Journal of Computer-Aided Molecular Design最新文献

Phosphoinositide kinase PIKfyve is a key regulator of endosomal trafficking and lysosomal function and is increasingly recognized as a therapeutic target in cancer, neurodegeneration, and viral infections. However, the discovery of selective and safe inhibitors remains limited. Here, we integrated computational screening with biochemical validation to identify natural products (NPs) as potential PIKfyve inhibitors. From a library of over 3500 NPs, trilobatin—a dihydrochalcone glycoside—emerged as the most promising hit. It exhibited strong binding affinity in docking and MM-GBSA analyses, favourable pharmacokinetic and drug-likeness profiles, and acceptable safety indices. Molecular dynamics simulations (100 ns) confirmed the stability of its interactions with key residues in the active site, supported by persistent hydrogen bonding and robust electrostatic and hydrophobic contributions. Experimental kinase inhibition assays validated these findings, revealing that trilobatin inhibits PIKfyve activity in a dose-dependent manner with an IC₅₀ of ~ 0.29 µM. These results establish trilobatin as a potent non-morpholine scaffold distinct from conventional morpholine-based PIKfyve inhibitors, thereby expanding the chemical diversity available for lipid kinase targeting. Importantly, this study demonstrates the value of integrating in silico screening, ADMET prediction, and biochemical validation to accelerate early-phase drug discovery. Trilobatin thus represents a promising lead for further structure–activity optimization, cellular assays, and in vivo evaluation. Beyond its immediate relevance, this work underscores the potential of NPs with favourable pharmacological profiles as sources of novel scaffolds for selective kinase inhibition and highlights trilobatin as a compelling candidate for therapeutic development across PIKfyve associated pathologies.

Curcumin, a natural polyphenol, exhibits broad anticancer properties but is limited by poor stability and bioavailability. To overcome these flaws while enhancing potency, a series of five novel difluoroboron curcumin analogues (Cox1-Cox5) were synthesized and characterized. Their cytotoxic activity was evaluated against the K562 human leukemia cell line using the MTT assay, revealing IC₅₀ values ranging from a potent 27.1 µg/mL to 58.6 µg/mL. To elucidate the structural basis of this activity, a comprehensive computational investigation was performed. Molecular docking studies identified human thymidylate synthase (hTS) as a plausible molecular target, with predicted binding energies showing a strong linear correlation (R² = 0.88) with the experimental IC₅₀ values. Density Functional Theory (DFT) calculations revealed that high electrophilicity is the key determinant of potent binding, with the nitro-substituted emerging as the most active compound. Finally, a Molecular Dynamics (MD) simulation of the Cox5-hTS complex confirmed the dynamic stability of the docked pose and the persistence of key hydrogen-bonding interactions. This integrated study validates a rational design strategy where stabilizing the curcumin scaffold and introducing potent electron-withdrawing groups yields compounds with superior and mechanistically understandable anti-leukemic activity.

Phosphodiesterases (PDEs), particularly PDE1C, regulate cyclic nucleotide signaling and are promising therapeutic targets for diseases such as cardiovascular disorders, pulmonary hypertension, neurocognitive conditions, and certain cancers. However, the development of selective PDE1C inhibitors is hindered by the structural diversity and functional redundancy within the PDE family, with only one inhibitor, ITI-214, reaching clinical trials. Traditional experimental screening methods are resource-intensive and often yield suboptimal results, necessitating more efficient approaches. In this study, we employed an integrated computational strategy combining machine learning (ML), molecular docking, and molecular dynamics (MD) simulations to rapidly screen for novel PDE1C inhibitors. An ML model was developed to predict PDE1C inhibitory activity, validated with an out-of-sample dataset, and applied to compounds pre-selected via molecular docking (docking score ≤ -10.00 kcal/mol) to estimate pIC₅₀ values. Six representative compounds were subjected to 100 ns MD simulations to assess binding stability with the PDE1C protein. Top-ranked compounds underwent in vitro validation, confirming two candidates with high PDE1C inhibitory potency. This multi-tiered approach enhances screening efficiency, mitigates individual method limitations, and provides a robust framework for identifying PDE1C inhibitors, paving the way for further lead optimization and preclinical development.

