Yunpeng Zhang

Breast cancer progression is closely linked to immune cells within the tumor microenvironment (TME), but the immune landscape differences across HER2-positive (HER2+), triple-negative (TNBC), and ER-positive (ER+) subtypes have yet to be systematically characterized. In this study, we characterized immune landscape differences across HER2-positive (HER2+), triple-negative (TNBC), and ER-positive (ER+) breast cancers using single-cell RNA sequencing data, focusing on T cells, macrophages, and B cells. Our analysis revealed that cytotoxic CD8 + T cells were more abundant in TNBC but showed signs of dysfunction. Macrophage subsets, including an NDUFB8 + population, exhibited divergent functional orientations across subtypes. B cells in TNBC exhibited functional alterations, characterized by a decreased proportion of plasmablasts and an increased proportion of memory B cells. Single-cell regulatory network inference and clustering (SCENIC) analysis predicted a set of subtype-specific candidate transcription factors (TFs), including E2F8 across all subtypes and ATF7 in HER2 + tumors, whose activity correlated with clinical outcomes. Finally, mapping multicellular communication networks revealed both conserved and subtype-specific signaling pathways. Together, these results present a detailed transcriptional landscape of immune heterogeneity in breast cancer and contribute novel insights that may inform the design of subtype-specific immunotherapies.

The updated CellMarker 3.0 resource (http://bio-bigdata.hrbmu.edu.cn/CellMarker) offers a comprehensive resource, integrating experimentally validated and robustly algorithm-predicted marker genes for diverse cell types across human and mouse tissues. In addition, web tools for analyzing single-cell RNA-seq (scRNA-seq) and spatial transcriptomics (ST) data are described. CellMarker 3.0 was updated with more data and new features, including (i) appending 274 tissues, 1877 cell types, 64 040 markers, and 1 245 923 tissue–cell type–marker entries over the previous version. The current release contains 930 tissues, 4455 cell types, and 90 955 cell markers, resulting in a total of 1 329 284 tissue–cell type–marker entries. (ii) Expanding the resource to include computationally predicted cell-type markers derived from six algorithms, namely Spapros, Seurat::FindAllMarkers, SEMITONES, Cepo, CelliD, and COSG. (iii) New marker information was incorporated from 168 sequencing technologies, such as CEL-Seq2 and SNARE-seq. (iv) Covering 35 types of cell markers, including protein–coding genes, rRNAs, and pseudogenes. (v) The cell tool module was comprehensively restructured and upgraded, with the original six scRNA-seq analysis modules systematically optimized and two new spatial transcriptomics modules, Spatial Expression, and Spatial Chat, added. CellMarker 3.0 is a valuable resource that integrates cell identity identification, single-cell transcriptomics, and spatial transcriptomics analysis.

The spatial microenvironment and cellular functional states are governed by spatiotemporally dynamic gene regulatory networks (GRNs), making their dissection essential for understanding disease pathogenesis and normal tissue programs. Despite the rapid development of multiple spatial transcriptomics databases, most lack systematic integration of intracellular GRNs and their spatial coupling patterns. Here, we present SpatialRegNet (https://bio-bigdata.hrbmu.edu.cn/SpatialRegNet/), a comprehensive database curating 1322 tissue slices derived from 152 human and mouse spatial transcriptomic datasets that span 68 tissue types and 93 distinct diseases or disease subtypes, encompassing 3 744 216 spatial spots. The database quantifies spot-level activity for 6733 transcription factor (TF)-mediated regulon networks, which collectively contain over 12.7 million TF–target gene pairs, and archives 802 935 regulon networks that are highly active in distinct spatial niches/cell types. SpatialRegNet embeds a suite of interactive online tools for analyzing regulon networks in spatial contexts, enabling users to interrogate the correlations between regulon network activity and cellular identities, map spatial regulatory heterogeneity and specificity, track dynamic regulon network activity across cell differentiation trajectories, reconstruct intercellular communication signaling cascades, and generate three-dimensional visualizations of regulon network activity. Collectively, SpatialRegNet constitutes a valuable public resource to advance mechanistic investigations of physiological tissue function, disease etiology, and biomarker discovery in spatially resolved GRNs.

