Journal of Computer-Aided Molecular Design最新文献

Protein pockets, or small cavities on the protein surface, are critical sites for enzymatic catalysis, molecular recognition, and drug binding. Accurately identifying these pockets is crucial for understanding protein function and designing therapeutic interventions. Traditional computational methods such as molecular docking, surface grid mapping, and molecular dynamics simulations are hampered by the use of fixed protein structures, and therefore it is challenging to identify cryptic pockets when they appear under physiological conditions. We propose a deep reinforcement learning (DRL) technique based on deep Q-networks (DQN) to identify precise protein pockets. Our strategy to improve the prediction of functional binding sites incorporates important molecular descriptors such as spatial coordinates, solvent-accessible surface area (SASA), hydrophobicity, and electrostatic charge. We pre-process protein structure data from the protein data bank (PDB) through feature extraction and selection methods, including variance threshold filtering and dimensionality reduction using an autoencoder. The sparse feature representation enables efficient training of a DQN agent, which navigates protein surfaces and iteratively optimizes pocket predictions. By using reinforcement learning concepts, the model adapts its pocket detection strategy according to the learned reward signals, increasing sensitivity and specificity. The method is tested on benchmark datasets and is found to exhibit superior performance in detecting well-defined and cryptic pockets over traditional computational methods. Experimental evidence suggests that our model successfully identifies binding sites in various protein families, with significant implications for drug discovery and protein-ligand interaction studies. Moreover, the model’s ability to incorporate geometric and biochemical features allows for a better understanding of pocket functionality. The scalability of our method makes it an important tool for large-scale virtual screening and personalized medicine. By using deep reinforcement learning, this research provides a new and effective framework for protein pocket prediction, opening up opportunities for developing new tools in structural bioinformatics, drug design, and molecular biology research.

The traditional drug discovery process is often lengthy, costly, and characterized by a high failure rate. There is a pressing need for innovative strategies to optimize this process and improve the chances of identifying effective therapeutic candidates. This study aims to utilize computational methods to develop a quantitative structure-activity relationship (QSAR) model that predicts the inhibitory activity of compounds against Fibroblast Growth Factor Receptor 1 (FGFR-1), which is associated with various cancers, including lung and breast cancer. The QSAR model was developed using multiple linear regression (MLR) on a dataset of 1779 compounds from the ChEMBL database. The dataset was curated, and molecular descriptors were calculated using Alvadesc software. Feature selection techniques refined the dataset, and the model’s predictive capability was validated through 10-fold cross-validation and external validation with a test set. In silico validation was further performed using molecular docking and molecular dynamics simulations. Additionally, in vitro validation was conducted using MTT, wound healing, and clonogenic assays on A549 (lung cancer), MCF-7 (breast cancer), HEK-293 (normal human embryonic kidney), and VERO (normal African green monkey kidney) cell lines. The QSAR model exhibited strong predictive performance with an R² value of 0.7869 for the training set and 0.7413 for the test set. Molecular docking and dynamics simulations further supported the model’s predictions, demonstrating stable interactions between the compounds and FGFR-1. Experimental validation through the MTT assay revealed a significant correlation between predicted and observed pIC50 values, confirming the model’s accuracy. Oleic acid, identified as the most promising compound, showed substantial inhibitory effects on A549 and MCF-7 cells, with low cytotoxicity observed on normal cell lines. The integration of computational and experimental methods significantly enhanced the efficiency and accuracy of the drug discovery process for FGFR-1 inhibitors.

Graphical abstract

Previous studies of the natural acetylenic acetogenin (2S,3R,4R)-3-hydroxy-4-methyl-2-(eicos-11′-yn-19′-enyl)butanolide (1), isolated from the plant Porcelia macrocarpa, indicated its in vitro activity against the clinically relevant form of Leishmania (L.) infantum, the intracellular amastigotes and no mammalian cytotoxicity. A second chemically related acetogenin, (2S,3R,4R)-3-hydroxy-4-methyl-2-(eicos-11′-ynyl) butanolide (2), exhibited a lack of antileishmanial activity at the highest tested concentration of 150 µM. These results suggest that the terminal double bond plays a crucial role in the antileishmanial activity of these compounds. Using a computational protocol to predict the metabolism of 1, the 19′-oxirane-derivative (3) was proposed, prepared, and experimentally tested against Leishmania (L.) infantum amastigotes. Compound 3 presented twofold more potency than 1, with an EC₅₀ value of 11.3 µM. Compounds 1–3 were also analyzed via molecular docking against L. (L.) infantum trypanothione reductase (TR) and thiol-dependent reductase 1 (TDR1), showing that the natural products 1 and 2 prefer specific regions in the active sites for lactone positioning. Docking of derivative 3 revealed interaction patterns between the different acetogenins, with the lactone moieties positioned in the same regions as compounds 1 and 2. Therefore, in silico prediction of metabolites from bioactive ligands can contribute to the design of potent derivatives, as demonstrated in this study, which aligns with our experimental findings.

