Journal of Computer-Aided Molecular Design最新文献

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder affecting millions of people worldwide, with its prevalence expected to rise in the coming years. Due to the complexity of AD and the intricate interplay among its pathological mechanisms, the development of multitarget-directed ligands (MTDLs) has emerged as a promising therapeutic strategy. These compounds could simultaneously modulate multiple pathogenic pathways. Specifically, cholinergic and amyloid mechanisms, implicated in the onset of the disease, are regulated by AChE and BACE1, respectively. Therefore, targeting both pathways offers substantial therapeutic potential for AD. Computational tools can be useful in the identification of potential MTDL for these enzymes, reducing both costs and time in the drug discovery process. This review explores the relevance of this approach in the research and development for novel AD therapies, highlighting ongoing efforts focused on the identification and development of MTDLs for AChE and BACE1 inhibition through in silico methods. Virtual screening was the most frequently applied technique for a fast selection of ligands based on their affinity for the enzymes of interest. The in silico ADMET prediction also appears with a technique that allows the screening of compounds with drug-likeness. Moreover, evidence suggests that combining multiple computational methods can effectively identify drug candidates with optimized properties for target modulation and brain bioavailability.

The alarming rise in flaviviruses like Zika virus (ZikV) and dengue virus (DenV) has made them a major global health concern, especially in tropical and subtropical regions where nearly half of the global population is at risk. The urgency for safe and effective antiviral treatment is underscored by the fact that, despite growing research efforts, there are still no FDA-approved drugs available. The methyltransferases of ZikV and DenV, i.e., non-structural protein-5 (NS5), stand out as a highly conserved enzyme that is involved in viral replication and evasion of the host immune system via a capping mechanism, making them a key target for antiflaviviral drug development. In this study, we employed a virtual reality and cheminformatics-assisted pipeline followed by biological validation to identify a potent natural inhibitor of NS5. The systematic computational analysis identified the natural compound ZINC8952607 as a putative inhibitor with high binding affinity towards the methyltransferases of ZikV and DenV. Initial screening and docking analysis reveal that the lead compound strongly binds at the active site of the NS5 protein with a higher affinity. Further, extensive analysis involving molecular dynamics and DFT establishes greater reactivity of the lead compound and its stable complex formation with the NS5 protein. In vitro assay determined the cytotoxicity of the lead molecule with a CC₅₀ value of 3.43 ± 0.17 µM, indicating its reasonable safety as a lead molecule. These findings highlight ZINC8952607 as a potential lead candidate and reinforce the importance of targeting NS5 to develop new antiviral drugs.

Graphical abstract

Mutations in the CFTR gene play a pivotal role in the onset/severity of cystic fibrosis (CF). While our understanding of CFTR mutation classes is fragmented, this study focused on missense single nucleotide variants (SNVs) in the ABC transporter-like conserved site of the CFTR protein, employing various bioinformatics tools to identify deleterious amino acid substitutions (A.A.S). To gain a comparative understanding of the deleterious A.A.S in CFTR mutation classes, the in silico prediction of classified A.A.S by MutPred2 was cross-referenced with predictions of unclassified A.A.S. The study revealed twenty-five deleterious A.A.S in the conserved site. Nine of these (S549R, S549N, G551S, G551D, L558S, A559T, R560T, R560S, and A561E) were already classified. The in silico predictions for the remaining sixteen A.A.S exhibited similarities to molecular variations predicted for classified CFTR mutations and identified nine A.A.S falling under class II and seven A.A.S falling under class III, while four of these A.A.S in this conserved site may have effects of class II and III mutations. However, this classification is relative, warranting a comprehensive analysis to elucidate the intricacies of these nsSNVs. The combined use of modulators in therapy holds promise for more effective CF management, recognizing that CFTR mutations may exert effects that extend beyond a single class of mutation.

The identification of protein target classes is a key step in drug discovery, as it enables prioritization of screening campaigns and supports target-based drug repurpose. In this study, we developed a deep-learning pipeline based on a multilayer perceptron (MLP) trained on 15,804 curated compounds representing four major pharmacological target classes: G protein–coupled receptors (GPCRs), kinases, nuclear receptors, and transporters. Using extended connectivity fingerprints (ECFP4) as molecular descriptors, the model achieved 96% accuracy in internal cross-validation and 87% accuracy on an external test set, demonstrating performance comparable to ensemble classifiers such as Random Forest, XGBoost, and LightGBM. Class-specific F1 scores confirmed robust and balanced predictions across GPCR, kinase, nuclear receptor, and transporter categories. Model interpretability was addressed using SHAP values, which highlighted pharmacophore-like substructures consistent with known ligand–target interactions. Application to reference drugs further validated predictive utility, with correct assignment of most compounds to their canonical protein target class. The final MLP model was deployed as a user-friendly web application to facilitate accessible protein class prediction for novel compounds. Overall, this work presents a reliable and interpretable computational framework to support target-class-based drug discovery and repositioning.

Activity cliffs (ACs) present a significant challenge in structure-activity relationship (SAR) studies, characterized by pairs of similar compounds exhibiting substantial differences in biological activity. This paper investigates the interactions between protein kinases and their inhibitors to predict ACs by utilizing advanced deep learning techniques. Matched Molecular Pairs (MMPs) and MMP-ACs are systematically defined based on a minimum 100-fold difference in potency. A deep autoencoder model is developed for feature extraction phase followed by classification phase using various algorithms, including Support Vector Machines (SVM) and neural networks. Our results demonstrate that deep learning approaches can effectively capture complex patterns in molecular data, leading to robust predictions of ACs. Across all classifiers, our experiments show a strong correlation between the structural properties of inhibitors and their activity profiles against specific protein targets.

