About

Chenxu Zhu, Ph.D., is an Assistant Professor of Physiology and Biophysics at Weill Cornell Medicine, specializing in Computational Cancer Systems Biology and Genomics in Computational Biomedicine. His research focuses on understanding complex molecular networks that control cell type-specific gene expression, which is critical for understanding the causes of human diseases and developing treatments. His work addresses the challenges in cancer research posed by the stochastic accumulation of genetic and epigenetic alterations in heterogeneous cellular populations. Dr. Zhu's laboratory develops cutting-edge genomic technologies for multimodal single-cell analysis and applies these technologies to study the combined effects of multiple regulatory layers in cell fate specification and maintenance during development, diseases, and cancer. His research aims to dissect the principles underlying gene regulation at the single-cell level, leveraging advancements in single-cell genomics to access comprehensive profiles from complex tissues.

Research topics

• Biology
• Computational biology
• Computer science
• Genetics
• Chemistry

Selected publications

• Integrative Single-Cell Epigenomic Atlas Annotates the Regulatory Genome of the Adult Mouse Brain

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-02-07

  articleOpen access

  SUMMARY Histone modifications underpin the cell-type-specific gene regulatory networks that drive the remarkable cellular heterogeneity of the adult mammalian brain. Here, we profiled four histone modifications jointly with transcriptome in 2.5 million nuclei across multiple adult mouse brain regions. By integrating these data with existing maps of chromatin accessibility, DNA methylation, and 3D genome organization, we established a unified regulatory framework for over 100 brain cell subclasses. This integrative epigenomic atlas annotates 81% of the genome, defining distinct active, primed, and repressive states. Notably, active chromatin states marked by combinatorial histone modifications more precisely identify functional enhancers than chromatin accessibility alone, while Polycomb- and H3K9me3-mediated repression contributes prominently to cell-type-specific regulation. Finally, this multi-modal resource enables deep learning models to predict epigenomic features and gene expression from DNA sequences. This work provides a comprehensive annotation of the mouse brain regulatory genome and a framework for interpreting non-coding variation in complex tissues. HIGHLIGHTS - A single-cell epigenome atlas of transcription and four histone modifications across 2.5M mouse brain cells. - Multi-modal integration functionally annotates ∼81% of the adult mouse brain regulatory genome - Cell-type-resolved chromatin landscapes reveal regulatory programs mediated by enhancers, Polycomb and H3K9me3 - Deep learning models predict cell-type-specific epigenomic features and gene expression from DNA sequence

  PublisherDOI

• Hierarchical NiCo2S4/NH2-UIO-66 nanoflower composites coupled with S, N-doped carbon for hybrid supercapacitors

  Journal of Energy Storage · 2026-04-14

  article
  PublisherDOI

• Live-cell decoding of labile post-translational modifications in APE1 with a rationally engineered nano-catcher

  Nucleic Acids Research · 2026-03-19

  articleOpen access

  Human apurinic/apyrimidinic endonuclease 1/redox effector factor 1 (APE1) is a multifunctional protein central to DNA repair and redox regulation, yet its dynamic post-translational modifications (PTMs) remain poorly understood. Here, we report a biotin-regulated avidin-based nano-catcher (bMIPAPE1) capable of capturing active APE1 in living cells. By leveraging biotin-saturated avidin assembled onto magnetic nanoparticles and surface-imprinted with polydopamine, we engineered highly specific binding cavities for APE1 that enable retention of labile PTMs. This platform revealed 25 previously unreported PTMs across 18 residues of APE1, encompassing acetylation, phosphorylation, ubiquitination, methylation, S-nitrosylation, palmitoylation, and succinylation, and highlighting several PTM hotspots. Representative modifications include phosphorylation at Y264 and Y269, and acetylation at K63, with several PTMs associated with APE1 nuclear export. In addition to high specificity and intracellular compatibility, bMIPAPE1 attenuated both the DNA repair and redox-related functions of APE1. Our findings demonstrate the utility of artificial nanocomposites as tools for live-cell PTM profiling and functional modulation of target proteins, offering a powerful approach to decode protein regulation in living systems and identify potential therapeutic targets in cancer.