The tumor ecosystem is organized into diverse ecological niches that undergo continuous regional remodeling and spatiotemporal evolution during tumor progression. Systematically characterizing the regional dynamics of these niches and reconstructing their spatiotemporal trajectories are essential for understanding tumor development, metastasis, and therapeutic response. Here, we present EvoNiche (https://bio-bigdata.hrbmu.edu.cn/EvoNiche or http://evoniche.bio-database.com/), a comprehensive database designed to systematically dissect the regional dynamics and spatiotemporal patterns of the tumor ecosystem by integrating 91 matched spatial and single-cell transcriptomic datasets spanning 35 cancer types. EvoNiche incorporates 712 tissue sections comprising 1 953 708 spatial spots and 58 single-cell datasets containing 2 945 536 cells, identifies 5609 spatial niches, and reconstructs tissue-level and cell-level spatiotemporal trajectories across tumor ecosystems. Through integrated analytical frameworks, EvoNiche enables comprehensive characterization of tumor ecosystem remodeling by resolving spatially organized cellular compositions, region-specific and trajectory-associated molecular programs and biological functions, spatially proximal cell–cell interactions, and spatial pseudotime-associated changes. Furthermore, EvoNiche prioritizes region-specific and transition genes as spatially derived candidate genes through systematic evaluation across pan-cancer and immunotherapy cohorts. Collectively, EvoNiche serves as a valuable resource for investigating regional ecosystem organization and spatiotemporal patterns across human cancers, and elucidating the regulatory mechanisms underlying tumor progression and therapeutic response.

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder driven by complex cellular changes. To identify transcriptional signatures involved in AD pathology, we first analyzed nine single-cell/nucleus RNA-seq datasets from 67 high-pathology cases (Braak stages V/VI), generating an atlas of 363 243 cells. Using a multi-sample integration strategy, we identified 51 cross-sample meta-programs (MPs) across seven cell types that collectively capture key AD-related processes, including synaptic dysfunction and neuroinflammation. By screening bulk transcriptomes from 1208 ad and 725 normal samples, we found that 26 MPs (13 pathogenic and 13 protective) displayed differential activity in AD, characterized by up-regulation of microglia MPs and down-regulated of neuron MPs. Furthermore, spatial transcriptomics revealed that these MPs form spatially coherent communities associated with distinct cellular neighborhoods. Finally, we performed an integrative drug repurposing screen to identify candidate drugs predicted to regulate pathogenic MPs. In conclusion, we conducted an integrated multi-omics study to identify AD cell-type-specific MPs, and this framework can be applied to deconvolve cellular heterogeneity and screen candidate drugs for AD.

Metabolic diseases are a substantial global health burden. However, how different metabolic pathways are connected to the clinical relevance in metabolic diseases remains unclear. Here, by integrating previously published single‐cell RNA sequencing datasets, a single‐cell metabolic atlas including 562 699 cells from 86 patients across seven major metabolic diseases is compiled. An advanced integrative analytical strategy to assess the activity of 227 metabolic pathways is developed. It is identified 89 disease‐enriched pathways. These pathways are further organized into 15 cooperative metabolic modules, revealing metabolic module remodeling associated with coordinated pathway activities and underlying regulatory mechanisms. By linking module activities to transcription factor dynamics and ligand–receptor signaling, cross‐cell‐type regulatory circuits underlying module‐specific remodeling and the identification of 63 potential therapeutic targets are delineated. Furthermore, an open‐access resource is developed to enable in‐depth exploration of the results in this study. In summary, this study illustrates conserved principles of modular metabolic remodeling across metabolic diseases and sheds light on therapeutic target discovery.