Cancer immune evasion is predominantly mediated through immune checkpoint pathways, such as the PD-1/PD-L1 axis. In this mechanism, PD-L1, which is often overexpressed on tumor cells, binds to PD-1 receptors on T cells, resulting in the inhibition of T cell activity and allowing tumors to evade immune surveillance. Targeting this interaction is of therapeutic significance. Marine fungal metabolites were investigated as potential PD-L1 inhibitors using a multi-level computational approach that combines quantum chemical, dynamic, energetic, and machine learning studies. Preliminary virtual screening narrowed down the list to the top four contenders, CMNPD20987, CMNPD20986, CMNPD24819, and CMNPD20907, with docking values ranging from − 10.7 to − 8.2 kcal/mol. HOMO-LUMO gap analysis based on the density functional theory demonstrated the highest electronic stability of CMNPD24819 (5.087 eV) and the highest reactivity of CMNPD20907 (3.954 eV). Redocking studies highlighted stable interactions with critical PD-L1 amino acid residues like Tyr⁵⁶ (π–π stacking), Asp¹²², Gln⁶⁶, Ile¹¹⁶, and Lys¹²⁴ (hydrogen bonding). Triplicate 200 ns MD simulations established the structural stability of the chosen complexes with low RMSD and RMSF values. MM/GBSA binding free energies estimated significant affinity, with notable affinity for CMNPD24819 (− 34.39 kcal/mol) and CMNPD20987 (− 30.63 kcal/mol). Analysis of free energy landscapes showed deep minima of the free energy basin, indicating stable conformational states. The machine learning regression model trained on ChEMBL PD-L1 inhibitors predicted high pIC50 values for the selected compounds, with CMNPD20907, CMNPD20986, CMNPD20987, and CMNPD24819scoring above the reference molecule. This holistic analysis highlights the electronic strength, beneficial binding profiles, and biomedical value of the marine fungal metabolites as potential future immune checkpoint inhibitors of cancer.

Integrating the techniques of deep learning, particularly graph neural network models, has made a significant advancement in drug discovery by facilitating effective exploration of chemical spaces and precise prediction of molecular properties. This paper presents a deep learning approach: molecular encoder, property predictor, molecular generator, and realness classifier (ME&PP–MG&RC). which integrates the molecular encoding and property prediction (ME&PP) component and the molecular generation and realness classification (MG&RC) component to help the prediction and discovery of new molecular compounds. Encoding molecular structures in latent space using graph neural network models, our approach allows for accurate prediction of properties such as quantum mechanical properties (HOMO, LUMO, and gap) for QM9 and pharmaceutical properties (QED, LogP, and SAS) for ZINC datasets. ME&PP component achieves the highest performance in property prediction with lower average absolute errors compared to existing methods after evaluation. Additionally, the approach produces molecules that are chemically diverse and valid, and they are checked by a realness classifier to make sure that they are practical. Results demonstrate the potential of ME&PP–MG&RC to revolutionize computational molecular design and drug discovery, providing a unified pipeline for precise property predictions, new molecular generation, and reality validation. This integrated approach represents a major step towards the automation of drug discovery and paves the way for the accelerated identification of innovative therapeutic candidates.

Pectic acid (polygalacturonic acid, PGA) and pectin oligosaccharides (POS) have shown selective anticancer activities with cytotoxicity against various cancer cells, including breast adenocarcinoma MCF-7. While promising, their anticancer activities against the nasopharyngeal carcinoma (NPC) have yet to be determined. To this end, the antiproliferation activities of pectic acid and enzymatically derived unsaturated POS trimer were investigated against the EBV-positive NPC cell line, C666-1, as well as on MCF-7 (as positive control) and NIH-3T3 cells (as non-tumorigenic cell model). Our study shows that both POS trimer and PGA can inhibit C666-1 cells’ proliferation at concentrations above 20 mg/mL and 2.5 mg/mL, respectively. Previous studies had also demonstrated that one of the key biological targets for these compounds is the human galectin-3 (hGal3). However, the binding mode of pectic acid and the unsaturated POS trimer to human galectin-3 is still lacking. To investigate this, in silico molecular docking and all-atom molecular dynamics simulations were performed. Analysis of the MD trajectories revealed that the unsaturated POS trimer prefers binding to hGal3 oligomer over hGal3 monomer with a ligand: protein monomer ratio of 1:2 and 1:3. On the other hand, both hGal3 monomer and dimer were found to bind stably to the POS 15-mer (as PGA model) along its linear chain throughout the simulation. In conclusion, both unsaturated POS trimer and PGA have the potential to be novel anticancer agents against EBV-positive NPC, though further development to increase their potency and/or selectivity would be necessary.

In recent decades, the rapid pace of digital transformation marks a transformative era for the healthcare and pharmaceutical industries. The incorporation of innovative technology, specifically Artificial Intelligence (AI) and its derivatives, has driven significant innovation and greatly enhanced the efficiency of biomedical research and drug discovery processes. Among critical biological targets, the p53 protein is essential for controlling cell cycle regulation and tumor suppression. Although p53 has long been considered undruggable, recent research has revived interest in targeting it with novel therapeutics. In this paper, A novel Hybrid Drug-Target Interaction IC50 (HDTI-IC50) prediction model is proposes to predict IC50 values. The model integrates Graph Convolutional Networks (GCNs) as well as Graph Attention Networks (GATs) by sequentially stacking their hidden layers. This hybrid architecture leverages the strengths of both models. Specifically, GCNs are first applied to effectively capture local structural information and perform well under homophily assumptions. Then, GAT is learned to model long-range dependencies and handle heterophilic graphs. By integrating both, the model learns richer node representations and can adapt to diverse graph structures. Following these layers, a global pooling mechanism follows, which combines Global Max Pooling (GMP) and Global Average Pooling (GAP). Compared to related approaches, which mainly perform general IC50 prediction or binary activity classification, the proposed HDTI-IC50 model provides a unified framework specifically tailored for p53 inhibitors. Unlike previous approaches that rely on conventional molecular descriptors and overlook structural topology, our model utilizes graph-based representations to capture both local and global molecular relationships. By sequentially integrating GCN and GAT layers, the model effectively combines localized structural learning with attention-based feature refinement, resulting in improved representation capability and predictive performance. The dataset applied in this paper is obtained from the database of the Genomics of Drug Sensitivity in Cancer (GDSC). Model performance is evaluated using standard regression metrics, involving Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and coefficient of determination (R²). The performance rate of MAE is 0.1, RMSE is 0.19, and R² is 0.8 demonstrating superior performance compared to state-of-the-art methods. It also achieves an average inference time of 7.70 s. This paper proposes a HDTI-IC50 model to predict IC50 for p53inhibitors. Results from experiments indicate that the proposed HDTI-IC50 model outperforms individual GCN, GAT-based, and other related drug-target models as well as baseline regression models. demonstrating both its predictive accuracy and computational economy.

