Biology subjects

Zhan, Y.

Publications and source records attributed to Zhan, Y..

At least 19 recordsLinked to original sources

bioRxiv2026

DAPE-PD: Diffusion adaptation of genomic foundation model enables tailored-generation of cis-regulatory DNA

Genomic foundation models (GFMs) have transformed our ability to learn the evolutionary language encoded in genomes, yet their potential for programmable regulatory DNA engineering remains underexplored. Regulatory function depends on combinatorial interactions among distributed sequence elements, motivating generation with bidirectional context. Here, we introduce DAPE-PD (Diffusion Adaptation of Pretrained Evo for Promoter Design), which adapts the autoregressive Evo model into a bidirectional masked diffusion model without foundation-scale retraining. DAPE-PD combines pretrained genomic representations with iterative unmasking to generate regulatory DNA sequences. Following genus-specific fine-tuning on genome-derived intergenic sequences, DAPE-PD generated sequences that retained genus-associated compositional features despite low similarity to the fine-tuning datasets. Experimental validation identified functional de novo promoters, while in silico prioritization recovered active designs from small experimental candidate sets across three non-model gut commensals. This framework connects genomic pretraining with diffusion-based generation, extending GFMs from sequence understanding toward functional cis-regulatory DNA engineering in living cells.

synthetic biology↗

bioRxiv2026

Defense-associated reverse transcriptases synthesize DNA in a telomerase-like manner

Defense-associated reverse transcriptases (DRTs) constitute an emerging class of bacterial defense systems that harness reverse-transcriptase activity to generate cDNA products and counter phage infection. A distinct subset of DRT10 systems evolutionarily related to telomerase generates tandem-repeat DNA to confer antiphage immunity. Here, we biochemically characterize the reverse transcription activity of DRT10 and determine cryo-EM structures of DRT10-ncRNA complex in three functional states, revealing an unanticipated 2:1 RT:ncRNA assembly mode. Two DRT10 protomers dimerize through a central pair of interlocked {beta}-hairpins and are further embraced by a single ncRNA. This architecture distinguishes DRT10 from previously characterized DRT complexes, which either lack an ncRNA component or exhibit an equivalent stoichiometry of RT:ncRNA. Strikingly, despite the nearly identical conformations of the two DRT10 protomers, cDNA product density was resolved in only one catalytic pocket, revealing different product occupancy within the dimer. Moreover, along with flanking linkers, the stem loops of the ncRNA as fixed points to supply steric barriers help define the template boundary. These findings reveal a distinctive ncRNA-mediated DRT10 dimeric assembly and establish a mechanistic framework for protein-primed reverse transcription constrained by the length of the RNA flanking linker.

biochemistry↗

bioRxiv2026

ProxiNet transfers spatially learned cellular proximity to dissociated single-cell transcriptomes

Spatial transcriptomics reveals cellular organization within intact tissues, whereas dissociated single-cell RNA sequencing provides broad transcriptomic coverage but loses information about cellular proximity and neighborhood structure. Here, we developed ProxiNet, a spatially supervised framework that learns transcriptomic signatures of pairwise cellular proximity from spatial reference datasets and transfers these relationships to dissociated single-cell transcriptomes. ProxiNet predicted cellular proximity across brain regions and spatial technologies, including zero-shot cross-technology transfer, and gradient-based attribution identified genes and broader transcriptional programs associated with proximity predictions. In spatial datasets with known coordinates, ProxiNet-derived cellular neighborhoods recovered reproducible tissue organization and anatomical structure, providing independent spatial validation of the inferred proximity relationships. Applying the spatially calibrated model to dissociated scRNA-seq revealed heterogeneous cellular neighborhoods with distinct cell-type compositions and candidate communication programs. In an Alzheimer dataset, ProxiNet further identified age-associated remodeling of inferred neighborhoods, including an AD-associated neighborhood at 8 months characterized by distinct astrocyte and neuronal transcriptional states and candidate intercellular communication programs. Together, these results establish pairwise cellular proximity as an interpretable and transferable representation for extending spatially learned tissue organization to dissociated single-cell transcriptomes.

bioinformatics↗

bioRxiv2026

Chimeric Induced Cooperativity Opens the Design Space of Eukaryotic Gene Regulation

Predictive engineering of eukaryotic transcription is limited by the coupling of signal sensing, DNA binding, TF abundance and promoter output. Here we establish chimeric induced cooperativity (CIC), a modular architecture that separates LBD, DBD and AD functions and links them to promoters with tunable basal and maximal output. Module parameters can be recombined to predict new CIC-TF configurations and guide design before construction. Ligand-induced cooperativity reduces basal DNA occupancy while increasing induced occupancy, and effective DBDs combine low OFF-state activity with strong ON-state promoter occupancy rather than binding strength alone. Synthetic promoters independently control occupancy gain and output range. The same framework extends to repression and can be recalibrated with limited measurements in mammalian cells. In yeast, CIC-12 achieved a mean fold induction of 298-fold across 12 orthogonal sensors; an earlier CIC-10 chassis enabled model-guided optimization of an eight-gene vitamin B5 biosynthetic pathway. CIC establishes a programmable, model-guided design space for eukaryotic transcriptional control.

