Xiao Dong

Cellular senescence is a heterogeneous cell state induced by diverse stressors, including telomere attrition, genotoxic agents, oxidative damage, and inflammation. Despite ongoing efforts to identify conserved senescence biomarkers, it remains unclear whether senescence-inducing stimuli converge at the level of individual genes or broader molecular processes. Here, we profiled transcriptomic changes in human primary lung fibroblasts (IMR-90) driven toward senescence by replicative exhaustion, bleomycin, H ₂ O ₂ , or ionizing radiation under matched, dose- or time-resolved conditions. Across all four senescent inducers, global transcriptomic variation aligned along a shared axis of senescence progression, consistent with established machine learning-based senescence classifiers. However, overlap at the level of individual genes was limited, with most responses being inducer-specific or only partially conserved. In contrast, pathway-level analysis revealed far more consistent enrichment across all conditions, including downregulation of proliferation-associated pathways and activation of stress-related and pro-inflammatory pathways, accompanied by distinct inducer-specific patterns. These results support a hierarchical organization of the senescent transcriptome, in which diverse senescence inducers converge on shared pathway-level features while maintaining gene-level heterogeneity. These results provide a foundational basis for interpreting senescence signatures and may facilitate the development of more robust transcriptome-based markers of cellular senescence in aging and disease.

Aging involves progressive physiological decline, yet the underlying transcriptomic patterns remain poorly understood. Although differentially expressed genes (DEGs) have been the primary focus of previous studies, here we investigate differentially variable genes (DVGs) using a novel Gene Stability Score (GSS). In 30 tissue types from nearly 1,000 individuals in the Genotype-Tissue Expression (GTEx) project, age- and sex-related DVGs account for approximately 15% of overall expression variability between samples of the same tissue, with age-related DVGs specifically contributing 7.7%. We further show that DEGs and DVGs affect distinct biological pathways, and that inter-individual instability is significantly correlated with cell-to-cell transcriptional noise. Moreover, gene regulatory network analysis reveals that this variability is not random but is shaped by local network architecture. Finally, we identify robust reference genes, including TBP, PUM1 , and TMEM199 , for RT-qPCR experiments in studying age-related gene expression changes in humans. Together, our findings suggest that aging involves both coordinated transcriptional programs and increased stochasticity across individuals and cells.

Single-cell RNA sequencing (scRNA-seq) has become a key technology for characterizing cell-type heterogeneity in complex tissues. However, its utility depends on accurate and reproducible cell-type annotation, which remains a major analytical challenge. Although hundreds of computational tools have been developed for automated annotation, there is currently no systematic framework to evaluate annotation robustness in a dataset-specific manner or within the context of complete analytical pipelines. Here, we present STEVE (Single-cell Transcriptomics Expression Visualization and Evaluation), a quantitative framework designed to assess the accuracy, robustness, and reproducibility of cell-type annotation in scRNA-seq studies. STEVE implements three complementary in silico evaluation modules: (i) Subsampling Evaluation to quantify annotation stability under varying reference sizes and data partitions; (ii) Novel Cell Evaluation to assess the ability to detect previously unseen cell types; and (iii) Annotation Benchmarking to compare alternative annotation tools against ground-truth labels. In addition, STEVE includes a Reference Transfer Annotation module that enables cross-dataset cell-type mapping using external reference datasets. All modules are built upon a unified probabilistic framework that provides consistent confidence estimation across evaluation scenarios. We evaluated STEVE across four independent scRNA-seq datasets with experimentally defined or expert-curated cell-type labels. Our results show that annotation robustness is strongly influenced by the annotation method, biological separability, dataset complexity, and batch effects. STEVE provides a practical framework for quantifying annotation uncertainty and improving reproducibility in single-cell transcriptomic analyses. STEVE is freely available at GitHub ( https://github.com/XiaoDongLab/STEVE ).

Background: The adult mammalian heart exhibits minimal regenerative capacity, rendering most major cardiac injuries irreversible. In contrast, neonatal mice retain the ability to regenerate cardiac tissue for a brief period after birth. Although single-cell RNA sequencing (scRNA-seq) has deepened our understanding of this regenerative window, existing datasets remain dispersed across studies, hindering integrative and comparative analyses. Methods and Results: To address this gap, we developed the Mouse CArdiac REgeneration DataBase (MCAREDB), a freely accessible, web-based resource that consolidates scRNA-seq datasets from multiple mouse cardiac regeneration studies. MCAREDB enables cell type-specific exploration of differential gene expression and pathway enrichment across key biological contexts, including neonatal versus postnatal stages and sham versus myocardial infarction surgeries. Core features include an interactive genome browser with epigenetic tracks, comparative gene expression visualization, pathway enrichment analysis, and bulk data download options. Conclusion: By integrating diverse datasets into a unified and interactive framework, MCAREDB lowers technical barriers to data exploration and accelerates hypothesis generation in cardiac regeneration research. The resource is freely available at https://MCAREDB.org:3305.

