Molecular Diversity最新文献

Predicting cancer drug responses is crucial for precision medicine. This study proposes AMDRP, a novel model that predicts drug responses by integrating drug features-represented as molecular graphs and extended connectivity fingerprints (ECFP)-with multi-omics data from cancer cell lines. AMDRP incorporates an Adaptive Feature Fusion (AFF) module to dynamically weight and fuse these drug features, resulting in enhanced drug representations. Furthermore, a multi-head bidirectional cross-attention (MBCA) module is introduced to model deep interactions between drug and cell line features. Extensive experiments demonstrate that AMDRP achieves significantly higher prediction accuracy than state-of-the-art baselines. Ablation studies confirm the critical contribution of both modules, with ECFP features providing substantial performance gains. The model's robustness and generalization capability were rigorously evaluated through cross-dataset validation and leave-one-out experiments, demonstrating its effectiveness against data distribution shifts. Predictions and enrichment analysis on unknown drug-cell line pairs underscore the model's predictive power and biological relevance. These results indicate that AMDRP is an effective tool for predicting cancer drug responses and holds potential value for guiding anticancer drug discovery.

Diabetes is a chronic medical disorder caused by insufficient production of the hormone insulin by the pancreas. Although there are various treatment options available for controlling diabetes, including non-peptide-based medications, the majority of these have adverse effects and are limited in comparison to peptide-based drugs. Protein drugs offer numerous benefits, including weight loss, significant reductions in blood glucose levels, and an extremely low risk of hypoglycemia. This article discusses treatment modalities, presents existing therapies, provides an in-depth comparison of peptide-based and other drugs, examines current development and barriers, offers some recommendations, and outlines future research directions for peptide drugs in the treatment of T2DM. In recent days, several computational tools and AI models, including ESMFold, ProteinMPNN, Schrödinger, and AutoDock Vina, have played an essential role in peptide-based drug discovery. Therefore, this article also highlights the significance of AI drug discovery, diverse AI models, and other computational tools to enhance peptide-based drug discovery and development.

Isoform-selective inhibition of class I phosphoinositide 3-kinases (PI3Ks) remains a major challenge in oncology and immune-mediated diseases, where dysregulated PI3K signaling drives tumor progression, therapeutic resistance, and aberrant immune activation. Efforts to achieve precise isoform selectivity are constrained by the high structural similarity among the catalytic subunits α, β, δ, and γ. In this regard, we developed an artificial intelligence (AI)-driven integrative framework that combines machine learning-based quantitative structure-activity relationship (ML-QSAR) modeling, fragment-level selectivity profiling, and reinforcement-learning generative chemistry to design isoform-selective PI3K inhibitors. Curated ChEMBL datasets were used to train independent XGBoost models for each isoform, achieving strong predictive performance (R² = 0.76-0.82; RMSE = 0.48-0.51) and interpretable SHapley Additive exPlanations (SHAP)-based feature attribution. Fragment analysis identified isoform-specific structural motifs that were used to guide targeted molecular exploration with the FREED +  + reinforcement-learning algorithm. The framework generated over 10,000 unique compounds, and molecular docking analysis showed favorable binding energies (- 7.9 to - 9.7 kcal/mol) and interactions consistent with known isoform-selective inhibitors. Generated molecules also exhibited suitable drug-likeness and synthetic accessibility, highlighting their potential as viable lead compounds. Collectively, this study demonstrates how combining predictive ML models with fragment-aware generative AI enables rapid discovery of selectivity-optimized PI3K inhibitors. The proposed pipeline is generalizable to other multi-isoform targets and establishes a scalable AI methodology for next-generation rational drug design in precision oncology and immune-modulating drug development.

Alzheimer's disease (AD) is a powerful neurodegenerative disease characterized by cholinergic deficiency, where the inhibition of acetylcholinesterase (AChE) remains a clinically validated strategy. In our current work, a virtual screening platform supported by machine learning identified new inhibitors of AChE out of a structurally diverse collection of 2,895 marine bacterial natural products. Following a curation based on a structure-based strategy, a robust regression model was constructed from the physicochemical and structural characteristics of the reported inhibitors of AChE in an attempt to predict the inhibitory strength (pIC₅₀) of the top-scored ligands. The model had high predictive fidelity and led to the selection of twenty prospective candidates, out of which three (CMNPD25858, CMNPD28646, and CMNPD28412) were shortlisted according to activity profiles and drug-likeness filters. The shortlisted compounds were prepared for quantum-level refinement through density functional theory in order to improve electronic and structural precision. These optimised ligands were then evaluated under physiological conditions in terms of binding stability, conformational study, and intermolecular interaction through all-atom molecular dynamics simulation. CMNPD25858 demonstrated outstanding structural retention, stable persistent hydrogen bonding, and negligible displacement in the catalytic site. Principal component analysis and free energy landscape mapping revealed a highly confined, energetically favorable conformational basin. Structural overlays of post-simulation minima with initial docking poses confirmed minimal divergence. MM-GBSA free energy calculations substantiated the superior binding affinities of CMNPD25858 (-87.90 kcal/mol) and CMNPD28646 (-83.44 kcal/mol) relative to the reference compound. In vitro AChE inhibition assays revealed that compound CMNPD25858 demonstrated the highest inhibition (75%) at 1 mg/ml, followed by CMNPD28646 (64%) and CMNPD28412 (57.81%), consistent with in silico predictions when compared to the standard Donepezil (95.27%). Therefore, these integrative studies highlight the strategic utility of machine learning in accelerating structure-activity prediction and rational hit selection, and identifies marine-derived CMNPD25858 and CMNPD28646 as potent, dynamically stable AChE inhibitors with high potential for anti-Alzheimer's therapeutic development.