synthetic biology↗

bioRxiv2026

SMARCA4, STK11, and KEAP1 co-inactivation associates with poor prognosis and upregulation of the TGF-β pathway in lung adenocarcinoma

BackgroundLung adenocarcinoma (LUAD) is clinically and molecularly defined by oncogenic driver mutations, identification of which has led to the development of driver-targeted therapies and substantial improvements in prognosis for subsets of LUAD patients. Recent studies assessing clinical outcomes in the context of multigenic alterations have identified secondary mutations that might explain differential responses to targeted therapies, chemotherapies and immunotherapies. Genetic inactivation or loss of SMARCA4, which frequently co-occurs with loss-of-function mutations in STK11 and KEAP1, is especially predictive of poor prognosis and shorter overall survival in LUAD patients, regardless of driver status. We sought to examine the clinical and functional associations of SMARCA4 deficiency in LUAD, with or without co-associated STK11/KEAP1 loss-of-function. MethodsWe examined correlation between SMARCA4 loss, gene expression and prognosis through genomic and transcriptomic profiling of clinically annotated LUAD samples. We generated isogenic cell line models with genetic knockouts of SMARCA4 with or without concomitant STK11 and KEAP1 to profile mutationally-defined genotypes of interest in vitro and in vivo. Lastly, we interrogated the functional dependency of SMARCA4/STK11/KEAP1 triple mutant models on TGF-{beta} signaling to assess its potential as a therapeutic target. ResultsSMARCA4/STK11/KEAP1 triple mutant LUAD is associated with poor survival and high frequency of multisite metastasis. SMARCA4/STK11/KEAP1 triple knockout models showed enhanced migration and invasion in vitro, and diversified organotropism in an in vivo intracardiac xenograft metastasis assay. RNA-Seq and DNaseI-Seq of these in vitro models and clinical samples identified upregulation of TGF-{beta} signaling and EMT gene expression signatures, and corresponding changes in chromatin accessibility, in SMARCA4/STK11/KEAP1 triple mutant LUAD. ConclusionsWe identify SMARCA4/STK11/KEAP1 triple mutant LUAD as a prognostically significant disease subset and nominate TGF-{beta} signaling as a potential therapeutic target.

cancer biology↗

bioRxiv2026

The +1 nucleosome functions in Pol II transcription initiation and the transition to elongation

Transcription initiation by RNA polymerase II (Pol II) occurs next to a +1 nucleosome that is positioned downstream of the transcription start site (TSS). The +1 nucleosome has been shown to influence the pre-initiation complex (PIC) assembly and Pol II pausing, but it is unclear whether and how it functions in transcription initiation and the transition to elongation. Here, we investigate the transcription initiation-elongation transition in vitro using DNA templates containing a +1 nucleosome, and we present cryo-EM structures of five intermediate states. First, after PIC assembly, the +1 nucleosome evicts TFIID from the PIC upon binding of ATP to TFIIH. Second, after DNA opening, the +1 nucleosome stimulates TFIIH translocase activity and initial RNA synthesis. Finally, after DNA bubble rewinding, the +1 nucleosome removes TFIIH from the early elongation complex for promoter escape. Our findings show that the +1 nucleosome not only acts as a passive border for PIC assembly and a passive barrier for Pol II pausing, but rather has active functions during the initiation-elongation transition of transcription.

molecular biology↗

bioRxiv2025

Structure of human cytoplasmic Pol II complex explains global transcription repression by Gdown1

RNA polymerase II (Pol II) is a 12-subunit enzyme crucial for gene transcription in the nucleus. However, its assembly in the cytoplasm, nuclear import, and nuclear function of assembly factors remain poorly understood. Here, we isolated Pol II from the cytoplasmic fraction of human cells (cfPol II) and determined its cryo-EM structure. The structure reveals that Pol II is fully assembled in the cytoplasm before nuclear import. We also found that Gdown1 binds Pol II through three distinct regions, indicating it may stabilize Pol II assembly intermediates. Notably, Gdown1 binding precludes the association of essential transcription factors IIB and IIF, rendering cfPol II inactive in promoter-dependent transcription in vitro. Our results provide a basis for Gdown1-dependent global transcription repression and suggest a model for the role of Gdown1 in Pol II assembly, import, and transcription regulation.

biochemistry↗

bioRxiv2025

Sourdough starter-inspired living materials grown from probiotic consortia with engineered division-of-labor