Recent advances in ultra-accurate sequencing technologies have revealed that somatic mutations accumulate throughout the human lifespan and may contribute to both normal aging and disease. These mutations are highly diverse, often non-recurrent, and functionally heterogeneous, making their biological impact difficult to evaluate systematically. Although many studies have profiled somatic mutations in individual tissues or limited cohorts, a centralized and scalable platform that integrates discoveries and supports functional interpretation has been lacking. To address this gap, we present SomaMutDB 2.0 (https://somamutdb.org/SomaMutDB/), a substantially expanded database cataloging 8.9 million mutations (8.57 million SNVs and 0.29 million INDELs) from 10 852 samples of 607 human subjects across 47 studies. Beyond expanded data coverage, SomaMutDB 2.0 introduces a comprehensive functional annotation framework applying 22 predictive models spanning coding, regulatory, expression-based, and ensemble predictors to systematically assess mutational impact. Users can browse pre-annotated variants through an interactive interface or upload their own variants for real-time analysis. Notably, all results are contextualized against mutations from normal, non-diseased tissues in the database, enabling more meaningful interpretation than raw scores alone. Together, these advances establish SomaMutDB 2.0 as the most comprehensive resource currently available for characterizing somatic mosaicism and its functional impact on human health and aging.

Somatic mutation is now recognized as a cause of multiple human diseases other than cancer. Osteoarthritis (OA), a highly prevalent age-related disease, has been associated with increased chromosomal abnormalities in articular cartilage. Here we characterize the somatic mutational landscape of chondrocytes during normal aging and in affected cartilage of patients with OA. We used single-cell whole-genome sequencing to analyze single-nucleotide variants (SNVs) and small insertions and deletions (InDels) in 100 chondrocytes isolated from the cartilage of hip femoral heads of 17 research participants aged 26−90 years, including 9 patients with OA and 8 non-OA donors. Both SNVs and InDels accumulate with age in chondrocytes with a clock-like mutational signature. Surprisingly, the age-related accumulation rate in OA chondrocytes is lower than that in non-OA control chondrocytes. Differences in mutational signatures and Gene Ontology term enrichment were found between OA and non-OA control samples. In this study, to understand the role of somatic mutation in the pathogenesis of OA, we characterized somatic SNV and InDel mutations. With further progress in analytical approaches, structural variations in the chondrocyte genome are also expected to provide valuable information.

Biological aging is associated with progressively more severe genetic and epigenetic alterations. While these changes are expected to affect the transcriptional profile of cells, the magnitude of that effect is unknown as the aging transcriptome is still poorly understood. Understanding the aging transcriptional landscape will give us greater insight into how cells are affected by and/or respond to the aging process. To facilitate the large-scale exploration of the aging transcriptome, we report the development of the Human Cell Aging Transcriptome Atlas (HCATA). HCATA, contains single-cell RNA-sequencing datasets from 76 publications totaling 92 million cells and 3,475 tissue-level samples across more than 50 tissue types with ages ranging from 0 to 103 years. HCATA includes a genome browser that allows users to interactively explore age-related differential expression, as well as functions to explore related pathways at the tissue and cell-type level. HCATA is publicly accessible at http://hcata-xiaodonglab.org:3304.

Recent advances in ultra-accurate sequencing technologies have revealed that somatic mutations accumulate across the human lifespan and may contribute to both normal aging and disease. These mutations are highly diverse, often non-recurrent, and functionally heterogeneous, which makes their biological impact difficult to evaluate systematically. Although many studies have profiled somatic mutations in individual tissues or limited cohorts, a centralized and scalable platform that integrates discoveries and supports functional interpretation has been lacking. To address this gap, we present SomaMutDB 2.0 ( https://somamutdb.org/SomaMutDB/ ), a substantially expanded database that catalogs 8.9 million mutations (8.57 million SNVs and 0.29 million INDELs) from 10,852 samples of 607 human subjects across 47 studies. Beyond expanded data coverage, SomaMutDB 2.0 introduces a comprehensive functional annotation framework that applies 22 predictive models, encompassing coding, regulatory, expression-based, and ensemble predictors, to systematically assess mutational impact. Users can browse pre-annotated variants through an interactive interface or upload their own variants for real-time analysis, with results contextualized against all mutations from normal, non-diseased tissues in the database. Together, these advances establish SomaMutDB 2.0 as the most comprehensive resource currently available for characterizing somatic mosaicism and functional interpretation in human health and aging. Graphical SomaMutDB 2.0 provides an expanded catalog of 8.9 million somatic mutations across 30 tissues, along with pipelines for mutational signature analysis and 22-tool functional annotation that enable user-submitted variant interpretation.

Accurate detection of somatic mutations in noncancerous cells is critical for studying somatic mosaicism, a process implicated in aging and multiple chronic diseases. However, single-cell and single-molecule DNA sequencing platforms differ in their error profiles, coverage biases, and sensitivity to specific mutation types, complicating cross-platform comparisons. Here, we present in vitro and in silico benchmarks to quantify true-positive and false-positive rates in single-cell whole-genome sequencing using Single-Cell Multiple Displacement Amplification, and in single-molecule sequencing using Nanorate Sequencing (NS) and whole-genome NS (WGNS). Using standard cell lines, we show that all three methods detect single-nucleotide variants (sSNVs) and small insertions and deletions (sINDELs) with high accuracy, but differ in genomic coverage and susceptibility to artifacts. Method-specific biases influence mutational signatures and hotspot detection. Applying results of the benchmark to IMR-90 fibroblasts, we estimate higher in vitro mutation rates using NS than expected from in vivo data, consistent with potential replication stress and culture-associated DNA damage. Overall, our study highlights the substantial impact of sequencing platform-specific biases on somatic mutation detection and interpretation, and lays the foundation for standardized, cross-platform-comparable analyses of somatic mosaicism in normal human tissues.