Multiscale molecular dynamics simulations were used to examine seven polymer-plasticizer solid-state electrolyte formulations. Formulation films were assembled and dried with Martini 3 coarse-grained models, backmapped to all-atom systems, hydrated with 150 mM NaCl, and simulated under a 1.6 V potential drop. Na⁺ diffusion, accessible free volume, and head-group coordination were associated with measured current density, while plasticizer water exposure and intermolecular organization were associated with cell viability, providing a molecular interpretation of the experimental screening trends.

Lipid nanoparticles (LNPs) are widely used for nucleic acid delivery, but formulation and lipid data remain fragmented across the literature. We developed LNPDB, an integrated database that standardizes LNP formulations, experimental readouts, ionizable lipid structures, and CHARMM force field files for molecular dynamics simulations. LNPDB supports property- and structure-based searches and is continuously expanding with newly reported datasets. Future developments will further integrate AI-based prediction and molecular simulation-derived descriptors to advance data-driven LNP design.

MolCube Peptide Designer is an integrated platform for designing and validating therapeutic peptides. It combines generative sequence design and target–peptide structure prediction with model construction, molecular dynamics simulation, and trajectory analysis. Applied to LC3, the workflow generated cyclic peptide candidates and evaluated their binding stability and energetics within the target pocket. Experimental testing confirmed a lead compound with a Ki of 145 nM. The platform therefore offers researchers an accessible end-to-end route for faster, more reliable peptide discovery.

MolCube-APPs provides a modular, cloud-native platform that streamlines molecular dynamics workflows from structure reading and modeling to system assembly and free energy simulation setup. MolCube-Designer supports structure prediction, peptide and antibody design, docking, ADMET prediction, and QSAR modeling. MolCube-Builder automates force field generation and complex system construction. Specialized modules support enhanced sampling, high-throughput simulation, ML/MM modeling, and alchemical free energy simulations. Integrated simulation execution and analysis through MolCube-Simulator and MolCube-Analyzer enable scalable, reproducible workflows for drug discovery and materials research.

Targeted protein degradation offers an alternative strategy for investigating PARP1 in breast cancer. This study combines structure-guided PROTAC design with exploration of established and underexplored E3 ubiquitin ligases, including cereblon. Molecular docking, binding-energy estimation, and 100-ns molecular dynamics simulations were used to evaluate predicted molecular interactions. Prioritized PROTAC molecules will be synthesized, structurally characterized, and evaluated for PARP1 degradation, selectivity, and cellular activity. This integrated computational and experimental approach aims to identify promising degraders for further investigation in breast cancer models.

Understanding the membrane-disrupting mechanisms of antimicrobial peptides is key to discovering new antibacterials. We conducted Umbrella Sampling and Gaussian Accelerated Molecular Dynamics to map free-energy landscapes of peptide insertion and envelope disintegration using all-atom Gram-negative (including LPS) and Gram-positive (with a peptidoglycan mesh) membranes. To validate our simulations, we synthesized the linear peptide and its hydrocarbon-stapled analogs, evaluating their secondary structures and bactericidal potential. This framework elucidates complex peptide-envelope interactions, providing guiding the design of potent antibacterials. We acknowledge support from National Science Centre, Poland (2023/49/N/NZ1/00989, 2025/57/B/NZ1/03369) and Foundation for Polish Science (START 012/2025).

Biomolecular simulations are powerful tools, but they remain expertise-intensive. We present Bioarchitect, a multi-agent framework that translates natural-language scientific instructions into complete biomolecular simulation workflows. Bioarchitect is evaluated on 14 tasks spanning membrane systems, nucleic acids, coarse-grained modelling and backmapping, and free-energy calculations. Across these tasks, the agents independently prepare, simulate, and analyse each system. Simulation results are assessed against established physical behaviour and published reference data

We present λ-walking, a reactive molecular dynamics framework that enables continuous charge transfer using the local electrostatic energy difference between proton donor and acceptor states. The method describes proton hopping associated with O–H bond breaking and formation while retaining an efficient empirical force field description. After validation in bulk water, λ-walking captures slower proton hopping at various aqueous interfaces. We further extend the framework to nonaqueous electrolytes, demonstrating its applicability to reactive charge transport across diverse solvation environments.