  PublisherDOI

• Single-cell Multiome Analysis of Chromatin State and Transcriptome in the Human Basal Ganglia

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-02-04 · 1 citations

  articleOpen access

  SUMMARY The basal ganglia play essential roles in motor control, emotion, learning and reward processing. Their dysfunction contributes to many neurological and psychiatric disorders. However, the gene regulatory programs defining basal ganglia cell-type identity and function remain poorly understood, limiting interpretation of disease-associated non-coding variants. Here, we present the first single-cell multiome atlas of histone modifications and transcriptomes across eight basal ganglia regions from neurotypical adult human donors. Joint profiling reveals cell-type-specific deployment of active and repressive cis -regulatory elements and gene regulatory networks, and suggests a combinatorial homeobox transcription factor code underlying cell identity. Integration with matched spatial transcriptomic MERFISH data uncovers regional heterogeneity of epigenomic landscapes. Comparative analysis between human and mouse medium spiny neurons uncovers conservation of core gene regulatory features. This atlas interprets non-coding risk variants of neuropsychiatric disorders and supports the development of a deep learning model to predict gene regulation and functional effects of disease-associated variants. HIGHLIGHTS Joint single-cell profiling of transcriptomes and three histone modifications across eight human basal ganglia regions characterizes active and repressive chromatin states at cell-type resolution. Cell-type-specific gene regulatory programs decode combinatorial homeobox TF grammar governing the identity and diversification of basal ganglia neurons. Intergrative analyses link noncoding neuropsychiatric risk variants to specific cell types, regulatory elements, and candidate target genes. A sequence-to-function deep-learning model predicts gene regulation from DNA sequence and prioritizes functional disease-associated variants.

  PublisherOA PDFDOI

• Single-cell 5-hydroxymethylcytosine landscapes of mouse early embryos at single-base resolution

  Cell Reports · 2025-04-01 · 2 citations

  articleOpen access

  DNA methylation and hydroxymethylation are extensively reprogrammed during mammalian early embryogenesis, and studying their regulatory functions requires comprehensive DNA hydroxymethylation maps at base resolution. Here, we develop single-cell 5-hydroxymethylcytosine (5hmC) chemical-assisted C-to-T conversion-enabled sequencing (schmC-CATCH), a method leveraging selective 5hmC labeling for a quantitative, base-resolution, genome-wide landscape of the DNA hydroxymethylome in mouse gametes and preimplantation embryos spanning from the zygote to blastocyst stage. We revealed that, in addition to late zygotic stages, onset of ten-eleven translocation (TET)-mediated DNA hydroxymethylation initiates immediately after fertilization and is characterized by the distinct 5hmC patterns on the parental genomes shaped by TET3 demethylase. We identified persistent clusters of 5hmC hotspots throughout early embryonic stages, which are highly associated with young retroelements. 5hmC is also associated with different regulatory elements, indicating a potential regulatory function during early embryogenesis. Collectively, our work elucidates the dynamics of active DNA demethylation during mouse preimplantation development and provides a valuable resource for functional studies of epigenetic reprogramming in early embryos.

  PublisherDOI

• SnapFISH: a computational pipeline to identify chromatin loops from multiplexed DNA FISH data

  UNC Libraries · 2025-10-21

  articleOpen accessSenior author
  PublisherDOI

• LLM4Tag: Automatic Tagging System for Information Retrieval via Large Language Models

