Drug-drug interactions (DDIs) are a major cause of adverse drug events in clinical practice, especially under polypharmacy settings where patients receive multiple medications simultaneously. Reliable computational prediction of DDIs is therefore essential for improving medication safety and supporting clinical decision-making. Despite recent advances in computational DDI prediction, existing methods often struggle to jointly model multi-granularity pharmacological semantics and stereochemical molecular characteristics, limiting their ability to generalize to previously unseen drugs under cold-start scenarios. To address these limitations, we propose DSMV-DDI, a multimodal representation learning framework for drug-drug interaction prediction that integrates biomedical knowledge graph topology, chemical substructure features, dual-level pharmacological semantic representations, and stereochemical molecular visual representations derived from three-dimensional molecular conformations. In particular, the proposed dual-level semantic strategy jointly characterizes interaction-level pharmacological associations and intrinsic single-drug functional properties, enabling complementary modeling of pharmacological information across different semantic granularities. Furthermore, molecular visual representation learning captures geometric and spatial characteristics beyond topology-based molecular representations, improving generalization to topologically unseen drugs. Extensive experiments on real-world DDI datasets demonstrate that DSMV-DDI outperforms state-of-the-art methods, achieving an accuracy of 0.967 and an AUPR of 0.992 under the conventional setting. The proposed framework also maintains strong performance under both partial and complete cold-start settings. Ablation analyses show that dual-level pharmacological semantics contribute most to overall performance, while molecular visual representations provide complementary geometric information that further improves prediction accuracy.

Second-neighbor interactions play a fundamental role in determining the higher-order organization of molecular graphs, yet they are not explicitly represented by conventional degree, distance or adjacency-based graph descriptors. This work introduces the Leap Architecture Matrix (LAM), a novel matrix representation that characterizes the distribution of graph edges between distinct leap-degree classes, thereby preserving the structural organization of second-neighbor connectivity within molecular networks. Based on this representation, three complementary descriptors are developed: the Transition Occupancy Ratio (TOR) which measures the occupancy of leap-transition classes; the Architecture Energy (AE) defined as the sum of the absolute eigenvalues of the Leap Architecture Matrix and the corresponding Spectral Radius (ρ) which quantifies the dominant architectural connectivity. The proposed framework is applied to a dataset of biologically important regulatory amino acids exhibiting diverse structural and physicochemical characteristics. Its discriminative capability is evaluated through comparisons with established molecular descriptors including the Wiener, Randic and first Zagreb indices, together with Pearson correlation analysis, principal component analysis and sensitivity analysis. The results demonstrate that the proposed descriptors capture complementary structural information associated with higher-order molecular organization while maintaining low computational complexity. In particular, the spectral descriptors derived from the Leap Architecture Matrix effectively distinguish subtle variations in molecular architecture that are not fully characterized by conventional topological indices. The proposed matrix framework provides an interpretable and computationally efficient approach for molecular structural characterization and offers a new family of graph-theoretical descriptors with potential applications in molecular similarity analysis, structural classification, spectral graph theory, QSAR/QSPR modeling and computational molecular informatics.

The hexafluoroisopropyl alcohol (HFIP) group is a strong hydrogen-bond donor that governs interaction geometry in anion complexes. In this study, three model systems with different hydrogen-bond donor acidities-HFIP-substituted ethylbenzene, ethylphenol, and ethylbenzene-were investigated using density functional theory (DFT). Rather than relying on energy-based ranking, the analysis focuses on geometric formation, anchor-induced constraints, and geometric fluctuation. In the HFIP-substituted ethylbenzene system, a strong O-H···O hydrogen-bond anchor is formed, leading to convergence toward an anchor-dominated interaction geometry. The secondary C-H···O contacts exhibit only limited geometric variability around the anchor-dominated geometry. In the ethylphenol system, the anchor is retained but weaker, allowing reproducible secondary C-H···O interactions and limited structural variability. In contrast, in the ethylbenzene system, no primary anchor is formed, and multiple weak interactions produce dispersed interaction geometries without convergence toward a dominant structural regime. These results demonstrate that donor acidity governs the strength of anchor constraints and geometric tolerance, producing distinct interaction geometry regimes separated by geometric boundary regions.

