Single-cell RNA-sequencing (scRNA-seq) can excavate cellular heterogeneity and distinguish different types of cells. Clustering cells into subpopulations is essential in analyzing scRNA-seq data as it can help subsequent downstream analysis. However, scRNA-seq data are high-dimensional, sparse, and contain erroneous zero counts, which poses a great challenge for clustering. Although various methods have emerged in recent years, they cannot fully grasp the information of cells by characterizing scRNA-seq data from a single perspective, resulting in poor learned embedding representation and poor clustering performance. In this paper, we propose a multi-view clustering framework scMVAF for scRNA-seq data, which can learn more discriminative embedding representations by integrating feature information from multiple cell views. First, to comprehensively capture the data information, we generate multiple diverse views by down-sampling features, and then scMVAF learns a strong embedding representation for each cell view using an autoencoder based on a denoising zero-inflated negative binomial model. Next, to explore the correlation between cells in different views, a multi-view fusion module is introduced to fuse the embeddings from different views into a unified feature space. Concurrently, the fused embeddings are clustered to generate pseudo labels to improve the embedding process, and finally updating the embedding features and pseudo-labels in turn to obtain better clustering performance. Experiments are implemented on 16 real datasets and verify that scMVAF is superior to the other eight advanced technologies. Our code script can be obtained at https://github.com/LQXLE/scMVAF/.

To address the challenge that existing fusion models struggle to capture complex associations in Alzheimer's disease (AD) multimodal data under linear assumptions, we propose a Deep Self-Attention Enhanced Multi-Attribute Constrained Joint Non-negative Matrix Factorization (DSAE-MACJNMF) framework. This framework leverages a deep self-attention encoder to capture nonlinear features, models cross-modal dependencies using the Hilbert-Schmidt Independence Criterion (HSIC), and incorporates Laplace regularization and label alignment constraints to preserve gene co-expression topology and integrate disease-specific prior knowledge, respectively. Experiments on the ADNI dataset demonstrate that our model reduces reconstruction error by 50% and improves cross-modal correlation by 40% compared to a state-of-the-art JNMF method. Importantly, the model identifies key AD-related brain regions (e.g., Cerebellum.Crus1-L) and genes (e.g.,CAPN1, SORL1), providing novel biomarkers and a promising analytical tool for understanding AD pathogenesis and facilitating early diagnosis.

The dynamic regulatory mechanisms of DNA methylation in the evolution of brain structure and function are key factors in neuroimaging genetics research, particularly in the multi-scale pathological analysis of Early Mild Cognitive Impairment (EMCI). However, existing neuroimaging genetics studies have overlooked the time-varying associations between genes and image features under epigenetic regulation. Besides, there is a lack of quantitative modeling of gene spatial consistency and heterogeneity in degenerative patterns across brain regions. Therefore, in this study, a Temporal-Spatial Sparse Canonical Correlation Analysis (TSSCCA) model is proposed, which integrates biological network priors and spatiotemporal modulation constraints. First, dynamic ternary coupling analysis of DNA methylation, gene expression, and longitudinal imaging phenotypes in EMCI is explored. Second, the model employs Laplacian regularization to address gene spatial consistency while integrating DNA methylation with gene expression data, thereby enhancing gene-imaging association accuracy through epigenetic regulation. Finally, TSSCCA uses the$\ell_{2,1}$-norm for feature selection across multi-time points and the$\ell_{1}$-norm to identify features at specific time points, thereby achieving accurate identification of disease-related genes and imaging features. Experimental results demonstrate complex spatiotemporal relationships between gene expression and neuroimaging data modulated by DNA methylation. This approach offers a novel framework for investigating neurodegenerative disease mechanisms. The source code for TSSCCA is available at https://github.com/Jindsmu/TSSCCA. The appendix section of the article is available at https://github.com/XT2024-OO/appendix/blob/main/Appendix.pdf.

Clustering methods for scRNA-seq have advanced cellular heterogeneity analysis, but most rely only on statistical features and incorporate biological priors weakly through simple concatenation or post hoc annotation. To address this, we propose GoMC-JNMF. It builds a topic–gene prior matrix using GoM-DE from high-confidence DEGs and enforces it as a hard constraint during multi-omics factorization. The model further integrates orthogonality, sparsity, and Frobenius regularization to improve interpretability. Evaluations on eight real-world datasets show that GoMC-JNMF achieves faster convergence and higher clustering accuracy than existing methods. Biological validation confirms that it captures more meaningful cellular structures. By embedding priors directly into its mathematical framework, GoMCJNMF enables more biologically grounded single-cell analysis. Source code is available at: https://github.com/Jindsmu/GoMCJNMF.

