Your browser does not allow saving data (e.g. private mode): bookmarks will be lost when the tab closes.
Search PubMed
Keywords, author, journal, years — results from PubMed, abstracts with Russian translation, bookmarks in your browser.
Found: 4 882
Page 37 of 489-
ProtoCloud: A prototypical self-explaining model for single-cell analysis.
Cell type annotation is a fundamental task in single-cell genomics. Although various methods have been developed for automatic annotation, they often function as black-box models lacking explainability, proper uncertainty estimation, and robustness for rare cell types. We introduce ProtoCloud, a self-explanatory deep generative model that embeds cells into a structured, low-dimensional space organized around cell-type-specific prototypes. ProtoCloud matches or outperforms existing methods across 11 large-scale datasets, particularly for rare cell types. Its built-in uncertainty quantification mechanism, based on cell-prototype similarity, identifies and re-annotates misannotated training cells. By backpropagating cell prototype similarities to the gene space, ProtoCloud identifies key genes driving its classifications, facilitating the discovery of both known and novel marker genes. Applied to a time-course dataset of post-injury retinal neurons, ProtoCloud successfully annotates previously unassigned cells; in an esophageal cell atlas, it identifies rare but potentially important cell populations and their marker genes associated with esophageal inflammation.
-
Systematic discovery of retina-enriched Rik genes identifies 1190005I06Rik as a novel modulator of visual signalling.
BACKGROUND: High‑throughput transcriptome projects have revealed thousands of mammalian genes with little or no functional annotation. Among these are hundreds of loci assigned provisional “Rik” identifiers following discovery in the RIKEN cDNA annotation effort. Although often dismissed as genomic dark matter, such genes may encode tissue‑restricted proteins that modulate physiologic functions and influence disease. The retina is a highly specialised neural tissue and a common site of inherited disorders; understanding its molecular repertoire could illuminate novel therapeutic avenues. METHODS: We integrated bulk RNA‑seq from ten adult mouse tissues, evolutionary and domain analysis, single‑cell RNA‑seq, and CRISPR/Cas9 gene disruption to systematically catalogue protein‑coding Rik genes enriched in the retina and test the function of a representative gene. RESULTS: A rigorous differential expression analysis identified 44 Rik genes with robust retina‑specific expression compared with nine non‑retinal tissues. Many of these genes lack orthologues beyond rodents, while others show broad conservation, illustrating a continuum from lineage‑restricted to conserved retinopathy candidates. Single‑cell transcriptomics revealed that these genes are expressed across retinal cell types, with the highest aggregate expression in cone photoreceptors and inner interneurons. To evaluate physiological significance, we generated a 1190005I06Rik knockout mouse. Although retinal architecture appeared normal, loss of 1190005I06Rik enhanced electroretinogram b‑wave amplitudes and altered light‑avoidance behaviour, indicating that this previously uncharacterised gene acts as a negative modulator of visual signalling. CONCLUSIONS: We present a curated atlas of retina‑enriched Rik genes and demonstrate that 1190005I06RIK modulates retinal circuit function. This resource expands the molecular landscape of the retina and provides new candidates for the genetic basis of inherited retinal disease. Our findings underscore that unannotated genes may exert measurable effects on sensory processing and warrant systematic exploration in the context of human ocular disorders.
-
Integrating Multi-Omics Atlas to Uncover Genetic and Epigenetic Mechanisms and Reveal Cell State Evolution Across Ecotypes in Male Urological Cancers.
Male urological cancers, including clear cell renal cell carcinoma (ccRCC), bladder cancer (BC), and prostate cancer (PCa), are characterized by extensive heterogeneity and complex ecosystems, yet the underlying mechanisms remain incompletely understood. In this study, scRNA-seq, scATAC-seq and spatial transcriptomics data are integrated to systematically characterize the features of the tumor microenvironment (TME). We identify tumor cell subclones and elucidate the impact of chromosomal abnormalities on their characteristic functions. We further identify transcription factor regulatory networks by analyzing tumor cell differentiation trajectories. Importantly, we integrate DNA methylation and SNP information to deeply dissect the tumor cell differentiation process, revealing the multilayer regulatory mechanisms of tumor-related genes. Additionally, we reveal the evolution of cellular states across ecotypes to provide a more comprehensive characterization of TME. Finally, we screened potential therapeutic agents targeting the molecular mechanisms underlying tumor cell differentiation (Amivantamab in ccRCC, Levothyroxine in BC, Ouabain in PCa) and signature genes of ecotype. In conclusion, our work establishes a comprehensive framework for tumor assessment and informs the development of precision therapeutic strategies.
-
BiGCN Learns B Cell Functional States by Integrating Single-Cell Transcriptomes and BCR Repertoires.