Molecular dynamics force fields are widely used to model sharply bent DNA, but their accuracy for sequence-dependent twist–bend coupling remains unclear. We evaluate three Amber-based force fields (parmbsc1, OL21, OL24) combined with TIP3P and OPC water in 92 bp DNA minicircles against PDB-derived nucleosome references. All six combinations underestimate coupling, though relative sequence trends are broadly reproduced, with CG/CG, GC/GC, and AG/CT as notable exceptions. OPC water substantially improves OL24, an effect linked to suppression of Type I kinks that otherwise distort minicircle geometry under TIP3P water.

Mitochondrial fission requires recruitment of dynamin-related protein 1 (DRP1) from the cytosol to the outer mitochondrial membrane (OMM), but how its receptor, mitochondrial fission factor (Mff), acts with the membrane though its transmembrane domain (TMD) remains unclear. We used all-atom MD to discover how the Mff transmembrane domain (TMD) engages mixed-lipid bilayers with or without cardiolipins (CL). The TMD remains stably inserted in all systems, showing only mild tilt and axial fluctuations. CL does not measurably alter bilayer thickness or integrity. Anchoring is supported by hydrogen bonds between R341/R342 and CL headgroups, with W337 further stabilizing the TMD. These findings reveal a robust insertion mode and pinpoint residues that likely enable Mff-driven DRP1 recruitment during mitochondrial fission.

Electrochemical doping at soft interfaces arises from concerted ion uptake, electronic charge formation, and polymer reorganization, but is difficult to represent in conventional molecular simulations. We develop an iterative framework that couples field-driven molecular dynamics with ion-exchange and charge-formation updates until steady state is reached. Applied to aqueous conjugated polymer–electrolyte interfaces, the simulations reveal how ion hydration influences the coupled doping process. This approach provides an atomistic model of coupled ionic and electronic processes in soft bioelectronic materials.

Fluorinated solvents have emerged as promising electrolyte components due to their oxidative stability and tunable solvation environments. Here, we develop ab-initio polarizable force fields to describe intermolecular interactions in fluorinated electrolyte systems. The resulting force field s are used to analyze Li⁺ solvation structures and ionic distributions in fluorinated electrolyte systems. These simulations provide molecular-level insight into ion motion in fluorinated solvents and establish a framework for understanding how fluorination influences electrolyte structure and transport.

We present a physics-based force field for water and aqueous ions combining a SAPT-based pairwise baseline with a Δ-machine-learning correction for many-body interactions. Trained only on CCSD(T) dimer and trimer interaction energies, the model reproduces water hexamer energies and bulk liquid properties, and extends naturally to ionic solutions. By capturing many-body effects at a fraction of the computational cost of potentials such as MB-pol, our model enables accurate molecular dynamics simulations at the nanometer scale in aqueous and ion-containing environments.

Beyond static structure prediction, understanding conformational changes between metastable states requires continuous, chemically meaningful transition pathways. Using G-quadruplex DNA as a challenging system, we use metadynamics and clustering to identify representative intermediates, then connect them with TPS-DPS, a CV-free generative path-sampling framework. TPS-DPS training incorporates a hydrogen-bond-guided loss, while generated pathways are refined by the string method. Together, these elements extend the framework beyond static structures toward physically grounded, time-ordered transition mechanisms and suggest a transferable strategy for diverse biomolecules.

Lipid flip-flop free-energy calculations commonly use the membrane-normal position as a reaction coordinate, but slow orthogonal degrees of freedom can remain poorly sampled near the bilayer center. In PSM translocation simulations, upper- and lower-derived configurations remained segregated despite replica exchange, with restrained simulations also showing persistent orientation differences. To improve sampling in this central region, we combine positional umbrella restraints with metadynamics along hidden collective variables, enabling more robust transition sampling and reliable lipid flip-flop free-energy profiles.

