex chemical processes. Our analysis suggests that this reduction is partly due to the interaction between the substrate and active site residues within the enzyme, which prearranges the substrate in a manner conducive to the [4+2] cycloaddition. Moreover, we observed a notable shift in the DA reaction mechanism from a concerted approach within the gas phase to an asynchronous approach within the enzymatic environment. This finding highlights the unique role of Plol4 in altering reaction pathways and underscores the enzyme's potential in synthetic applications. Overall, this research provides valuable insights into the mechanisms of enzyme-catalyzed reactions and their implications for the synthesis of polycyclic compounds.

Expanded porphyrins serve as promising candidates for MRI contrast agents and sensitizers in photodynamic therapy. In this study, we theoretically designed a series of expanded porphyrins incorporating thiophene and selenophene moieties to investigate their optoelectronic properties. Density functional theory (DFT) and time-dependent DFT (TD-DFT) calculations were performed to assess their aromaticity, stability, and photophysical characteristics. Our results reveal that all designed molecules exhibit superior optoelectronic performance, with enhanced aromaticity compared to conventional porphyrins. The absorption spectra of the molecules closely resemble that of porphyrins, suggesting potential applicability in related biomedical and photonic applications. Notably, molecule 4, featuring both a thiophene moiety and a conventional selenophene ring, demonstrates the highest stability, an increased energy gap highest between occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO), and a planar geometry, leading to strengthened aromaticity. These findings provide valuable insights for the rational design of next-generation porphyrin-based materials.

Accurately modeling the dynamics of open quantum systems is critical for advancing quantum technologies, yet traditional methods often struggle with balancing accuracy and efficiency. Machine learning (ML) offers a promising alternative, particularly through recursive models that predict system evolution based on the past history. While these models have shown success in predicting single observables, their effectiveness in more complex tasks, such as forecasting the full reduced density matrix (RDM), remains unclear. In this work, we extend history-based recursive ML approaches to complex quantum systems, comparing four physics-informed neural network (PINN) architectures: (i) single-RDM-predicting PINN (SR-PINN), (ii) SR-PINN with simulation parameters (PSR-PINN), (iii) multi-RDMs-predicting PINN (MR-PINN), and (iv) MR-PINN with simulation parameters (PMR-PINN). We apply these models to two representative open quantum systems: the spin-boson (SB) model and the Fenna-Matthews-Olson (FMO) complex. Our results demonstrate that single-RDM-predicting models (SR-PINN and PSR-PINN) are limited by a narrow history window, failing to capture the full complexity of quantum evolution and resulting in unstable long-term predictions, especially in nonlinear and highly correlated dynamics. In contrast, multi-RDMs-predicting models (MR-PINN and PMR-PINN) provide more accurate predictions by extending the forecast horizon, incorporating long-range temporal correlations, and mitigating error propagation. Surprisingly, including simulation parameters explicitly, such as temperature and reorganization energy, in PSR-PINN and PMR-PINN does not consistently improve accuracy and, in some cases, even reduces performance. This suggests that these parameters are already implicitly encoded in the RDM evolution, making their inclusion redundant and adding unnecessary complexity. These findings highlight the limitations of short-sighted recursive forecasting in complex quantum systems and demonstrate the superior stability and accuracy of far-sighted approaches for long-term predictions.

Semiclassical mapping dynamics offer a computationally tractable approach for simulating nonadiabatic processes in complex molecular systems. This work presents a comparative study of purely data-driven (DD) neural networks and physics-informed neural networks (PINNs) for learning the Markovian propagation of trajectories within the Meyer-Miller-Stock-Thoss (MMST) mapping Hamiltonian framework. Using the spin-boson model as a benchmark, we assess the performance of both approaches in reproducing the nonadiabatic dynamical details for classical mapping model (CMM) and symmetrical quasiclassical (SQC) dynamics. Our results demonstrate that PINNs, which explicitly incorporate the physical equations of motion, significantly outperform DD models, especially when trained with limited datasets. PINNs accurately capture the population dynamics and preserve the physical correlations in phase space, regardless of the underlying neural network architecture (fully connected network or gated recurrent unit) or the specific MMST Hamiltonian formulation. In contrast, DD models exhibit substantial inaccuracies and unphysical behaviors. This clearly shows the key benefit of embedding physical laws into machine learning frameworks to achieve data-efficient, accurate, and reliable simulations of nonadiabatic phenomena.

