The tumor microenvironment is a dynamic eco system where cellular interactions drive cancer progression. However, inferring cell-cell communication from non-spatial scRNA-seq data remains challenging due to incomplete li gand-receptor databases and noisy cell type annotations. H ere, we propose scGraphDap, a graph neural network frame work that integrates functional state pseudo-labels and graph structure learning to improve both cell type annotation an d CCC inference. By leveraging pathway activity scores (e. g., angiogenesis, apoptosis) as pseudo-labels, scGraphDap optimizes cell-cell graphs to capture functional proximity be yond geometric similarity. Furthermore, a graph domain adaptation module aligns cell embeddings across patients, enhancing cross-individual generalization. Evaluated on 38,667 cells from 15 patients across three cancers, scGraphDap ac hieved an average accuracy of 82.82%. Statistical validation confirmed its ability to recover disease-specific gene interactions (e.g., STAT3-CD274 in breast invasive carcinoma) without prior knowledge. The source code and data used in this paper can be found in https://github.com/LiYuechao1998/sc GraphDap. Our work provides a unified framework for TME analysis, offering insights into therapeutic target discovery.

Functional MRI (fMRI) and single-cell transcriptomics are pivotal in Alzheimer's disease (AD) research, each providing unique insights into neural function and molecular mechanisms. However, integrating these complementary modalities remains largely unexplored. Here, we introduce scBIT, a novel method for enhancing AD prediction by combining fMRI with single-nucleus RNA (snRNA). scBIT leverages snRNA as an auxiliary modality, significantly improving fMRI-based prediction models and providing comprehensive interpretability. It employs a sampling strategy to segment snRNA data into cell-type-specific gene networks and utilizes a self-explainable graph neural network to extract critical subgraphs. Additionally, we use demographic and genetic similarities to pair snRNA and fMRI data across individuals, enabling robust cross-modal learning. Extensive experiments validate scBIT's effectiveness in revealing intricate brain region-gene associations and enhancing diagnostic prediction accuracy. By advancing brain imaging transcriptomics to the single-cell level, scBIT sheds new light on biomarker discovery in AD research. Experimental results show that incorporating snRNA data into the scBIT model significantly boosts accuracy, improving binary classification by 3.39% and five-class classification by 26.59%. The codes were implemented in Python and have been released on GitHub (https://github.com/77YQ77/scBIT) and Zenodo (https://zenodo.org/records/11599030) with detailed instructions.

Accurate prediction of circRNA-disease associations (CDAs) is essential for elucidating disease mechanisms and identifying potential biomarkers. However, existing computational methods often struggle to fully capture multi-scale topological patterns and molecular attribute information within biological networks. To address this, we introduce MuseCDA, a novel framework designed to predict CDAs via multi-scale structure embedding. MuseCDA first leverages bidirectional transformer to generate bio-semantics from circRNA sequences, while incorporating disease similarities based on Gaussian kernel distance. It then employs random walks to capture both local neighborhood structures and global high-order network patterns, enabling comprehensive modeling of biological relationships across hierarchical levels. Finally, MuseCDA utilizes a multilayer perceptron to predict association scores. Through the seamless integration of sequence bio-semantics, Gaussian similarities, and multiscale structural features, MuseCDA achieves highly accurate CDA predictions. Evaluated on the CircR2Disease 2.0 dataset, MuseCDA demonstrates superior performance with an AUC of 0.9147, significantly surpassing baseline methods. Case studies further confirm its capability to identify novel candidates, offering valuable insights for biomarker discovery and the exploration of regulatory networks.

Single-cell RNA sequencing (scRNA-seq) technology reveals cellular heterogeneity and functional diversity, but its high dimensionality and sparsity pose challenges for analysis. Cell clustering is a crucial task in scRNA-seq data analysis. While supervised methods rely on extensive annotations, traditional unsupervised clustering methods often ignore intercellular relationships between cells. Recent works have shown that pathway-aware, multi-view graph constructions improve robustness and annotation accuracy across platforms and tissues [1] [2]. We propose scMGTC, a novel unsupervised multi-graph transformer-based cell clustering model. scMGTC integrates biological prior knowledge into representation learning by extracting gene sets from KEGG pathways and constructing distinct cell-cell graphs for each pathway. Extensive analysis conducted on six real scRNA-seq datasets demonstrates that scMGTC achieves promising performance in scRNA-seq clustering.