  ArXiv.org · 2025-02-19

  preprintOpen access

  Tagging systems play an essential role in various information retrieval applications such as search engines and recommender systems. Recently, Large Language Models (LLMs) have been applied in tagging systems due to their extensive world knowledge, semantic understanding, and reasoning capabilities. Despite achieving remarkable performance, existing methods still have limitations, including difficulties in retrieving relevant candidate tags comprehensively, challenges in adapting to emerging domain-specific knowledge, and the lack of reliable tag confidence quantification. To address these three limitations above, we propose an automatic tagging system LLM4Tag. First, a graph-based tag recall module is designed to effectively and comprehensively construct a small-scale highly relevant candidate tag set. Subsequently, a knowledge-enhanced tag generation module is employed to generate accurate tags with long-term and short-term knowledge injection. Finally, a tag confidence calibration module is introduced to generate reliable tag confidence scores. Extensive experiments over three large-scale industrial datasets show that LLM4Tag significantly outperforms the state-of-the-art baselines and LLM4Tag has been deployed online for content tagging to serve hundreds of millions of users.

  PublisherOA PDFDOI

• The ‘vulnerability code’: Is cell identity the architect of its own decay?

  Clinical and Translational Medicine · 2025-12-01

  articleOpen accessSenior authorCorresponding

  The establishment and maintenance of cellular identity depend on two fundamental yet historically separately studied mechanisms: DNA repair machineries that safeguard genome fidelity, and epigenetic programs that regulate cell-type-specific gene expression patterns.1 Research on DNA damage repair has primarily emphasised the molecular pathways and kinetics of the DNA damage response.2 Conversely, investigations on epigenetics have focused on how cells establish and sustain their transcriptional memory.3 This disciplinary separation has created a blind spot: while we understand how the hardware (genome integrity) fails and how the software (epigenetic regulation) becomes compromised, the question of whether and how the former impairs the latter remains unresolved. Recent research indicates that these two processes are interconnected. The link between them is recognised through the understanding that DNA damage constitutes an effective ‘toxic modification’ that influences gene regulation. When cells experience genotoxic stress, chromatin temporarily relaxes to permit the access of repair factors to DNA; this process involves the phosphorylation of histone H2AX and the recruitment of chromatin remodellers.4 Ideally, these changes are transient and are reversed once the damage has been repaired. However, some modifications may not be fully reversed, leaving persistent epigenetic ‘scars’ that affect gene expression long after the repair process.2 Conversely, the existing epigenetic context influences the genome's vulnerability: heterochromatin regions tend to undergo slower repair, while active regions such as promoters and enhancers are more accessible but also more prone to damage caused by transcriptional activity or toxins.5 This results in a complex, system-wide challenge: the concept that chromatin modifiers are repurposed for DNA repair suggests that the balance of cell survival and proper functioning involves an ongoing conflict at the molecular level. This bidirectional relationship constitutes a ‘systems level’ issue: the ‘relocalization of chromatin modifiers’ theory proposes that the machinery used for repairing DNA breaks is frequently borrowed from the epigenetic maintenance system, leading to a direct conflict between the maintenance of cell survival and the preservation of cellular function.6 These findings challenge the long-held view in the field that DNA damage occurs randomly primarily due to thermodynamic noise and environmental factors. If DNA damage hotspots can be precisely identified, the trajectory of cellular decline may become more predictable, as the initial regulatory failure can be specifically targeted. The susceptibility of organs to DNA damage varies considerably among mammals: the most vulnerable are energy-dense tissues such as the brain, owing to their reliance on abundant redox-active compounds. Indeed, within the nervous system, long-lived cells tend to accumulate damage over extended periods, highlighting the importance of these mechanisms in neural aging and pathology. Recent studies have demonstrated significant correlations between elevated DNA damage, the gradual decline of the epigenetic