The integration of artificial intelligence (AI) and robotics into chemical synthesis has given rise to autonomous laboratories, transformative systems designed to overcome limitations in traditional experimental approaches. This review synthesizes recent advancements in autonomous laboratory systems, highlighting their applications in chemical synthesis and the innovations driving their evolution. Autonomous laboratories combine automated hardware, intelligent software, and adaptive systems to optimize experimental workflows, reduce human intervention, and enhance efficiency in complex reaction environments. Key developments include AI-driven reaction pathway planning, closed-loop optimization frameworks, and robotic platforms capable of executing multi-step synthesis with minimal expert oversight. Leading research groups have demonstrated significant progress, such as AI-guided discovery of functional materials, automated photocatalytic reaction optimization, and self-learning microfluidic systems. This review provides a comprehensive analysis of current achievements and remaining gaps, offering insights for researchers and policymakers in advancing this transformative technology.

In this work, we perform all-atom and coarse-grained dynamics simulations to predict the mechanical properties of a typical synthetic protein system. Previous experiments showed that proteins with a larger molecular weight exhibit better mechanical performance. Our steered molecular dynamics (SMD) simulations at the all-atom level only capture intermolecular interactions and fail to reproduce this tendency. The results of the dissipative particle dynamics (DPD) simulations at the coarse-grained level are consistent with experiments. The comparison between two levels of resolution highlights the importance of simulation scales in predicting mechanical properties of complex systems. We also reveal some underlying factors correlated with the mechanical properties of synthetic proteins, such as molecular weights, fabrication processes, the ratio of hydrophobic to hydrophilic segments and their order in the amino acid sequences.

Iron-sulfur ([FeS]) clusters are among the most ancient and functionally diverse cofactors in biology, mediating processes ranging from electron transfer to substrate activation and catalysis. Within this broad class, non-cubane [FeS] clusters exhibit outstanding structural flexibility and chemical versatility. This review summarizes recent computational advances in elucidating the mechanisms of non-cubane [FeS] cluster-dependent enzymes, focusing on systems containing [2Fe–2S] and distorted [4Fe–4S] clusters. Representative enzymes discussed include biotin synthase (BioB), Rieske dioxygenases (NDO, BphA, NBDO), heterodisulfide reductase (Hdr), and carbon monoxide dehydrogenases (CODHs), that perform diverse chemical transformations such as sulfur insertion, aromatic cis-dihydroxylation, S–S bond cleavage, and CO₂/CO interconversion. Special emphasis is placed on how quantum chemical cluster modelling and QM/MM simulations have provided insights into transient intermediates, rate-determining steps, and the roles of metal nuclearity and heterometal incorporation in tuning catalytic reactivity.

This study employed molecular dynamics (MD) simulations, utilizing both a machine learning force field (MACE-OFF) and a traditional force field (PCFF), to predict the thermal properties of poly(hexamethylene terephthalamide-co-isophthalamide) (PA6T/6I) copolymers. The simulations are benchmarked against experimental data to assess the predictive accuracy of these two methodologies for the thermal properties of PA6T/6I copolymers. Our findings reveal that the MACE-OFF force field, after calibration for the PA6T/6I copolymer, offers significant precision in modeling π-π and hydrogen-bonding interactions, closely mirroring the results from M06 functional simulations. The MD simulations underscore the MACE-OFF model's ability to deliver more stable thermal properties, including the glass transition temperature (Tg) and density, for copolymer systems with varying PA6T content, aligning well with experimental observations. Furthermore, a comprehensive analysis of dynamic properties, such as mean squared displacement and free volume, within the PA6T/6I copolymers was performed to decipher the mechanisms underlying the temperature-dependent changes in thermal properties observed throughout the simulation process. A thorough examination of the fluctuations in inter-chain and intra-chain hydrogen bonding within the copolymer systems has unveiled the correlation between the molecular packing arrangement and thermal properties. This research establishes that the MACE-OFF model accurately simulates the thermal dynamical behavior of PA6T/6I copolymers, a capability that could be extended to other polyamide systems.