The advancement of single-cell sequencing technology has provided unprecedented resolution for investigating cellular heterogeneity. Methods based on non-negative matrix factorization (NMF) and autoencoders are widely applied in single-cell sequencing analysis. However, current analytical models for single-cell sequencing data still face challenges such as high noise and limited applicability to specific scenarios, leading to suboptimal clustering performance. To address this issue, this study proposes an Autoencoder-like Dual-Graph Nonnegative Matrix Factorization (ADGNMF) model for single-cell multiomics analysis. The proposed method first modifies the joint NMF into an autoencoder-like architecture, followed by construction of multi-omics graph regularization and co-cluster graph regularization to enhance clustering performance and representational capability of the model. Experimental results on 8 multi-source transcriptomic datasets, 2 transcriptomic-epigenomic datasets, and 2 transcriptomic-proteomic datasets validate superior clustering performance and biological interpretability of the model. The source code of ADGNMF is available at https://github.com/jj-LanJADGNMF.

The high recurrence rate of tumor limits the growth of precision medicine, whereas the exploration of correlations in multimodal data enables mining of features linked to tumor recurrence, ultimately identifying prospective biomarkers. Nevertheless, existing multimodal approaches centered on genetic molecular data inadequately leveraged data structure and ignored the involvement of genes in the pathway or biological processes, thereby hampering interpretability of association models. In this study, a novel joint similarity nonnegative matrix factorization (JSNMF) model based on data-driven idea was proposed by adding pathway scoring data based on utilizing pathological images of tumor, gene expression data. The similarity network fusion model was applied to calculate the fusion matrices of the three-modality data with tumor recurrence as the label. Additionally, the prior information was calculated using the principal component analysis method, which was then applied to the joint nonnegative matrix factorization model with network regularization constraints. The solving efficiency of JSNMF model was enhanced by incorporating sparse orthogonality constraints on objective function. Experimental results demonstrate that incorporating prior knowledge enhances the search efficiency for joint patterns across multimodal data. The model identified recurrence-related common modules, including cellular features, genes, and pathways. Bioinformatics analysis indicated that the model can identify potential biomarkers associated with immune cell infiltration levels for recurrence diagnosis. Furthermore, the proposed method provides a new perspective for mining task-specific associations in multimodal data. This study also improves understanding of association patterns among genetic molecular features linked to tumor recurrence.

Landslides in mountainous regions exhibit widespread distribution and high recurrence, posing critical threats to human safety and socioeconomic assets. Interferometric Synthetic Aperture Radar (InSAR) has proven instrumental in landslide identification, continuous monitoring, and early-warning systems. Nevertheless, the inherent side-looking geometry of SAR systems restricts displacement measurements to the line-of-sight (LOS) direction, which may misrepresent true slope-parallel displacement and cause substantial inconsistencies between ascending/descending track-derived results and actual ground movement. To address this limitation, this study introduces the concept of LOS sensitivity (LS) and systematically quantifies geometric distortion effects on LOS measurements across diverse slope gradients and aspects. We propose an innovative LOS Sensitivity Correction (LSC) framework that integrates geometric distortion correction through joint optimization of topographic parameters and SAR interferometric processing configurations, enabling precise retrieval of slope-parallel displacement from single-track InSAR observations. Validation experiments utilizing Sentinel-1 ascending/descending datasets over the Ermulin and Huanglianping landslides demonstrate that while conventional D-InSAR results exhibit significant inter-track discrepancies (exceeding 40 mm in magnitude), LSC-processed displacements achieve remarkable consistency in deformation patterns (correlation coefficient >0.92) and spatial distribution. We further analyzed the influence of topographic parameters and interferometric processing configurations on LSC, revealing their effects on the correction results. The proposed methodology comprehensively resolves LOS sensitivity challenges for slopes with arbitrary geometries, advancing the fundamental understanding of terrain-dependent SAR detection limitations while transcending the simplistic formulaic application of conventional correction frameworks. This work establishes a paradigm for robust interpretation of slope-parallel kinematics from single-track InSAR products, with direct implications for landslide hazard assessment and mitigation strategies in complex terrains.

An automatic vision-based sewer inspection plays a vital role of sewage system in a modern city. Recent advances focus on modeling a deep learning-based method to realize the sewer inspection system, benefiting from the capability of data-driven feature extraction. Although the acceptable performances of sewer defect classification are achieved, there is still a gap between the emerged methods and actual application scenarios. The first issue is that the multi-focus complementarity is ignored to represent the sewer defect, resulting in capturing the multi-scale information of sewer defect inefficiently. Second, the inherent uncertainty of sewer defect is not considered, while the serious unknown sewer defect categories would be missed, resulting in the untrustworthy sewer inspection. In this paper, we focus on quick-view (QV)-based sewer inspection, while a trustworthy multi-focus fusion framework (TMFF) is proposed, jointly combining multi-label classification and uncertainty estimation. Specifically, focal segment module (FSM) is designed based on optical flow to split the QV sewer video into long-focus and short-focus segments, where the multi-focus segments can be modeled to represent the multi-scale information of sewer defect. Then, evidential deep learning (EDL) is introduced to quantify the uncertainty, while joint expert scheme (JES) is designed to aggregate the expert opinions of multi-focus segments. Moreover, evidential disambiguating strategy (EDS) is proposed to alleviate the ambiguity of uncertainty estimation. Extensive experiments are conducted on VideoPipe, in which the superiority of TMFF is demonstrated compared with the state-of-the-art methods. Furthermore, we validate the potential capability of TMFF against the unknown cases of sewer defects.