Radiosensitizers are agents that make tumour cells more sensitive to radiation therapy. One key mechanism involves inhibition of the DNA-dependent protein kinase (DNA-PK), an enzyme crucial for repairing DNA double-strand breaks in mammalian cells. Suppression of the DNA-PK enzyme compromises the double-strand break repairs to amplify the radiation induced toxicity among the tumour cells. In this study, 73 6‑Anilino Imidazo[4,5‑c]pyridin-2-one derivatives were curated as potent DNA-PK inhibitors and subjected them to 2D -and 3D-Quantitative Structure Activity Relationship analyses to explore their structural requirements. Apart from conventional methodology, we implemented newly developed MolSHAP analyses for R-group analyses. Significant information regarding structural requirements were retrieved from each of these cheminformatic analyses. Additionally, to understand the interaction between the ligands and the DNA-PK receptor, molecular dynamics (MD) simulation analysis of 100 ns were carried out for the most and the least potent compounds among the dataset. The findings indicated H-bond and π-π interactions to be the key factors for binding interactions. Furthermore, novel ligands were designed through the MolSHAP tool and were validated through the chemometric model developed in this investigation. The designed compound exhibited favourable predicted activity and replicated key interaction profiles of the co-crystallized bound ligand in MD simulations. The investigation was carried out through open-access tools to safeguard reproducibility and accessibility among researchers.

The SARS-CoV-2 papain-like protease (PLpro) represents a crucial therapeutic target due to its dual role in viral polyprotein processing and suppression of host immune responses through de-ubiquitination and de-ISGylation activities. To identify novel allosteric druggable sites on PLpro, we developed a molecular dynamics approach with flooding fragments (MDFFr), which extends a previously established method –Molecular Dynamics flooding– enabling broader applicability across biological targets. Using MDFFr, we evaluated interactions of known phenolic inhibitors with SARS-CoV-2 PLpro and identified several biologically significant sites, encompassing allosteric hotspots, cryptic pockets, and regions involved in protein–protein interactions. Our simulations not only confirmed experimentally characterized binding sites, including fragment-binding and protein–protein interaction regions for ubiquitin and ISG15 (Interferon-Stimulated Gene 15), but also uncovered previously unrecognized hotspots for further investigation. These results establish MDFFr as a suitable approach for physics-based druggability assessment of biological targets using only protein 3D structure, while providing detailed insights into fragment-protein interactions at both druggable sites and protein–protein interfaces. These findings also unveil new opportunities for allosteric inhibition of PLpro, potentially advancing therapeutic strategies against SARS-CoV-2 and other coronavirus-related diseases. Furthermore, by using “real” drug-like fragments (rather than standard cosolvent “probes”), MDFFr enhances translational relevance and directly informs drug repurposing and ligand discovery efforts.

Two series of alizarine derivatives containing vanillin scaffold (10a-h) or aromatic amide function (12a-h) were synthesized and structurally characterized. The cytotoxic evaluation revealed higher activity towards leukemia cancer cell lines (K562 and HL-60) than solid tumor cells (HeLa and MCF-7). The compound 10 h, containing a benzyl group, showed the most prominent activity against K562 cells, and the lowest toxicity towards healthy cells among all active derivatives. The most active compounds 10f, 10 h, and 12 h were further investigated and induced a significant increase in the percentage of HeLa, K562, and HL-60 cells in the subG1 cell cycle phase in comparison with the control cells. Compounds 10f and 10 h activated apoptosis in K562 cells through all three tested caspases, while derivative 12 h only induced the activation of the main effector caspase-3. Molecular docking simulations suggest that these compounds can form stable complexes with caspase-3, consistent with their experimentally confirmed involvement in caspase-dependent apoptotic pathways. All three tested derivatives demonstrated moderate to strong binding to bovine serum albumin (BSA), with preferential occupation of subdomain IIA (site I), as supported both experimentally and through docking studies. The interaction study of these compounds with DNA indicated their ability to interact with ct-DNA through the minor groove.