Water is essential for biological function, yet in living systems it often experiences anisotropic environments created by interfaces and nanoscale confinement. Using MD simulations, we show that confinement reduces the dielectric constant as anisotropic positive dipolar correlations suppress collective polarization along the confined direction. The collective character also emerges in lateral transport, where correlated displacements generate cooperative motion among neighboring water molecules. Together, orientational and translational correlations provide a molecular picture of hydration and transport in heterogeneous biological systems.

Rechargeable lithium-ion battery performance, including cyclability and thermal stability, is strongly governed by local solvation environments and atomistic ion–ion and ion–solvent interactions within electrolytes. Molecular dynamics simulations provide molecular-level insight into ion coordination, clustering, and correlated motion that is difficult to access experimentally. Here, we developed a transferable ab initio polarizable force field for lithium-containing ionic-liquid and organic-solvent electrolytes. The model accurately captures lithium-ion clustering and dynamical properties across diverse electrolyte environments, highlighting the importance of polarization for reliable electrolyte simulations.

Supramolecular stickers drive polymer network assembly through physical crosslinking, yet their influence on network topology has not been extensively investigated. We use coarse-grained molecular dynamics simulations to investigate how sticker chirality affects network topology and mechanical properties. In homochiral networks, helical stacking imposes a unidirectional rotational bias on polymer chains, promoting entanglements. These networks thus exhibit higher entanglement density and toughness than racemic networks. These results highlight sticker chirality as a design parameter for tough supramolecular polymer networks.

Poly(alkylacrylic acid)s, including poly(acrylic acid) (PAA) and poly(methacrylic acid) (PMA), are ionizable polymers used in diverse applications. To elucidate their pH-dependent conformational behavior, we performed explicit-solvent MD simulations across a range of ionization levels under good and poor solvent conditions. PAA typically shows an approximately linear increase in apparent pKₐ with α′, whereas PMA displays a plateau near α′ ≈ 0.2, suggesting resistance to initial ionization and distinct underlying molecular mechanisms.

Azobenzene photoisomerization involves electronic excitation, which classical molecular dynamics (MD) simulations cannot describe explicitly. We therefore applied torsional biasing to mimic the photoisomerization-induced conformational changes and examine their effects on polymer morphology and chain relaxation. For polymers containing azobenzene units in the backbone, we induced trans-to-cis conversion by gradually modifying azobenzene’s C-N=N-C torsional potential. The resulting cis configurations exhibited looser packing and reduced parallel alignment of aromatic rings, accompanied by faster chain relaxation than their trans counterparts.

Peptide–graphene biosensors use probe peptides to recognize odorants. However, the molecular basis of their selectivity remains unclear. In this study, replica exchange umbrella sampling was performed to compare peptide–odorant binding free energies in bulk water. Peptide organization on graphene was also examined at different surface densities. Binding affinities differed among pairs, while peptide organization and the local water environment varied with surface density. These analyses characterize molecular interactions and surface properties relevant to odor sensing.

Potassium ion channels are membrane proteins that selectively conduct K⁺ ions across cell membranes. Recent single-molecule measurements have revealed that membrane thickness can modulate ion channel conductance. Here, we used enhanced sampling molecular dynamics simulations to characterize the free energy landscape of a potassium ion channel in lipid membranes of different thicknesses. Our results show that thicker membranes stabilize the closed conformation, whereas thinner membranes favor the open conformation. These results suggest that membrane thickness regulates ion channel function by altering its conformational stability.

VacA, a major Helicobacter pylori virulence factor, forms an anion-selective hexameric membrane channel whose atomic structure remains unresolved. We developed a hierarchical modeling workflow combining AlphaFold-derived helices, symmetry, packing constraints and MD simulations. The resulting channel model is structurally stable, water-permeable, and reproduces experimentally observed mutational effects and chloride-over-sodium selectivity. Simulations further predicted proton conduction, which was supported by liposome-based proton transport assays. This work demonstrates the strength of an integrated, simulation-driven strategy for modeling membrane-inserted pore-forming toxins.