The growing demand for advanced optoelectronic technologies has intensified the search for efficient nonlinear optical (NLO) materials for photonic and telecommunication applications. Herein, we employed a density functional theory (DFT)-based computational framework to design and evaluate coinage-metal-functionalized P-flower complexes, M@CPF (M = Cu, Ag, Au), based on the π-conjugated C16(PF)8 scaffold. The incorporation of coinage metals significantly modifies the electronic structure of CPF by promoting metal-to-framework charge transfer, reducing excitation energies, and enhancing electronic polarization. All M@CPF complexes exhibit improved polarizability and hyperpolarizability compared with pristine CPF. Among them, Ag@CPF shows the highest static first hyperpolarizability (β0) value of 15935.63 a.u., which is 17.76 times larger than that of pristine CPF (897.31 a.u.), followed by Au@CPF and Cu@CPF. Frequency-dependent NLO analysis further reveals a remarkable dynamic response, particularly for Au@CPF, which exhibits exceptionally high first hyperpolarizability values at 1064 nm under SHG and EOPE processes. In addition, the third-order NLO response is enhanced by up to 1.68 × 105 times compared with pristine CPF. UV-vis-NIR analysis confirms significant bathochromic shifts and enhanced charge-transfer transitions in the metal-functionalized complexes. These findings demonstrate that coinage-metal-functionalized P-flower architectures are promising candidates for next-generation NLO and optoelectronic applications.

The current study aims at exploring structural, electronic, and thermodynamic properties of complex metal hydrides NaXH6 (X = P, As, Sb) with DFT (Density Functional Theory) as the main calculation software. The tolerance factor with volume optimization has been used to attain structural integrity. The lattice parameters of NaPH6, NaAsH6 and NaSbH6 are 7.47, 8.03 and 8.89 Å, respectively. Among the three compounds, NaPH6 has the highest values for both volumetric hydrogen capacity (96.3 g H2/L) and gravimetric hydrogen storage (10.1%). The desorption temperature is observed to be 422 K, 199 K, and 347 K for NaPH6, NaAsH6 and NaSbH6, respectively. The computed density of states and band structures (DOS/PDOS) indicate that the entire three hydrides are of semiconducting nature with definite energy separations among the valence bands and the conduction bands. The series of band gap reductions observed is NaPH6 > NaAsH6 > NaSbH6. At room temperature, NaPH6 shows the highest value of electrical and thermal conductivity. AIMD simulations have been extensively used to verify the thermal stability. Elastic properties show that NaPH6 has the highest length flexibility, and shape elasticity also shows more resistance to change in volume and has the strongest resistance against deformation among the three compounds. Negative values of Cauchy pressure show covalent bonding in these compounds, while the B/G ratio indicates that all three compounds are brittle. Furthermore, the calculated Debye temperature (ѲD) progressively decreases when passing from NaPH6 to NaSbH6, indicating weaker lattice vibrations and lower phonon frequencies in heavier counterparts. These outcomes offer significant insights for designing and optimizing the performance of hydrogen storage materials, leading to the advancement of hydrogen storage technologies.

The main challenge in extracting and separating rare earth elements from industrial waste is their chemical similarity. Thus, this study focuses on the solvent extraction and separation of lanthanum and cerium using diethyl-2-hexylphosphoric acid (D2EHPA). Molecular Dynamics (MD) simulations and Density Functional Theory (DFT) were employed to determine the stoichiometry and the stability of the complexes formed. The results show that the extraction percentage efficiency and the separation factor Dce/DLa increased in the pH range of 1-5. FTIR studies of loaded D2EHPA organic solutions indicate cation-exchange behavior through the formation of P-O-La and P-O-Ce bonds. The analysis of logD vs. pHeq and logD-2pH vs. log [D2EHPA] determined that the extracted complex stoichiometries were [La(NO3) (D2EHPA)2] and [Ce(NO)3 (D2EHPA)2]. The radial distribution function (RDF) and coordination number of La-H2O and La-NO3[-]/Ce-H2O and Ce-NO3[-] gave 8 and 1, respectively, for the number of water molecules and nitrate ions around La[3+] and Ce[3+]. Optimizing the geometry through the density functional theory (DFT) approach on the optimized structures of the [La(NO3)(H2O)8][2+] and [Ce(NO3)(H2O)8][2+] ionic complexes illustrated the monodentate form of nitrate ions around the central atom in the first coordination sphere. However, bidentate structures were found for D2EHPA in the extracted complexes. The Born Haber thermodynamic cycle and the Conductor-like Polarizable Continuum Model (CPCM) were applied to calculate the free energies ΔGext of lanthanum (-179.02 kcal/mol) and cerium (-358.61 kcal/mol). The energy gap of HOMO-LUMO between [La(NO3)(D2EPA)2] = 0.17 eV) and [Ce(NO3)(D2EPA)2] = 0.03 eV), the Mayer Population Analysis (MPA) and Wiberg Bound Index (WBI) corroborated the stability of the D2EHPA complex with cerium over that with lanthanum. These results could help us to understand how organic extractants interact with rare earth elements and radionuclides. They could also have applications in nuclear technology and the sustainable remediation of polluted environments.