B lymphocytes are pivotal players in immune regulation and disease progression, yet their full functional diversity has remained elusive due to the siloed analysis of single-cell transcriptomes and B cell receptor (BCR) repertoires. Here, we present BiGCN, a novel method that explicitly integrates transcriptomic and BCR data at single-cell resolution. By uniting these two modalities within a unified embedding using a graph convolutional network (GCN) framework, BiGCN enables high-fidelity characterization of B cell states and clonal relationships. Applying BiGCN across a wide spectrum of paired single - cell RNA sequencing (scRNA-seq) and single -cell BCR sequencing (scBCR-seq) datasets, including infectious disease, autoimmunity, and a comprehensive pan-cancer atlas, we demonstrate that BiGCN markedly outperforms existing approaches in information integration, functional annotation, and biological insight. BiGCN not only recapitulates the trajectory from B cell maturation to immunoglobulin class switching but also discovers intermediate B cell subtypes and reveals functional states that are otherwise obscured by conventional analyses. Our results establish BiGCN as a versatile and robust platform for B cells research, accelerating discoveries in immunology, cancer biology, and systems medicine.
-
Single-cell profiling decodes patagium development in gliding mammal.
The gliding patagium represents a key adaptation for mammalian flight, but its cellular development remains unexplored. Using single-nucleus RNA sequencing of embryonic flying squirrel patagium and dorsal skin, we construct a single-cell atlas of patagium development and identify two distinct fibroblast subpopulations (Fp2 and Fr) highly enriched in the patagium. These fibroblasts are characterized by the patagium upregulation of Wnt5a, Fgf7, and Fgf10, and are associated with patagium morphogenesis through dermal-epidermal putative communication interactions between dermal fibroblasts (Fp2 and Fr) and epithelial basal keratinocytes. Specifically, Fp2 fibroblasts are potentially involved in distal dermal condensation and epithelial thickening together with elevated Wnt5a expression, while both Fp2 and Fr fibroblasts could play a role in epithelial polarization and thickening through Fgf7 and Fgf10, as suggested by ex vivo assays. Our data suggest that gliding patagium development results from the co-option of conserved WNT and FGF signaling pathways within a specialized fibroblast-epithelial context, illustrating how modifications of conserved developmental programs give rise to derived morphological traits.
-
Evaluating genetic ancestry inference from single-cell transcriptomic datasets.
Characterizing the ancestry of donors in single-cell transcriptomic studies is crucial to ensure genetic homogeneity, reduce biases in analyses, identify ancestry-specific regulatory mechanisms and their downstream roles in disease, and ensure that existing datasets are representative of human genetic diversity. While these datasets are now widely available, information on the ancestry of donors is often missing, hindering further analysis. Here, we propose a framework to evaluate methods for inferring genetic ancestry from genetic polymorphisms detected in single-cell sequencing reads. We demonstrate that widely used tools (e.g., ADMIXTURE) provide accurate inference of genetic ancestry and admixture proportions, despite the limited number of genetic polymorphisms identified and imperfect variant calling from sequencing reads. We infer genetic ancestry for 401 donors from 10 Human Cell Atlas datasets and report a high proportion of donors of European ancestry in this resource. For researchers generating single-cell transcriptomic datasets, we recommend reporting genetic ancestry inference for all donors and generating datasets that represent diverse ancestries.
-
Multi-organ single-cell transcriptomic atlas identifies QrIAA14 as a candidate negative regulator of adventitious root development in Quercus robur.
Plant organ development involves coordinated cell fate transitions across multiple tissues, yet the cellular programs underlying organ-specific differentiation in woody plants remain poorly understood, particularly the mechanisms limiting efficient root development during vegetative propagation of oak species. Here, we generated a comprehensive single-cell transcriptomic landscape of leaf, stem, and root tissues of Quercus robur to resolve developmental trajectories at cellular resolution. A total of 41,471 high-quality cells were classified into 30 distinct clusters, enabling the identification of major cell types and organ-specific transcriptional features across three vegetative organs. Pseudotime analyses exhibited the developmental programs related to guard cell differentiation in leaves, vascular formation in stems, and root tissue development. Additionally, combining scRNA-seq, bulk transcriptome profiling, and phytohormone investigations, we identified auxin signaling as an important regulator during adventitious root development process. Notably, QrIAA14-1, an IAA14 homolog, was preferential enrichment in root hair, near-root hair cells and root cap along root developmental trajectories, which was further supported by RT-qPCR and in situ hybridization assays. Furthermore, the overexpression of QrIAA14-1 significantly inhibited oak root elongation, resulting in around 64.46% reduction in adventitious root length compared with control plants, providing mechanistic insight into the limitations of root development in oak. Together, this study provides the first high-resolution single-cell atlas of cellular organization and developmental dynamics across oak vegetative organs and identifies the candidate regulator genes associated with root development, offering new insights into the regulatory mechanisms of woody plant root regeneration and clonal propagation.
-
Based on scRNA-seq and bulk RNA-seq to establish tumor immune microenvironment-associated signature of skin melanoma and predict immunotherapy response.