Viruses continuously evolve through mutations that can alter viral fitness and antigenicity. Here, we develop a computational framework to explore protein sequence space through conformational dynamics, using the SARS-CoV-2 RBD/ACE2 complex as a case study. Molecular dynamics simulations and sequence design were integrated to characterize variant-specific conformational changes and identify sequence preferences compatible with different structural states. Comparison with Deep Mutational Scanning (DMS) and GISAID data provides validation of the predicted sequence space. This framework may enable prediction of feasible evolutionary mutations and inform future antigen design.

Hyperactive antifreeze proteins bind several ice planes, but how adsorption interacts with plane-specific growth kinetics remains unclear. Using molecular dynamics of Tenebrio molitor antifreeze protein on basal and primary prism ice, we compare protein-free and protein-bound growth fronts. The intrinsic growth-rate anisotropy between the planes is strongly reduced on the protein-bound face, where the prism front develops larger curvature than the basal front. We test whether curvature and pinning account for this convergence, aiming at quantitative design criteria for ice-binding materials.

The EGFR kinase domain's C-terminal tail docks onto the kinase domain surface, but the range of kinase conformations this contact co-occurs with is unclear. We use multiscale molecular dynamics (Martini 3+Gō, all-atom CHARMM36m, CALVADOS3) and unsupervised clustering on salt-bridge, R-spine, activation-loop, and αC-helix markers to map the full conformational landscape — active-like, inactive-like, and intermediate states — sampled alongside C-tail-kinase domain contact. We identify force-field-specific limitations in this sampling, and outline a framework for extending this classification to C-tail phosphorylation and oncogenic mutation variants.

In multicomponent membranes, liquid-ordered and liquid-disordered domains organize through competition between line energy and curvature elasticity; how this balance shifts at the nanoscale remains unclear. Using coarse-grained molecular dynamics of 30 nm DIPC/DPPC/CHOL vesicles and Helfrich analysis, we show this balance inverts at the nanoscale: liquid-ordered/liquid-disordered demixing drives solvent efflux and relaxation into an oblate shape, with a continuous liquid-disordered equatorial band and liquid-ordered polar caps. Curvature elasticity favors this band over a two-domain state below ~65 nm, establishing a size-dependent principle for nanoscale membrane organization.

Understanding how metal binding regulates enzyme function requires connecting conformational dynamics with bioactivity. We combine molecular dynamics simulations, RMSD-based conformational scoring, and enzyme assays to investigate calcium activation of a fungal decarboxylase PcncAAAD using a series of simulation-guided mutants. Calcium binding at site A stabilizes a lid–rim structure, preserving the substrate-binding environment, whereas binding at site B stabilizes the C-terminal domain. Simulation-guided mutagenesis and experimental validation link structural stability to catalytic activity, providing mechanistic insights for engineering metal-activated enzymes.

Molecular dynamics simulations of aquaporin-4 (AQP4) embedded in a lipid membrane were performed to investigate temporal fluctuations in pore-water motion. We analyzed time series of the collective mean displacement of water molecules within the pore using power spectral density. Low-frequency enhancement was observed in a reference system without an ion concentration difference across the membrane. We discuss the characteristics of these fluctuations and consider how transmembrane ion concentration differences may affect pore-water dynamics and asymmetric water transport.

Magnetosomes are membrane-bound magnetic organelles in magnetotactic bacteria and contain a variety of membrane-associated proteins. Although several of these proteins are predicted to contain transmembrane regions, their structures and stability within lipid bilayers remain poorly understood. Here, we investigated the stability of the putative transmembrane region of MamJ as a single helix in a lipid bilayer using molecular dynamics (MD) simulations. Replica Exchange with Solute Tempering 2 (REST2) enabled extensive conformational sampling and showed that the region stably adopts a transmembrane helical structure. Moreover, this approach revealed multiple helix conformations with distinct bending patterns, indicating conformational flexibility within the membrane.