landscape and the development of age-related neurodegenerative disorders,7, 8 indicating that an imbalance between DNA repair mechanisms and epigenetic stability may contribute to neuronal aging and disease progression. This creates a major technical challenge: precisely identifying damage hotspots across different cell types within complex, heterogeneous tissues. To bridge this gap, we developed Paired-Damage-seq, a method for single-cell parallel analysis of oxidative and single-strand DNA breaks alongside the transcriptome.9 We first benchmarked Paired-Damage-seq in a cultured cell line: the specificity of damage detection was validated against previously established bulk assays, and the sensitivity was evaluated using publicly available single-cell datasets (of transcriptome and epigenome). Additionally, analysis of stress-treated cells showed coordinated changes between DNA damage accumulation and epigenomic shifts, highlighting the method's ability to detect both baseline and stress-induced patterns. By applying this technology to the mouse cerebral cortex, we confirmed that DNA damage hotspots are not randomly distributed: instead, regulatory hotspots, including nucleosome-depleted regions, bear the bulk of the burden. As expected, Paired-Damage-seq demonstrated that distinct cell populations (e.g., neurons versus glia) exhibit unique vulnerability profiles associated with their specific cellular functions. For example, neuron cells tend to accumulate damage in synaptic genes, whereas glial cell-specific DNA damage hotspots are enriched in metabolic regulatory regions. This indicates that the ‘cost’ of cellular identity constitutes a particular vulnerability to damage within the specific genomic elements that establish that identity. The identification of DNA damage hotspots offers valuable mechanistic insights: DNA damage repair not only maintains the structural integrity of the genome but also preserves epigenetic stability. Our findings indicate that these damage hotspots may serve as early markers of epigenetic changes, as they tend to lose epigenetic memory faster than other genomic regions (the ‘vulnerability code’). This process potentially results in a ‘scarring’ effect on the chromatin landscape. When repair factors are repeatedly recruited to the same hotspots, the local epigenetic states (such as histone and DNA modifications) may not be perfectly restored (Figure 1). This aligns with the ‘Information Theory of Aging’, which proposes that the noise produced by DNA repair processes gradually damages epigenetic memory, causing a decline in cellular identity.10 Consequently, a map of DNA damage can serve as a useful prediction tool for the cell's future state. By locating ongoing genomic breaks, it becomes possible to anticipate which genes might soon lose their regulatory control. This shifts the perspective on aging from a stochastic to a more deterministic process; by identifying key hotspots, we can improve predictions of future epigenomic changes. This shift from ‘stochastic’ to ‘deterministic’ has immediate translational implications for the diagnosis and treatment of age-related diseases. By identifying these ‘fragile’ regulatory regions, we can detect the earliest signs of cellular decline, potentially years before pathology manifests or global epigenetic changes become noticeable. Furthermore, this understanding opens new avenues for therapeutic intervention. Current therapeutic approaches often depend on broad-spectrum antioxidants or general epigenetic modulators,1, 2 which may lack sufficient specificity to achieve optimal efficacy. Future interventions could focus on reinforcing local epigenetic states (such as chromatin accessibility) at the specific DNA damage hotspots, rather than attempting to globally protect the entire genome. Ultimately, by integrating the study of DNA damage and epigenetics, we are entering a new era of precision medicine. By mapping the process of cellular identity erosion, we can potentially intervene to arrest or even reverse it. C.Z. is supported by Weill Cornell Medicine and New York Genome Center startup funds, National Institutes of Health (NIH)/National Institute of General Medical Sciences (grant no. DP2GM154011), NIH/National Human Genome Research Institute (grant nos. R00HG011483 and RM1HG011014) and the MacMillan Center for the Study of the Noncoding Cancer Genome at the New York Genome Center. C.Z. and D.B. are listed as inventors of a provisional patent application related to Paired-Damage-seq.

  PublisherOA PDFDOI

• Cardiac fibroblast-derived mitochondria-enriched sEVs regulate tissue inflammation and ventricular remodeling post-myocardial infarction through NLRP3 pathway