Cellular heterogeneity and dynamic interactions within the tumor microenvironment are critical drivers of cancer initiation and progression. Single-cell RNA sequencing, with its high-resolution capabilities, has significantly advanced the study of cellular heterogeneity in the tumor microenvironment. However, existing single-cell annotation methods are limited by data sparsity, biological heterogeneity, and batch effects, which hinder their broader application in this context. To address this, we propose scTECTA, an innovative graph neural network-based method that employs transfer learning to seamlessly transfer cell-type annotation knowledge from a well-annotated source domain to an unannotated target domain. This approach leverages graph domain adaptation, integrating novel asymmetric neural network architecture and domain-adversarial learning framework. By harnessing the generalization capabilities of graph convolutional network to correct distribution shifts and employing adversarial training to further align expression profiles across batches, scTECTA substantially enhances predictive precision and robustness. We performed a systematic evaluation across multiple datasets from diverse sources, encompassing six cancer types from 34 patients, to compare the cell-type classification performance of scTECTA against 10 benchmark methods. The results demonstrate that scTECTA markedly outperforms benchmark methods in cell-type classification and exhibits robust batch-effect correction, establishing it as an efficient and powerful tool for tumor microenvironment cell-type annotation.

MOTIVATION: A large number of studies have shown that circular RNA (circRNA) affects biological processes by competitively binding miRNA, providing a new perspective for the diagnosis, and treatment of human diseases. Therefore, exploring the potential circRNA-miRNA interactions (CMIs) is an important and urgent task at present. Although some computational methods have been tried, their performance is limited by the incompleteness of feature extraction in sparse networks and the low computational efficiency of lengthy data. RESULTS: In this paper, we proposed JSNDCMI, which combines the multi-structure feature extraction framework and Denoising Autoencoder (DAE) to meet the challenge of CMI prediction in sparse networks. In detail, JSNDCMI integrates functional similarity and local topological structure similarity in the CMI network through the multi-structure feature extraction framework, then forces the neural network to learn the robust representation of features through DAE and finally uses the Gradient Boosting Decision Tree classifier to predict the potential CMIs. JSNDCMI produces the best performance in the 5-fold cross-validation of all data sets. In the case study, seven of the top 10 CMIs with the highest score were verified in PubMed. AVAILABILITY: The data and source code can be found at https://github.com/1axin/JSNDCMI.

Identifying protein targets for drugs establishes an indispensable knowledge foundation for drug repurposing and drug development. Though expensive and time-consuming, vitro trials are widely employed to discover drug targets, and the existing relevant computational algorithms still cannot satisfy the demand for real application in drug R&D with regards to the prediction accuracy and performance efficiency, which are urgently needed to be improved. To this end, we propose here the PPAEDTI model, which uses the graph personalized propagation technique to predict drug-target interactions from the known interaction network. To evaluate the prediction performance, six benchmark datasets were used for testing with some state-of-the-art methods compared. As a result, using the 5-fold cross-validation, the proposed PPAEDTI model achieves average AUCs>90% on 5 collected datasets. We also manually checked the top-20 prediction list for 2 proteins (hsa:775 and hsa:779) and a kind of drug (D00618), and successfully confirmed 18, 17, and 20 items from the public datasets, respectively. The experimental results indicate that, given known drug-target interactions, the PPAEDTI model can provide accurate predictions for the new ones, which is anticipated to serve as a useful tool for pharmacology research. Using the proposed model that was trained with the collected datasets, we have built a computational platform that is accessible at http://120.77.11.78/PPAEDTI/ and corresponding codes and datasets are also released.