The Ups1–Mdm35 complex transports phosphatidic acid between mitochondrial membranes, supporting cardiolipin synthesis. Using all-atom molecular dynamics simulations, we identified three Ups1 membrane-binding modes and propose distinct roles for T2 in productive membrane recognition and T1 in lipid delivery. Extended simulations capture complete lipid release, with pocket opening involving the ω loop and α3 helix and membrane insertion preceding final dissociation. Targeted mutagenesis and lipid-transfer assays support these functional assignments. Our findings reveal how membrane orientation and conformational dynamics coordinate successive steps of mitochondrial lipid transport.

Small-molecule coacervates are emerging as programmable delivery platforms, yet the molecular mechanisms underlying condensate formation, cargo encapsulation, stimulus-triggered release, and membrane entry remain poorly understood. Here we use the photoresponsive phase-separating fluorescent molecule (PPFM) as a model system and conduct all-atom simulations to resolve their coacervation-to-delivery process at molecular scale. We show that PPFM condensates are heterogeneous assemblies with water-rich internal cavities instead of homogeneous liquid droplets. This nanoscale architecture enables volumetric loading of the cargo KWK peptide (KRWWKWWRR), which penetrates the PPFM-rich interior through hydrated domains while maintaining non-specific hydrophobic contacts with PPFM. Upon conversion to their photolysis products pyrene ketone (PYKE), the cavity-filled hydrated PPFM architecture collapses into more compact, water-poor PYKE-rich structures, leading to cargo expulsion and providing a molecular mechanism for light-triggered cargo release. Simulations with lipid bilayers further reveal spontaneous membrane insertion of cohesive PPFM-rich assemblies, accompanied by local lipid reorganization. Consistently, giant unilamellar vesicle experiments reveal rapid passive transbilayer passage of PPFM coacervates and show that KWK-loaded assemblies can cross the membrane while remaining associated. Together, these results identify hydrated nanoscale heterogeneity, photoinduced structural collapse and cohesive membrane partitioning as key molecular features underlying PPFM-mediated cargo delivery, providing a comprehensive molecular foundation for engineering photoresponsive small-molecule coacervates as programmable delivery platforms.

Throughout evolution, organisms have adapted to both hotter and cooler environments, and these adaptations are reflected in their protein sequences, allowing enzymes to function under different thermal conditions. Adenylate kinase (AK), a ubiquitous enzyme found in nearly all forms of life, plays a crucial role in cellular energy homeostasis by catalyzing the reversible conversion of ATP and AMP into two ADP molecules. Interestingly, within the ancient bacterial phylum Firmicutes lineage, looking at approximately 3 billion years of evolutionary history, key active-site residues surrounding the nucleotide substrates remain highly conserved despite extensive sequence variations associated with thermal adaptation. This raises an intriguing question: how has evolution influenced enzyme catalysis through distal sequence variations while largely preserving the residues directly involved in the chemical reaction? In this study, we employ nanosecond-scale molecular dynamics simulations of AK homologs from several species adapted to different temperatures. Starting from catalytically relevant conformations, we investigate how thermal adaptation influences large-scale conformational fluctuations and their connection to the local reaction geometry. Our observations reveal an intriguing relationship between ATP-lid domain conformational dynamics in AK and the alignment of reactive groups involved in phosphoryl transfer, with distinct conformational preferences across the investigated systems at different temperatures. These findings suggest that thermo-adaptive distal sequence variations could influence enzyme catalysis by modulating conformational fluctuations and promoting the sampling of favorable reaction geometries at their respective operating temperatures.