Clear cell renal cell carcinoma (ccRCC) exhibits pronounced heterogeneity across WHO histological grades, yet systematic single-cell multi-omics studies characterizing these transitions remain limited. We integrated scRNA-seq and scATAC-seq data across ccRCC WHO grades to establish a multi-omics framework encompassing tumor cells and immune populations. Using pseudotime trajectory analysis and machine learning ensembles, we developed a prognostic signature (CBG) from core nodes of transcriptional regulatory networks. We found that in tumor cells, epigenetic alterations consistently precede metabolic reprogramming and invasive adaptation. CD8+ T cell exhaustion followed a trajectory shifting from IRF7- to ZNF683-regulated states, while monocytes differentiated toward M1 and M2 macrophages orchestrated by NFIC/IL1B and CEBPD/GLI2. Intercellular communication networks showed a temporal progression from inflammation, through vascular remodeling, to immunosuppression dominance. The CBG signature demonstrated robust performance in independent cohorts, identifying SLC11A1 and SH3YL1 as antagonistic survival determinants. This study elucidates the dynamic molecular and immunological mechanisms underlying ccRCC grade progression, providing a robust framework for subtype-specific prognostication and precision therapeutic targeting

Epilepsy is mainly characterized by spontaneous seizures caused by hyperactive neural circuits. To delineate the cell-type-specific mechanisms underlying neuronal hyperexcitability, we resolve the hyperexcitability of excitatory neurons across epileptic human brain trans-foci at single-cell resolution to identify the key drivers and potential diagnostic signatures. We constructed a comprehensive atlas encompassing 240,000 cells derived from the temporal cortex and hippocampus, detecting trans-regional cellular and molecular diversity. We further delineated dynamic trajectories, gene expression patterns, and functional reorganization across cell types. Using the LASSO and random forest algorithms, we prioritized the core genes and developed a logistic regression-based diagnostic model. Despite transregional cellular landscape conservation, major cell types varied in abundance. Detailed analysis delineated various excitatory neuron subtypes’ dynamic trajectories, intricate expression, and functional reorganization, with pronounced dysfunction in the posterior hippocampal and temporal cortex networks, indicating hyperactive pro-epileptic effects. Excitatory neurons exhibit an intrinsic ability to autonomously organize themselves into distinct, highly active modules, characterized by a high activation state during epileptogenesis, as illustrated by ten epilepsy-associated functions. Transcription circuits FOSL2/FOS/EGR3/EGR1 promote neuronal hyperexcitability. Integrating epilepsy bulk RNA-seq data, we identified 24 overlapping genes between differential genes and circuit targets. The LASSO and random forest algorithms prioritized three core genes (IL1B, SOCS6, and COL4A1). A logistic regression model based on these three genes showed variable performance, with an apparent AUC of 1.000 in the discovery cohort (GSE256068) and AUCs of 0.974 and 0.722 in and two validation cohorts, indicating the need for further validation. Our study establishes the FOSL2/FOS/EGR3/EGR1 circuit as a master regulator of pathological neuronal hyperactivity across epileptic foci, linking transcriptional activation to network dysfunction. Identifying overactive factors may represent a candidate molecular pathway for future therapeutic exploration against hyperexcitability.

Esophageal squamous cell carcinoma (ESCC) progression involves dynamic cellular state transitions and tumor microenvironment remodeling, accompanied by extensive transcriptional regulation reprogramming. Here, we systematically mapped the TF-mediated regulatory landscape underlying ESCC progression at single-cell resolution by integrating stage-specific ESCC single-cell transcriptomic datasets comprising over 200,000 cells with TF–target interaction networks. Using a random walk algorithm combined with hypergeometric testing, we identified malignant progression-associated TFs (mpTFs) across multiple cell types and disease stages. Our analysis revealed extensive stage-dependent regulatory remodeling during ESCC progression. TCF4 was identified as an early-stage regulator associated with epithelial–mesenchymal transition activation and malignant invasive phenotypes. In immune lineages, BATF and IRF4 exhibited trajectory-associated activation during CD4⁺ T-cell differentiation and CD8⁺ T-cell exhaustion, suggesting critical roles in immunosuppressive T-cell state transitions. Additionally, mpTF-mediated remodeling of M2 macrophage subpopulations contributed to immunosuppressive tumor microenvironment formation during advanced ESCC progression. We further identified prognosis-associated cell-type-specific and shared mpTFs, including TFAP2C, which was associated with stabilized fibroblast and monocyte functional states and a less aggressive tumor microenvironment phenotype. Collectively, this study provides a comprehensive single-cell atlas of TF-mediated regulatory programs during ESCC progression and offers potential therapeutic targets for precision oncology.