Coronaviruses, including SARS-CoV (2003), MERS-CoV (2012), and SARS-CoV-2 (2019), have caused major global outbreaks, highlighting the importance of structural studies of conserved viral proteins for antiviral drug discovery. Therefore, the structural characterization of coronavirus proteins is essential for understanding their function and supporting the development of new therapeutic strategies. In present work, a structural analysis of the MERS-CoV nsp5 protein was done in detail. Specifically, here for the first time all the MERS-CoV nsp5 unique sequences were selected, then used in homology modeling and molecular dynamics methods to identify all possible binding pockets that could potentially be exploited in further drug discovery projects. Despite high overall sequence similarity, significant differences in conformational dynamics were observed, indicating that minor sequence variations can induce local structural rearrangements. This flexibility, particularly in loop regions, contributes to the formation and modulation of binding pockets. In total, seven regions with ligand-binding potential were identified. Most of the identified sites, except for the canonical catalytic site, represent previously uncharacterized binding regions distributed across the protein. Among these, pocket 3 remained highly stable throughout simulations, while pockets 4 and 5 exhibited potential cryptic behavior, showing moderate expansion into surrounding protein regions under dynamic simulation. Suggesting their binding potential to bind small ligands upon conformational rearrangement. These findings provide a comprehensive structural framework for MERS-CoV nsp5 and highlight several promising binding sites for future structure-based antiviral drug design.

Ferrochelatase (FECH), the terminal enzyme in the heme biosynthetic pathway, has emerged as a compelling therapeutic target for enhancing photodynamic therapy (PDT) and suppressing angiogenesis. Consequently, the design of FECH inhibitors represents a highly promising strategy for discovering dual-modality therapeutics, potentially offering a foundation to simultaneously explore phototoxicity and anti-angiogenic mechanisms. In this study, we deployed a rigorous and multi-tiered computational pipeline to identify potent heterocyclic FECH inhibitors from a compound library. The screening protocol integrated deep learning (DL)-assisted molecular docking and a dual-algorithm machine learning (ML) consensus classifier to systematically eliminate false positives. The orthogonal artificial intelligence filters isolated two structurally distinct and highly viable candidates: AZD-8055 and hSMG-1 inhibitor 11e. Subsequent 200 ns molecular dynamics simulations and Molecular Mechanics/Poisson-Boltzmann Surface Area (MM/PBSA) thermodynamic evaluations reveal that AZD-8055 demonstrated a stabilizing effect on the structure, exhibiting the most stable properties and favorable binding free energy. Conversely, hSMG-1 inhibitor 11e achieved the most potent binding free energy. Ultimately, this study demonstrates the immense utility of constructing an integrated, AI-driven structure- and ligand-based workflow to explore FECH inhibitors. By successfully identifying AZD-8055 and hSMG-1 inhibitor 11e as potential FECH inhibitors in silico, this approach provides a strong theoretical foundation for the development of dual-modality photodynamic and anti-angiogenic therapeutics.

The advancement of CCR5 antagonists is a vital approach in HIV-1 treatment to mitigate drug resistance and the toxicities of long-term antiretroviral therapy. This work utilized a combined ligand and receptor-based virtual screening workflow to discover novel Maraviroc derivatives. Initial similarity search of the ZINC in-stock database using ChemFP yielded 18 candidates, which were analyzed by molecular docking against the CCR5 receptor (PDB: 4MBS). Eight molecules displayed predicted docking scores more favorable than Maraviroc (-8.84 kcal/mol), led by ZINC001542979606 at -9.74 kcal/mol, although these differences fall within the inherent uncertainty of the scoring function. Multi-platform in silico ADMET analysis and molecular dynamics simulations revealed that the identified subset of analogues exhibits distinct trade-offs between computationally predicted safety profiles and thermodynamic binding stability. To assess temporal stability, molecular dynamics simulations were executed in a hydrated POPC lipid bilayer in triplicate (200 ns per replica), and MM-PBSA and LIE analyses were used to estimate binding free energies. These simulations demonstrated that while some analogues show improved predicted toxicological profiles, they often incur significant thermodynamic penalties or higher structural drift than Maraviroc. These results highlight ZINC001542979606 and ZINC000030691387 as theoretical leads that require in vitro validation to navigate the complex trade-offs involved in developing safer HIV-1 entry inhibitors.