  Pharmacological Research · 2025-02-25 · 13 citations

  articleOpen access

  Resident cardiac fibroblasts (CFs) play crucial roles in sensing injury signals and regulating inflammatory responses post-myocardial infarction (MI). Damaged mitochondria can be transferred extracellularly via various mechanisms, including extracellular vesicles (EVs). In this study, we aimed to investigate whether CFs could transfer damaged mitochondrial components via small EVs (sEVs) and elucidate their role in regulating inflammatory responses post-MI. Left anterior descending coronary artery ligation was performed in mice. Mitochondrial components in sEVs were detected using nanoflow cytometry. Differential protein expression in sEVs from normoxia and normoglycemia CFs (CFs-Nor-sEVs) and CFs post oxygen-glucose deprivation (CFs-OGD-sEVs) was identified using label-free proteomics. CFs-sEVs were co-cultured with mouse bone marrow-derived macrophages (BMDMs) to assess macrophage inflammatory responses. Effects of intramyocardial injection of CFs-sEVs were assessed in MI mice in the absence or presence of NLRP3 inhibitor CY-09. Results demonstrated that mitochondrial components were detected in CFs-derived sEVs post-MI. Damaged mitochondrial components were enriched in CFs-OGD-sEVs (CFs-mt-sEVs), which promoted pro-inflammatory phenotype activation of BMDMs in vitro. Myocardial injection of CFs-mt-sEVs enhanced tissue inflammation, aggravated cardiac dysfunction, and exacerbated maladaptive ventricular remodeling post-MI in vivo. Mechanistically, above effects were achieved via activation of NLRP3 and above effects could be reversed by NLRP3 inhibitor CY-09. This study indicates that CFs could transfer damaged mitochondrial components via the sEVs post-MI, promote macrophage inflammatory activation and exacerbate maladaptive ventricular remodeling post MI by activating NLRP3. Our findings highlight the potential therapeutic effects of inhibiting CFs-mt-sEVs and NLRP3 to improve cardiac function and attenuate ventricular remodeling post-MI. • CFs could transfer damaged mitochondrial components via sEVs (CFs-mt-sEVs) post-MI. • NLRP3 mediates the proinflammatory effects of CFs-mt-sEVs. • The NLRP3 inhibitor CY-09 could reverse the negative effects of CFs-mt-sEVs post-MI. • Detecting mitochondrial components in plasma sEVs may aid risk stratification in MI patients. • Providing Evidence for future clinical studies testing the effects of CY-09 on MI patients.

  PublisherDOI

• SnapHiC-D: a computational pipeline to identify differential chromatin contacts from single-cell Hi-C data

  UNC Libraries · 2025-10-21

  articleOpen access

  Single-cell high-throughput chromatin conformation capture technologies (scHi-C) has been used to map chromatin spatial organization in complex tissues. However, computational tools to detect differential chromatin contacts (DCCs) from scHi-C datasets in development and through disease pathogenesis are still lacking. Here, we present SnapHiC-D, a computational pipeline to identify DCCs between two scHi-C datasets. Compared to methods designed for bulk Hi-C data, SnapHiC-D detects DCCs with high sensitivity and accuracy. We used SnapHiC-D to identify cell-type-specific chromatin contacts at 10 Kb resolution in mouse hippocampal and human prefrontal cortical tissues, demonstrating that DCCs detected in the hippocampal and cortical cell types are generally associated with cell-type-specific gene expression patterns and epigenomic features. SnapHiC-D is freely available at https://github.com/HuMingLab/SnapHiC-D.

  PublisherDOI

Recent grants

• Studying the mammalian regulatory circuits by developing single-cell multi-omics technologies

  NIH · $180k · 2020–2022

Frequent coauthors

• Bing Ren

  University of California, San Diego

  51 shared

• Chengqi Yi

  Center for Life Sciences

  40 shared

• Yongping Huang

  Center for Excellence in Molecular Plant Sciences

  33 shared

• Dehong Yang

  Nanfang Hospital

  23 shared

• Yang Xu

  Shanghai Jiao Tong University

  21 shared

• Yanxiao Zhang

  Westlake University

  21 shared

• Lansa Qian

  Center for Excellence in Molecular Plant Sciences

  20 shared

• Yaohui Wang

  Chinese Academy of Sciences

  18 shared

Education

• Ph.D., School of Life Sciences

  Peking University

  2017

• B.S., College of Chemistry and Molecular Science

  Wuhan University

  2011