Skin cutaneous melanoma (SKCM), a form of skin cancer, ranks among the most formidable and lethal malignancies. Exploring tumor microenvironment (TME)-based prognostic indicators would help improve the efficacy of immunotherapy for SKCM patients. This study analyzed SKCM scRNA-seq data to cluster non-malignant cells that could be used to explore the TME into nine immune/stromal cell types, including B cells, CD4 T cells, CD8 T cells, dendritic cells, endothelial cells, Fibroblasts, macrophages, neurons, and natural killer (NK) cells. Using data from The Cancer Genome Atlas (TCGA), we employed SKCM expression profiling to identify differentially expressed immune-associated genes (DEIAGs), which were then incorporated into weighted gene co-expression network analysis (WGCNA) to investigate TME-associated hub genes. Discover candidate small molecule drugs based on pivotal genes. Tumor immune microenvironment-associated genes (TIMAGs) for constructing TIMAS were identified and validated. Finally, the characteristics of TIAMS subgroups and the ability of TIMAS to predict immunotherapy outcomes were analyzed. We identified five TIMAGs (CD86, CD80, SEMA4D, C1QA, and IRF1) and used them to construct TIMAS. In addition, five potential SKCM drugs were identified. The results showed that TIMAS-low patients were associated with immune-related signaling pathways, high MUC16 mutation frequency, high T cell infiltration, and M1 macrophages, and were more favorable for immunotherapy. Collectively, TIMAS constructed by comprehensive analysis of scRNA-seq and bulk RNA-seq data is a promising marker for predicting ICI treatment outcomes and improving individualized therapy for SKCM patients.
-
Domain adaptation for supervised integration of scRNA-seq data.
Large-scale scRNA-seq studies typically generate data in batches, which often induce nontrivial batch effects that need to be corrected. Given the global efforts for building cell atlases and the increasing number of annotated scRNA-seq datasets accumulated, we propose a supervised strategy for scRNA-seq data integration called SIDA (Supervised Integration using Domain Adaptation), which uses the cell type annotations to guide the integration of diverse batches. The supervised strategy is based on domain adaptation that was initially proposed in the computer vision field. We demonstrate that SIDA is able to generate comprehensive reference datasets that lead to improved accuracy in automated cell type mapping analyses.
-
Chemical induction enhances patient-derived organoid fidelity to primary colorectal cancer at the single-cell level: A re-analysis of public scRNA-seq dataset GSE261012.
BACKGROUND: Patient-derived organoids (PDOs) are widely used preclinical models for colorectal cancer (CRC). However, conventional PDO cultures frequently lose tumor microenvironment (TME) components and undergo culture-induced transcriptional drift. A chemically defined induction system (CiPDO) has been proposed to preserve fetal-like plasticity states, yet a direct single-cell comparison of CiPDO versus conventional PDO against the matched primary tumor is lacking. OBJECTIVE: To evaluate and compare the single-cell transcriptional fidelity of chemically induced PDO (CiPDO) and conventional PDO (current_PDO) relative to the matched primary colorectal tumor, using a publicly available scRNA-seq dataset. METHODS: We re-analyzed the public scRNA-seq dataset GSE261012, which profiled cells from the primary tumor, conventional PDO (current_PDO), and chemically induced PDO (CiPDO) of the same CRC patient. Using Scanpy, we performed quality control, clustering, cell type annotation, and module scoring for oncofetal-like state (OnFS), proliferation, stress/hypoxia, epithelial-mesenchymal transition (EMT), and differentiation signatures. All possible pairwise comparisons among the three conditions were performed using two-sided Mann-Whitney U tests. For each module score, we provided biological interpretation of the score magnitude, reference ranges from established CRC single-cell atlases, and contextualization of the observed values. RESULTS: After quality control, 16,249 cells were retained across 19 clusters. Both PDOs completely lost TME components (fibroblasts, myeloid, T cells) compared with the primary tumor. At the epithelial-cell level, CiPDO showed significantly higher OnFS scores than current_PDO (0.189 vs. 0.102, p < 0.001) and both PDOs exhibited significantly higher OnFS than the primary tumor (-0.232; both p < 0.001). Proliferation scores were highest in the primary tumor (1.461), intermediate in current_PDO (0.983), and lowest in CiPDO (-0.314); all pairwise differences were significant (p < 0.001) except primary tumor vs. current_PDO (p = 0.082). Stress/hypoxia scores were significantly elevated in current_PDO (0.286) compared with both CiPDO (0.052, p < 0.001) and the primary tumor (0.195, p = 0.041). Differentiation scores did not differ significantly between the two PDO conditions (p = 0.171), whereas both PDOs showed significantly higher differentiation than the primary tumor (both p < 0.001). EMT scores showed a small but statistically significant difference between CiPDO (-0.078) and current_PDO (-0.098, p < 0.001), and both were significantly higher than the primary tumor (-0.128; both p < 0.001). CONCLUSION: Within the constraints of a single-patient design, CiPDO better preserves the oncofetal-like plasticity state of primary CRC epithelial cells while exhibiting fewer culture-induced proliferation and stress signatures than conventional PDO. Differentiation states were comparable between the two PDO conditions. These findings support the use of chemically defined induction to improve PDO fidelity for studying tumor cell plasticity, although validation in larger cohorts is required.