北京义翘神州科技股份有限公司

The upregulation of interleukin-1 receptor-associated kinase 4 (IRAK4) drives pro-tumorigenic signaling across various malignancies. Currently available IRAK4 inhibitors are all limited by suboptimal selectivity and off-target toxicity. To develop non-toxic and mechanistically focused IRAK4 inhibitors, an in silico machine learning (ML) pipeline combined with structure-driven drug repositioning was employed. An IRAK4-targeting dataset of bioactive compounds was systematically filtered and curated to train a random forest (RF) regression model. Comparative cross-validation against 41 independent ML frameworks demonstrated reasonable predictive accuracy, and the optimized model was deployed to screen a library of 1040 FDA-approved therapeutics. Orthogonal molecular docking supported the binding efficacy of RF-identified lead compounds, including udenafil, linagliptin, benflumetol, and nimodipine. Thermodynamic and conformational stability were validated using molecular dynamics (MD) simulations, yielding stable root mean square deviation (RMSD), radius of gyration (Rg), root mean square fluctuation (RMSF), principal component analysis (PCA), hydrogen bonding, and MMGBSA/MM-PBSA profiles. The efficacy of these candidates was further supported by their low toxicity profiles, via graph neural network (GNN)-based toxicity analysis. These findings establish a scalable computational strategy for identifying repurposable oncology therapeutics and propose udenafil, linagliptin, benflumetol, and nimodipine as mechanistically credible candidates for experimental validation.

A convenient, catalyst-free protocol for the synthesis of hitherto unknown alkyne-tethered N-acyl pyrazolines from readily available hydrazides and skipped diynones has been developed. The subsequent acid-catalyzed dehydration of the synthesized hydroxy derivatives opened a shortcut toward a representative set of 5-alkynyl-1-acylpyrazoles. The elaborated approaches possess competitive advantages such as complete chemo- and regioselectivity, availability of starting materials and excellent yields (up to quantitative) of target products having rich functionality patterns for further derivatization.

Transformer architectures, capable of capturing sequential dependencies in the history of user interactions, have become the dominant approach in sequential recommender systems. Despite their success, such models consider sequence elements in isolation, implicitly accounting for the complex relationships between them. Graph neural networks, in contrast, explicitly model these relationships through higher order interactions but are often unable to adequately capture their evolution over time, limiting their use for predicting the next interaction. To fill this gap, we present a new framework that combines transformers and graph neural networks and aligns different representations for solving next-item prediction task. Our solution simultaneously encodes structural dependencies in the interaction graph and tracks their dynamic change. Experimental results on a number of open datasets demonstrate that the proposed framework consistently outperforms both pure sequential and graph approaches in terms of recommendation quality, as well as recent methods that combine both types of signals.

Functional Magnetic Resonance Imaging (fMRI) enables the non-invasive mapping of brain activity during various cognitive processes, and understanding how specific cognitive tasks activate distinct brain regions is crucial for both neuroscience and clinical applications. However, acquiring task-based fMRI (tfMRI) is costly, time-consuming, and often infeasible for individuals unable to perform tasks. Resting-state fMRI (rsfMRI), which is more widely available and does not require task compliance, can instead be leveraged to infer task-related activations. In this work, we propose BrainSparseCNN, a sparse 3D convolutional neural network that exploits the brain’s spatial structure to predict tfMRI contrasts from rsfMRI while efficiently processing high-dimensional neuroimaging data. BrainSparseCNN achieves up to 7% higher Pearson correlation than surface-based and dense volumetric competitors, with statistically significant gains (p < 0.01) across all tasks, furthermore improving spatial alignment, subject identification accuracy, and saliency interpretability while maintaining computational efficiency. Overall, BrainSparseCNN provides a more precise, interpretable, and scalable framework for inferring individual functional activation maps from resting-state data, with code publicly available at https://github.com/univanxx/brainsparsecnn.

Women’s mental health is a vital yet underexplored aspect of empowerment and social progress, particularly in rural India. This study focuses on Bangalore Rural, Chikkaballapur. where women encounter multiple barriers to psychological well-being, including poverty, patriarchal norms, limited healthcare facilities, and domestic violence. These stressors often lead to high rates of depression, anxiety, and trauma, while stigma and lack of awareness further restrict help-seeking behavior. Although programs such as the National Mental Health Programme (NMHP) and the District Mental Health Programme (DMHP) exist, their outreach in rural communities remains inadequate. Informal networks like families and Self-Help Groups (SHGs) provide some level of emotional and social support, but they are fragmented and lack professional guidance. The research highlights the pressing need to strengthen rural mental health services through awareness initiatives, accessible professional care, and community-based interventions. Addressing these gaps is essential for improving women’s well-being, ensuring family stability, and fostering inclusive rural development.