Fungal infections, especially those caused by naturally drug-resistant strains such as Candida glabrata, present a significant global health concern due to the limitations of existing antifungal treatments, which are often compromised by toxicity and resistance. In this study, we investigate the potential SUMO-specific protease, Ulp2, in C. glabrata (CgUlp2), a deSUMOylating enzyme essential for maintaining protein homeostasis, as a target for antifungals. Structural analyses revealed significant differences between CgUlp2 and its human counterpart. Molecular docking studies identified a potential binding region between CgUlp2 and CgSmt3, which led us to screen for small molecules that could interfere with this protein–protein interaction. We identified the FDA-approved compounds silymarin and honokiol as promising candidates through pharmacophore-based virtual screening. Docking results revealed that silymarin and honokiol interact with the CgUlp2–CgSmt3 protein complex. Consistent with these computational findings, our growth assays show that silymarin and honokiol inhibit the growth of C. glabrata, and overexpression of CgUlp2 could partially rescue this growth defect. These findings underscore the potential of CgUlp2 as a key target for developing new antifungals, representing a significant step forward in the fight against fungal infections. While this study offers promising computational insights, it is limited to in silico predictions. Experimental validation, including enzyme inhibition assays and in vivo efficacy testing, is essential for clinical translation.

Metal-based drugs have historically been used for different therapeutic and diagnostic applications. The need to advance metal-based drugs and drug candidates has contributed to the development of computational strategies to handle metal-containing chemical structures for drug discovery applications. However, most of them have limitations related to their computational cost, scalability, capacity to be used in different chemical contexts, and data accessibility. This study introduces the first open-access metal-based approved drug database (MetAP DB) with FDA-approved drugs across various therapeutic applications and clinical diagnostic outcomes. It also introduces the first versions of molecular metal-based fingerprints (Metal-FP) representations proposed as a general protocol for representing metal-based compounds. The proposed fingerprints encode chemical data related to the presence of metals, their valence and oxidation states, the presence of specific functional groups, and the atom connectivity of metal-based compounds. The fingerprints Metal-FP2 and Metal-FP3 showed highlighting the effectiveness in differentiating metal-based compounds according to distance metrics, data visualizations, random forest, and logistic regression algorithms.

The global dynamics of pyruvate kinase were examined using molecular dynamics (MD) simulations to investigate the effects of allosteric inhibition through bond restraints applied at two key allosteric sites. The study employed the experimentally resolved structure of the enzyme complexed with the allosteric inhibitor IS-130 at the small C–C interface, serving as a reference for analyzing an additional, computationally predicted allosteric site at the large A-A interface. Simulations identified the B and CT domains as the most mobile regions, with bond restraints at either interface significantly reducing CT domain flexibility up to 9 Å across all chains. Restraints at the C–C interface limited minimal global conformational sampling, whereas restraints at the A-A interface altered the dynamic profile without narrowing the sampled conformational space, suggesting distinct regulatory roles for each interface. Distance fluctuation analyses revealed enhanced interchain communication and reduced mobility near restrained sites, suggesting that these restraints reinforce allosteric inhibition by stabilizing otherwise flexible domains. Cross-correlation analysis showed a marked reduction in long-range residue-residue correspondence, especially under C–C restraints, indicating disrupted dynamic coordination essential for catalytic activity. Mutual information analysis, capturing both linear and non-linear dependencies, further supported these findings by showing a widespread loss of dynamic correspondence in positional fluctuations across the receptor upon restraint application. Notably, although the C–C interface has been experimentally linked to inhibition, these results suggest that the computationally predicted large A-A interface may also contribute to allosteric regulation. Together, these findings highlight the distributed and cooperative nature of allosteric control in pyruvate kinase.

Antimicrobial Resistance (AMR) is a global concern demanding high-throughput and precise AMR surveillance strategies. This review provides a comprehensive list of Artificial Intelligence (AI) driven frameworks widely employed in the early detection, structural characterization, and designing of novel inhibitors to block the resistance pathways critical for AMR. Deep learning algorithms including DeepGO, DeepGOPlus, DeepGO-SE, PFresGO, DPFunc, ProtENN and graph-based architectures of GraphSite, GrASP enables precise functional annotation of resistance-associated proteins. AI-guided protein modeling performed by AlphaFold, RoseTTAFold, ProtGPT-2, ESMFold etc. generates high resolution 3D conformations, further utilized in performing molecular docking via tools like AutoDock, DeepDocking and DeepChem and analyzed with tools like DeepDriveMD, TorchMD, and PRITHVI, which can perform real-time molecular dynamics simulations. Identification of relevant resistant biomarkers from mass-spectrometry profiles can also be achieved with the help of DeepNovo, Casanovo, or Prosit. Tools like DeepARG, HMD-ARG, and BacEffluxPred enables identification of unannotated resistance genes from metagenomic samples. Natural Language Processing (NLP) and Large Language-based models (LLM) facilitate identification of resistant determinants via literature mining enabling regulatory network mapping and rational inhibitor design. Furthermore, AI-mediated de-novo inhibitor design is achieved using Variational Autoencoders (VAE), Generative Adversarial Networks (GAN), diffusion and flow-matching based frameworks serve as potential options for enhancing diagnostic interventions against resistant phenotypes. AI-based protein–protein interaction predictors include DeepInteract, Pred_PPI, PLIP, DeepAIPs-Pred, DeepAIPs-SFLA, SBSM-Pro, Deep Stacked-AVPs, and pNPs-CapsNet help in understanding how resistance proteins interact with each other enabling precise identification of AMR-modulating peptides and supports the modeling of novel antibiotics for blocking interactions and disrupting resistance pathways.

Graphical abstract