Gelsemium has a long history of medicinal use but is also a poisonous plant. Some low-toxicity alkaloids in Gelsemium exhibit anxiolytic, anti-inflammatory, analgesic, and other pharmacological effects; however, certain alkaloids in Gelsemium are highly toxic. Nevertheless, the molecular targets underlying the biological effects of Gelsemium alkaloids remain poorly understood. We employed electrophysiological techniques and molecular modeling to examine the modulatory effects of Gelsemium alkaloids on inhibitory neurotransmitter receptors, as well as to elucidate the mechanisms underlying their molecular interactions. Our findings indicate that low-toxicity alkaloids primarily exert their pharmacological effects through actions on glycine receptors, with the binding site located at the orthosteric site between two α-subunits. Both highly toxic and low-toxicity alkaloids target GABAA receptors, using the β+/α− interface transmembrane structural domains as common binding sites. These results identify the targets through which Gelsemium alkaloids affect the central nervous system and predict the binding modes and key amino acids involved from a computational modeling perspective. However, further experimental validation through mutational studies is necessary to strengthen these findings.

TRPV1 channel agonists and antagonists, which have powerful analgesic effects without the addictive qualities associated with traditional analgesics, have become a focus area for the development of novel analgesics. In this study, quantitative structure–activity relationship (QSAR) models for three bioactive endpoints (K_i, IC₅₀, and EC₅₀) were successfully constructed using four machine learning algorithms: SVM, Bagging, GBDT, and XGBoost. These models were based on 2922 TRPV1 modulators and incorporated four types of molecular descriptors: Daylight, E-state, ECFP4, and MACCS. After the rigorous five-fold cross-validation and external test set validation, the optimal models for the three endpoints were obtained. For the K_i endpoint, the Bagging-ECFP4 model had a Q² value of 0.778 and an R² value of 0.780. For the IC₅₀ endpoint, the XGBoost-ECFP4 model had a Q² value of 0.806 and an R² value of 0.784. For the EC₅₀ endpoint, the SVM-Daylight model had a Q² value of 0.784 and an R² value of 0.809. These results demonstrate that the constructed models exhibit good predictive performance. In addition, based on the model feature importance analysis, the influence between substructure and biological activity was also explored, which can provide important theoretical guidance for the efficient virtual screening and structural optimization of novel TRPV1 analgesics. And subsequent studies on novel TRPV1 modulators will be based on the feature substructures of the three endpoints.

Central nervous system (CNS) disorders such as depression severely impair human health. Targeted inhibition of the GluN1-GluN2A receptor is a promising therapeutic strategy, but current drugs often have adverse effects. To develop novel candidate drugs, this study utilized the (S)-ketamine and GluN1-GluN2A receptor complex as a structural template and conducted de novo drug design with the DrugFlow platform. An integrated strategy of molecular docking-based virtual screening combined with high-throughput binding free energy (∆G_binding) calculations from large-scale molecular dynamics (MD) simulations identified three promising antagonists. The ∆G_binding values of these compounds are all below −18.98 kcal/mol, indicating stronger binding affinity than (S)-ketamine, and they demonstrate promising drug-like properties and development potential. 200-ns MD simulations confirmed their stable receptor binding and mechanism consistent with (S)-ketamine. Electrophysiological recordings revealed that, at a concentration of 10 μM, Compounds A1, A2, and A3 produced concentration-dependent inhibition of GluN1-GluN2A receptor-mediated currents, with fractional inhibition values of 24.26%, 35.36%, and 41.76%, respectively. These findings demonstrate the compounds’ potential as CNS disorder therapeutics, requiring further experiments to validate efficacy and advance development for conditions like depression.

To tackle the collaborative optimization of high energy, low sensitivity, and thermal stability in energetic materials (EMs), we developed an interpretable multitask graph network (QGeoSEP++) using 7,498 experimental data points from 1,847 EMs, covering six key properties. Integrating 3D geometry-enhanced representation, atomic multihead self-attention, and a knowledge embedding module for expert descriptors (e.g., oxygen balance), QGeoSEP++ achieves both high prediction accuracy and strong interpretability. A dynamic loss weighting strategy based on predictive uncertainty further optimizes multi-property synergy. Ablation studies show the combined modules reduce the average RMSE by 16.2% and boost the average R2 by 3.8%, outperforming mainstream models (D-MPNN, SchNet, MTL-GNN) across all tasks. Leveraging attention weights, QGeoSEP++ identifies critical substructural features governing multi-dimensional properties, overcoming the interpretability bottleneck of black-box models. An online server (https://emp.cmdrg.com) is deployed to enable rapid property prediction, structural visualization, and efficient virtual screening of novel EMs.