Judging and identifying biological activities and biomarkers inside tissues from imaging features of diseases is challenging, so correlating pathological image data with genes inside organisms is of great significance for clinical diagnosis. This paper proposes a high-dimensional kernel non-negative matrix factorization (NMF) method based on muti-modal information fusion. This algorithm can project RNA gene expression data and pathological images (WSI) into a common feature space, where the heterogeneous variables with the largest coefficient in the same projection direction form a co-module. In addition, the miRNA-mRNA and miRNA-lncRNA interaction networks in the ceRNA network are added to the algorithm as a priori information to explore the relationship between the images and the internal activities of the gene. Furthermore, the radial basis kernel function is used to calculate the feature proportion between different kinds of genes mapped in the high-dimensional feature space and projected into the common feature space to explore the gene interaction in the high-dimensional situation. The original feature matrix is regularized to improve biological correlation, and the feature factors are sparse by orthogonal constraints to reduce redundancy. Experimental results show that the proposed NMF method is better than the traditional NMF method in stability, decomposition accuracy, and robustness. Through data analysis applied to lung cancer, genes related to tissue morphology are found, such as COL7A1, CENPF and BIRC5. In addition, gene pairs with a correlation degree exceeding 0.8 are found, and potential biomarkers of significant correlation with survival are obtained such as CAPN8. It has potential application value for the clinical diagnosis of lung cancer.

The maximum detectable deformation gradients (MDDG) for interferometric synthetic aperture radar (InSAR) technology is important for the selection of SAR images and processing algorithms to perform accurate slope displacement monitoring, which is strongly influenced by terrain factors in mountainous areas. In this paper, a functional model is proposed to determine the MDDG of InSAR with respect to arbitrary slope gradients/aspects and wavelengths. Based on this model, regional MDDG characteristics are explored and compared in Mao County, Sichuan Province, China. The MDDG distribution regarding on Sentinel-1, ALOS-2/PALSAR-2 and TerraSAR-X SAR satellite data using arbitrary slope gradient/aspect are derived. Furthermore, the MDDG from variable satellites for three different bands (X/C/L-band) are compared and the influence factors with respect to the wavelength and resolution on MDDG are discussed. The proposed model is helpful in selecting of SAR data or processing algorithms based on calculated MDDG, in the meanwhile, it has significant implications on the understanding and analyzing real slope displacement monitored by InSAR regarding on different SAR images in mountainous areas.

Circular ribonucleic acids (RNAs) (circRNAs) are formed by covalently linking the downstream splice donor and the upstream splice acceptor. One of the most important functions of circRNAs is mainly exerted through binding RNA-binding proteins (RBPs). However, there is no efficient algorithm for identifying genome-wide circRNA-RBP interactions. Here, we developed a unique algorithm, circRIP, for identifying circRNA-RBP interactions from RNA immunoprecipitation sequencing (RIP-Seq) data. A simulation test demonstrated the sensitivity and specificity of circRIP. By applying circRIP, we identified 95 IGF2BP3-binding circRNAs based on the IGF2BP3 RIP-Seq dataset. We further identified 2823 and 1333 circRNAs binding to >100 RBPs in K562 and HepG2 cell lines, respectively, based on enhanced cross-linking immunoprecipitation (eCLIP) data, demonstrating the significance to survey the potential interactions between circRNAs and RBPs. In this study, we provide an accurate and sensitive tool, circRIP (https://github.com/bioinfolabwhu/circRIP), to systematically identify RBP and circRNA interactions from RIP-Seq and eCLIP data, which can significantly benefit the research community for the functional exploration of circRNAs.

In the context of global warming, monitoring glacial lakes is of great significance for revealing climate change and mitigating glacier lake outburst floods (GLOFs). Multitemporal mapping of glacier lakes through optical satellite remote-sensing imageries can only provide us comparative analysis of glacier lake changes during long time intervals. In recent years, the rapid development of synthetic aperture radar (SAR) has made it possible to observe seasonal changes of glacial lakes. This letter proposed a polarization enhancement-based maximum interclass variance (PE-MIV) method to accurately and automatically mapping the seasonal cycles of glacial lakes based on multipolarization SAR data. Taking the Gongcuo and Langcuo glacial lakes in Laigu village, Tibetan Plateau, as examples, based on 113 Sentinel-1 SAR imageries covering 2017 to 2020 with a 12-day revisiting time, the seasonal boundary changes of both lakes were extracted automatically with a relative accuracy of 93.59%. The seasonal variations of glacial lake area are closely related to temperature and precipitation. This letter demonstrated that the satellite SAR (short revisiting time, multipolarization, etc.) is an effective tool to monitor areal changes of glacier lakes, which can be widely applied in detecting glacier lakes, GLOFs, and their response to climate change.