Transglutaminase 2 (TG2) serves as a modifiable transamidating acyltransferase that precipitates calcium-induced protein alterations. The enzyme plays a crucial role in the cell and disease states, such as tissue repair, calcium signal transduction, celiac disease, and cancer. It is implicated in protein crosslinking and has been found in high concentrations in the small intestines of those with celiac disease. The function of TG2 hinges upon calcium ions binding to particular sites on the enzyme. In this study, we delve into the contribution of calcium-responsive transglutaminase 2 (TG2) in celiac disease, utilizing both molecular dynamics simulations and coarse-grained models, and investigate the impact of non-synonymous single nucleotide polymorphisms (nsSNPs) on TG2. Molecular dynamics reveal prominent conformational differences between the open and closed conformations. In the coarse-grained model, key residues are found adjacent to the active site in the open conformation, while in the closed conformation, key residues are distant from the active site. We further explore the functional impact of nsSNPs in TG2 using both sequence-based and structure-based computational tools. Through a consensus approach, we identify ten nsSNPs that are predicted to destabilize TG2 or alter its structural flexibility, with mutations such as R48H, E186Q, C277S, and E549G likely to influence active site accessibility and calcium coordination. The findings from this research enhance our understanding of the molecular processes underpinning celiac disease and helps facilitate innovative treatment approaches that target calcium-responsive TG2.

Dantrolene is a clinically approved therapeutic used to suppress abnormal calcium release associated with malignant hyperthermia linked to RyR1 mutations. Cryo-electron microscopy studies have revealed a putative dantrolene-binding cavity within the RyR1 P1 domain and highlighted several residues associated with ligand recognition. However, the dynamic molecular interactions governing ligand stabilization within this region remain incompletely understood. MM/GBSA calculations, conformational entropy analyses, and potential of mean force (PMF) profiling were employed to investigate the local binding dynamics of dantrolene within isolated RyR1 P1-domain complexes. Comparative analyses of the P1 domain from open- and closed-state conformations of RyR1 revealed that the closed-state complex provides a more stable binding environment characterized by lower backbone fluctuations, enhanced ligand-protein contacts, stabilization of a distal loop region (residues 1006-1028), and higher ligand dissociation barriers. Per-residue energetic decomposition and interaction analyses identified W882 and W996 as major contributors to ligand stabilization through cooperative hydrogen bonding and π-mediated interactions. Alanine substitution of these residues increased ligand conformational flexibility and substantially reduced the energetic barriers associated with ligand dissociation, whereas R1000 exhibited comparatively smaller contributions to overall binding stability. Overall, these computational findings are consistent with previous experimental observations and provide atomistic insight into the interaction networks and structural determinants underlying state-dependent stabilization of dantrolene in a cryo-EM-derived RyR1 P1-domain binding model.

In recent years, in the treatment of non-small-cell lung cancer (NSCLC), epidermal growth factor receptor (EGFR) inhibitors have demonstrated ideal clinical efficacy. Unfortunately, a significant obstacle to targeted lung cancer therapy is the unavoidable emergence of acquired resistance to EGFR inhibitors through a variety of pathways during a period of medication. The third-generation EGFR inhibitor Lazertinib, which is potent, irreversible, brain-penetrant, mutant-selective, and wild type-sparing, was used to treat patients with advanced or metastatic NSCLC. Lazertinib can bind to EGFR[T790M] in different conformations, identified by a 180° rotation of the pyrazole moiety, according to the X-ray co-crystal structure. A molecular modeling study integrating molecular dynamics and free energy calculation was conducted to comprehend the distinct binding manner of Lazertinib binding to EGFR and the structural need for the inhibitory activity. According to binding free energy calculations, Lazertinib has a greater binding affinity with EGFR[T790M] than EGFR[WT], which is in accordance with the experimental observations. Additionally, it confirms that Lazertinib preferentially binds to EGFR[T790M] with the same conformation as in EGFR[WT]. The residues that made a greater contribution to the binding of Lazertinib to EGFR were identified using the per-residue energy decomposition. It is anticipated that these findings will be helpful to the rational development of new EGFR inhibitors.