Graphical abstract

Detailed molecular potential models of three major representative substances of V type agents were created and tested against the scarce available experimental results. Molecular Dynamics simulations were conducted, and first main focus was elucidating thermodynamic and transport properties of these highly toxic organophosphorus compounds. Alongside, an in-depth investigation of their intermolecular structure and vibrational spectra calculations were performed. Using classical simulations key thermodynamic quantities such as density, enthalpy of vaporization, heat capacity under constant pressure as well as transport properties such as viscosity and self-diffusion coefficient were computed. Molecular level structural organization was probed through pair radial distribution functions, providing insight into short range interactions and ordering of molecular sites and atoms, as well as coordination numbers. Furthermore, infrared spectra concerning vibrational states were derived from inverse Fourier transform of the total dipole autocorrelation function, revealing signature vibrational modes in the infrared fingerprint region for functional group identification. This combined approach offers a critical molecular insight into the behavior of V type chemical warfare agents under ambient conditions, contributing to predictive modelling and safe handling of these hazardous substances. This study presents the first comprehensive atomistic simulation of VX, RVX, and CVX, offering detailed thermodynamic, transport, and spectroscopic insights through refined OPLS-based potential models.

Kinase profiling is an essential step in both hit identification and selectivity evaluation. Since in vitro testing of large chemical libraries is costly and time-consuming, a computational approach can be applied to narrow down the reasonable chemical space. In this work, we collected data from several sources and prepared a curated, comprehensive database for training machine learning (ML) models to predict selectivity towards 75 kinases. We demonstrated the usefulness of this database by preparing several ML models with various molecular representations and model architectures. Among these, a graph neural network-based model enhanced by utilizing 3D pharmacophore ensembles showed the best performance. Finally, the developed model was applied to a library of in-stock compounds to facilitate kinase-focused drug discovery.

Graphical abstract

The degree and pattern of deacetylation in chitin-derived polymers critically determine their physicochemical properties and functional potential. In this study, a comprehensive theoretical analysis is performed of decameric chitin-like chains with systematically varied degrees of deacetylation (DDA), using density functional theory (DFT), electrostatic surface mapping, noncovalent interaction (NCI) analysis, and global reactivity descriptors. Structural optimizations revealed that partial deacetylation induces significant torsional rearrangements and enhanced intra-chain hydrogen bonding, leading to increased conformational flexibility. Molecular electrostatic potential (MEP) surfaces demonstrated a transition from neutral, acetyl-dominated topologies to highly polarized and reactive amine-rich domains. NCI analysis confirmed the emergence of cooperative hydrogen bonding and van der Waals networks in mid-range DDA structures. Furthermore, HOMO–LUMO analysis and TAFF-derived descriptors identified 20% ([[GlcNac]₄- GlcN]₂)–60% ([[GlcN]₃-[GlcNac]₂]₂) DDA chains as electronically soft, highly polarizable, and capable of dual electron donation and acceptance. These findings suggest that partially deacetylated chitosan chains exhibit a unique combination of flexibility, reactivity, and internal cohesion, providing a molecular rationale for their superior performance in biomedical and functional materials applications.

Direct inhibition of the Kelch-like ECH-associated protein 1 (Keap1)-nuclear factor erythroid 2-related factor 2 (Nrf2) protein–protein interaction (PPI) represents a critical pathway for enhancing the antioxidant response. Therefore, screening for direct Keap1-Nrf2 PPI inhibitors holds significant potential for addressing oxidative stress-related diseases. This study aims to develop an integrated approach to identify direct Keap1-Nrf2 PPI inhibitors from traditional Chinese medicine (TCM) using Achyranthis bidentatae Radix (ABR) as a case study. The approach incorporated ultrahigh-performance liquid chromatography-quadrupole-orbitrap mass spectrometry analysis, data mining, drug-like property evaluation, molecular docking, chemical structure clustering, molecular dynamics (MD) simulations, in vitro experimental validation, and density functional theory (DFT) calculations. A total of 517 compounds were identified in ABR, of which 248 met the drug-likeness criteria. Additionally, seventeen compounds from six structural clusters were identified as having theoretical Keap1-Nrf2 PPI inhibitory activity. Among these compounds, shidasterone, nortrachelogenin, wogonin, and N-trans-feruloylmethoxytyramine were subjected to experimental evaluation for their Keap1-Nrf2 PPI inhibitory and free radical scavenging activities. MD simulations and DFT calculations demonstrated that these compounds directly inhibited Keap1-Nrf2 PPI through hydrophobic interactions, hydrogen bonds, and salt bridges. Moreover, DFT calculations confirmed that these compounds scavenged free radicals via the hydrogen atom transfer mechanism. In conclusion, the strategy presented herein offers a robust framework for screening direct Keap1-Nrf2 PPI inhibitors with structural diversity from ABR and other TCM sources.