(Yichao Wu, Yangyang Zhang, Ai Shinobu)Protein kinases recognize their substrates through specific short peptide motifs, and this specificity has motivated broad interest in developing peptide-based kinase inhibitors. Traditional studies have focused mainly on this specific, bound-pose interaction, whereas the nonspecific interactions that precede specific binding have received comparatively little attention. Such nonspecific interactions are transient, making them difficult to capture using either experimental techniques or conventional molecular dynamics simulations, and consequently their biological significance and functional role remain poorly understood. Here, we used two-dimensional replica exchange (gREST/REUS) to study the binding process between Abl kinase and its optimal synthetic substrate, Abltide. The simulations revealed that, in addition to the specific bound pose, the surface of Abl kinase harbors multiple nonspecific binding sites that constitute distinct encounter states, mediated primarily by electrostatic interactions with additional aromatic and polar contributions. Among them, the encounter states located adjacent to the binding pocket direct Abltide toward it, consistent with proposed mechanisms by which nonspecific binding facilitates substrate recognition. As Abltide approaches the binding pocket, the αEF/αG/β11 patch plays a central role in stabilizing the resulting intermediate states, primarily through hydrophobic interactions with additional electrostatic and polar contributions. Within the substrate-binding site, the Abltide backbone also shifts progressively out of the register of the specific bound pose, forming backbone hydrogen bonds with the kinase that are absent from the bound pose. These findings indicate that substrate recognition by Abl kinase is encoded not only at the specific binding site but also across the surrounding kinase surface, and that peptide-based inhibitor design should consider both the specific and nonspecific binding behavior of the peptide.

Authors: Yangyang ZHANG, Yichao WU, Ai SHINOBU. Abstract: The BCR-ABL fusion causes constitutive overactivation of Abl kinase and drives chronic myeloid leukemia. Peptide binders offer a selective alternative to small-molecule inhibitors. Over the years, experiments have screened a series of linear peptides, producing abltide and several optimized sequences together with a ranking of their selectivity. We performed FEP MD simulations on abltide variants mutated at the non-conserved positions and calculated their relative binding free energies. Our results follow the experimental trend closely, though the ranking is not fully reproduced. This suggests that binding affinity alone may not be sufficient to guide the design of an optimal competitive binder. Experiments have also found that grafting abltide onto a macrocyclic scaffold improves inhibition. We used a deep-learning structure predictor (Boltz-2) to graft the linear peptides favored by experiment and by our calculations onto such a scaffold, and selected suitable complex structures for future MD simulations. By combining AI-generated structures with MD simulation, we establish a practical pipeline for studying how peptide kinase inhibitors can be optimized in both sequence and scaffold.

Ethylene–EtOP copolymers exhibit autonomous self-healing near equimolar composition, whereas neither homopolymer heals detectably. All-atom molecular dynamics simulations show that increasing ethylene content enhances chain mobility but reduces interchain π–π contacts. The EtOP homopolymer favors intrachain aromatic association. The self-healing 49:51 copolymer retains substantial interchain contacts with fewer intrachain interactions, while the ethylene-richer 32:68 copolymer is more mobile but has fewer interchain contacts. These results suggest self-healing is associated with balancing local mobility and interchain aromatic interactions.

Unlike molecular property prediction, reaction yield prediction is inherently relational, as outcomes depend critically on intermolecular interactions. We propose RelaRink, which represents each reaction as a hierarchical graph over all combinatorial molecular pairs. A source–sink pre-training strategy learns electron-flow directionality, which fine-tuning leverages to reinterpret atomic representations within pairwise interactions. On an amide coupling dataset, RelaRink achieves an MAE of ~8%, outperforming other approaches. Ablations confirm that this pre-training contributes beyond parameter initialization and is central to capturing intermolecular interactions.

AI-based protein design can rapidly generate thousands of candidate binders, but most fail to fold or bind productively, creating a critical need for robust prioritization. We present a generalizable hybrid pipeline that integrates deep-learning design and MELD physics-based simulations to filter large libraries down to a handful of high-confidence candidates. As a case study, we apply this approach to the BRD3 ET domain of BET proteins, an important but difficult-to-target interaction hub in gene regulation.

Solvent effects on reaction barriers arise from solvation differences between reactants and transition state (TS). We examined how reaction representation, geometry fidelity, and modality affect prediction of activation solvation free energy. Adding the TS geometry improved accuracy over the 2D baseline (test R² 0.70 to 0.90), whereas reactant–product geometries did not. With low-cost TS geometries at inference, accuracy fell below the 2D baseline in distribution but exceeded it on two external reactions, where reactant–TS geometries alone outperformed the dual-modality model.