4-HNE carbonylation and phosphorylation in the Zu5-ANK domain have been observed in human erythrocytes with sickle cell trait and G6PD deficiency. However, the atomic-level effects of phosphorylation and multiple carbonylations by 4-HNE on the Zu5-ANK domain are not yet fully understood. The present work provides further insight into understanding the changes caused by polymodifications in Zu5-ANK due to oxidative stress and phosphorylation. Here, we present a molecular dynamics simulation-based study of the polycarbonylated, phosphorylated, and phosphocarbonylated Zu5-ANK domain. Our results show that these modifications induce local conformational and mobility changes in the protein, particularly within three regions, which include specific residues that interact with the β-spectrin protein. Furthermore, our analyses revealed that while the domain exhibits remarkable global structural resilience to isolated modifications, the combination of phosphorylation and polycarbonylation exerts a synergistic destabilizing effect. This polymodified state alters the spatial geometry of the basic residue patch essential for β-spectrin binding. Subsequent protein-protein docking with β-spectrin repeats 14 and 15 demonstrated that although isolated carbonylation is structurally tolerated at the interface, the simultaneous presence of both modifications disrupts the canonical electrostatic complementarity. This perturbation leads to a reduced buried surface area, diminished interaction specificity, and a highly heterogeneous binding mode. Overall, these in silico findings propose a theoretical molecular rationale for how the accumulation of multiple post-translational modifications might weaken the ankyrin-spectrin anchorage, providing a mechanistic basis for future experimental studies on erythrocyte membrane instability under oxidative stress.

Benzothiadiazole derivatives (BTZ) have established themselves as benchmarks in nonlinear optical (NLO) research, owing to their exceptional charge-transfer capabilities and extensive electronic delocalization. With a unique blend of structural versatility and electronic tunability, these compounds hold immense potential for revolutionary modern optical applications. We developed a tailored set of BTZ derivatives, comprising four core-centered (C-1PhTDZ to C-2PhDTDZ) and four terminal-substituted (T-1PhTDZ to T-2PhDTDZ) compounds, by systematically modifying the acceptor unit size and placement to promote extended π-conjugation and enhance NLO responses. Overall, T-2PhDTDZ showed exceptional planarity (molecular planarity measurement; MPP 0.002 Å; span of deviation from plane, SDP 0.006 Å) due to enhanced intramolecular charge transfer (ICT) from terminal BTZ placement, boosting charge separation and performance. C-2PhDTDZ and T-2PhDTDZ exhibit the notable isotropic (αiso) and anisotropic (αaniso) polarizabilities, with values of 365.1 × 10[-24] esu, 93.8 × 10[-24] esu, 879.10 × 10[-24] esu, and 113 × 10[-24] esu, respectively. Within the terminal series, T-2PhDTDZ shows the highest total third-order NLO polarizability (<γ>) of 5128.8 × 10[-36] esu, which is approximately 1.2 to 6.4 times larger than other derivatives. While among core series, C-1PhDTDZ exhibits a remarkably enhanced NLO response, attaining the maximum γ value of 1906.0 × 10[-36] esu. Additionally, frequency-dependent γ was computed, with T-2PhDTDZ exhibiting the largest amplitudes for γ(-ω;ω,0,0) and γ(-2ω;ω,ω,0) at 1400 nm, with values of 247.90 × 10[-36] esu and 324.53 × 10[-36] esu, respectively. Moreover, the density of states (DOS) and natural bond orbital (NBO) analyses provide further insights into the electronic structure and charge transfer characteristics. We compared the Voc values of our designed derivatives with those of PC61BM. Notably, T-2PhDTDZ shows a pronounced offset value of 3.1 eV, which favors rapid electron injection and reduces energy loss. These findings suggest that our designed core-versus terminal-based BTZ derivatives possess significant advancements in NLO and photovoltaic technologies, paving the way for next-generation optoelectronics and photonic applications.

The SPICE dataset is widely used to train and benchmark machine learning force fields, yet the numerical consistency of its underlying density functional theory (DFT) force calculations has received limited systematic evaluation. We present a large-scale analysis of molecular net forces in the B3LYP-D3BJ/DZVP subset of SPICE 1 OpenFF, comprising 16,560 molecules and approximately one million conformations. For isolated systems, fully converged DFT calculations are expected to satisfy translational invariance, yielding near-zero molecular net forces within numerical precision. We observed a mean molecular net force of 0.0001166 Hartree/Bohr (5.99 meV/Å), with 98% of molecules exceeding 0.000038 Hartree/Bohr (1.95 meV/Å). Molecular net force increased with molecular size (r = 0.59) and conformational energy spread (r = 0.61). Multivariable, bootstrap, and robust regression analyses consistently identified molecular size, conformational energy spread, mean atomic force norm, and molecular charge as independent predictors of net force (adjusted R[2] = 0.45). Monte Carlo simulations showed that the observed size dependence is substantially more consistent with stochastic accumulation of small force imbalances than with fully systematic directional bias. Filtering molecules by net force reduced baseline prediction error primarily because smaller molecules were preferentially retained, rather than through improved force consistency after accounting for molecular size. Molecular net force therefore provides a complementary diagnostic of force consistency but should not be interpreted as a direct measure of atomic-force accuracy. The proposed framework is readily applicable to quality assessment of DFT datasets used for machine learning.