Graphical abstract

An integrated strategy was developed to screen direct Keap1-Nrf2 PPI inhibitors from TCM taking Achyranthis bidentatae Radix as an example.

A series of ten chloro- and bromo-substituted isatin derivatives were synthesized and evaluated for their ability to inhibit the monoamine oxidase (MAO) enzymes. All compounds demonstrated more potent inhibition of MAO-A compared to MAO-B. The most potent MAO-A inhibitor was HIB2 (IC₅₀ = 0.037 μM), followed by HIB4 (IC₅₀ = 0.039 μM), while HIB10 (IC₅₀ = 0.125 μM) exhibited the most potent inhibition of MAO-B. HIB2 was identified as a specific MAO inhibitor with a selectivity index of 29 for MAO-A over MAO-B. The enzyme-inhibitor dissociation constants (K_i) for HIB2 and HIB10 were 0.031 μM and 0.036 μM, respectively, for MAO-A and MAO-B. Both HIB2 and HIB10 exhibited competitive and reversible inhibition. An analysis of the ADMET and PAMPA suggested that HIB2 is permeable to the blood–brain barrier (BBB). Molecular docking analysis revealed that HIB2 forms stable hydrogen bonds with Asn181 and Gln215 in the MAO-A ligand–protein complex. Dynamic analysis indicated the stability of HIB2 with MAO-A. These findings suggest that HIB2 is potent reversible MAO-A inhibitor, making this class of compounds potential therapeutic agents for neurological disorders.

Diabetes mellitus remains a major global health challenge, necessitating the search for potent and safer therapeutic agents. In this study, a series of novel pyrrolo-imidazolidinone derivatives (1–10) was designed and synthesized as potential anti-diabetic agents. Structural elucidation was carried out using HREI-MS, ¹H-NMR and ¹³C-NMR spectroscopy. The anti-diabetic potential of the compounds was evaluated in vitro against α-amylase and α-glucosidase enzymes. Among the synthesized derivatives, compounds 4, 5, and 7 exhibited the most potent inhibitory activity, with IC₅₀ valuesranging between 4.10 ± 0.30 to 2.10 ± 0.10 µM (α-amylase) and 4.80 ± 0.40 to 2.60 ± 0.20 µM (α-glucosidase), surpassing the reference drug acarbose (IC₅₀ = 4.20 ± 0.60 µM and 5.10 ± 0.10 µM, respectively). In silico studies, including molecular docking, pharmacophore modeling, and ADMET profiling, supported the experimental findings and provided insights into the structural features governing enzyme inhibition and drug-likeness. The results highlight pyrrolo-imidazolidinone derivatives as promising scaffolds for further development of effective anti-glycemic agents.

Multidrug-resistant (MDR) Klebsiella pneumoniae poses a significant global health concern, particularly in hospital setting where it causes severe and hard-to-treat infections. In this study, 329 fungal-derived compounds were screened for their potential to inhibit MrkD1P, a key fimbrial adhesin protein (PDB ID: 3U4K) involved in host tissue adhesion. Molecular docking analysis identified ochratoxin A (− 9.1 kcal/mol), bromadiolone (− 8.6 kcal/mol), and permethrin (− 8.2 kcal/mol) as top-performing candidates, exhibiting strong binding affinities and stable molecular interactions, including hydrogen bonding and hydrophobic contacts. These findings were reinforced by 100-nanosecond molecular dynamics (MD) simulations, which showed sustained ligand–protein stability, particularly for ochratoxin A. Free energy estimations using the MM/PBSA method further suggested the thermodynamic favourability of these interactions. Pharmacokinetic profiling (ADMET) indicated favourable absorption and distribution properties for all three compounds, with low toxicity predictions, though some hepatotoxicity was noted. Principal component analysis (PCA) demonstrated that ochratoxin A and permethrin induced substantial alterations in protein dynamics, suggesting ligand-specific structural effects. Experimental validation confirmed the antibacterial activity of ochratoxin A against K. pneumoniae, producing a 34 ± 0.67 mm inhibition zone at 100 µg/disc, surpassing ciprofloxacin (33 mm) with a MIC of 18.33 ± 0.72 µg/mL and MBC of 39.33 ± 1.36 µg/mL (p < 0.05). Collectively, these in silico and in vitro results highlight fungal metabolites, particularly ochratoxin A, as promising therapeutic leads against MDR K. pneumoniae. However, further in vivo investigations are required to establish their safety and clinical potential.

Graphical abstract