Genetic circuits enable precise control over cellular behavior. As circuits become more complex, implementing them within a single cell population becomes difficult. Tasks can instead be distributed in engineered consortia. However, selecting appropriate consortia and harnessing the natural traits of each species for such a division of labor remain unsolved challenges. Here, we report the creation of SINERGY, an engineered synthetic sourdough starter in which functional modules are distributed between the two constituent species. These modules, consisting of sensing, communication, and response, enable in vitro biosensing and in vivo drug delivery and modulate the gut ecosystem in an animal disease model. SINERGY not only preserves the safety and health-promoting properties of the natural components but also endows microbes with programmable functionalities, facilitating broad applications in biomedicine.

synthetic biology↗

bioRxiv2025

MiTo: tracing the phenotypic evolution of somatic cell lineages via mitochondrial single-cell multi-omics

Mitochondrial single-cell lineage tracing (MT-scLT) has recently emerged as a scalable and non-invasive tool to trace somatic cell lineages. However, the reliability and resolution of MT-scLT remains highly debated. Here, we present MiTo, the first end-to-end framework for robust MT-scLT data analysis. Thanks to highly-optimized algorithms and user-friendly interfaces, this modular toolkit offers unprecedented control across the entire MT-scLT workflow. Benchmarked against novel real-world datasets (375-2,757 cells; 8-216 lentiviral clones), MiTo outperformed state-of-the-art methods and baselines in MT-scLT data pre-processing and clonal inference. Applied to a time-resolved dataset of breast cancer evolution (>2,500 cells), MiTo accurately inferred ground-truth cell lineages (ARI=0.94) and cell state transitions, detected clonal fitness markers, and quantified heritability of gene regulatory networks. Comparing alternative lineage markers, MiTo quantified the resolution limit of existing MT-scLT systems, which currently enable reliable inference of coarse-grained cellular ancestries, but not high-resolution phylogenetic inference. In conclusion, this work provides robust tools and practical guidelines to dissect somatic evolution with single-cell multi-omics.

bioinformatics↗

bioRxiv2025

Systematic evaluation of blood contamination in nanoparticle-based plasma proteomics

Circulating blood proteomics enables minimally invasive biomarker discovery. Nanoparticle (NP)-based circulating plasma proteomics studies have reported varying number of proteins, ranging from ca 2000 to 7000 but its not clear whether higher protein number is more informative. Here, we first develop OmniProt - a silica-NP workflow for plasma proteomics which is optimized through systematic evaluation of NP types and protein corona formation parameters. Next, we present an Astral spectral library for 10,109 protein groups expressed in human plasma. Using Astral with a throughput of 60 sample-per-day, OmniProt identifies ca 3000 to 6000 protein groups from human plasma. Notably, we found that platelet/erythrocyte/coagulation-related contamination artificially elevates protein identifications and compromises quantification accuracy for circulating plasma proteins in NP-enriched samples. Through controlled contamination experiments, we identified and validated protein biomarkers for contaminations of platelet, erythrocyte, and coagulation, respectively, in NP-based plasma proteomics experiments. We further developed an open-access software Baize for accessing plasma proteomics contamination. Finally, we validated our OmniProt workflow and Baize software in a clinical cohort of 193 patients with CT-indistinct benign nodules or early lung cancers. On average, we identified ca 4400 protein groups for each plasma sample without significant contamination. Five samples were flagged out as contaminated. In conclusion, this study reveals that contamination of plasma samples significantly alters the protein identification and quantification in NP-based plasma proteomics experiments, and presents an open-access software Baize to evaluate plasma proteome contamination.

biochemistry↗

bioRxiv2025

Maternal age and genome-wide failure of meiotic recombination are associated with triploid conceptions in humans

Triploid and haploid conceptions are not viable and are a common occurrence in humans, where they account for 10% of all pregnancy losses. Despite the parent-of-origin being important in the etiology of the pregnancy, our knowledge of their causes is limited, especially at the point of conception. Using a dataset of 96,660 biopsies and a validation dataset of 44,324 from human blastocysts embryos generated by intra-cytoplasmic sperm injection (ICSI), we estimate that 1.1% of human conceptions (n=1,063) contain extra or missing chromosome sets in zygotes. We identify a maternal age effect, with a 1.059 per year increased risk in triploidy/haploidy (p=0.0008). In 0.03% of couples, we identified three or more triploid/haploid embryos, suggesting a personal risk effect (p=0.03). Genotype analysis of 55 triploid embryo biopsies and their parents show that one third of maternal triploid conceptions originate in meiosis I and two thirds in meiosis II. Seven of these embryos are inferred to have entirely failed to initiate meiotic recombination genome-wide, suggesting that human oocytes with pervasive meiotic recombination failure that are formed during fetal development are capable of ovulation in adult life. Finally, we identify a new type of genome-wide maternal isodiploidy (two maternal chromosome sets) in 0.05% of embryos (41 of 74,009). Collectively, our findings shed light on the biology of meiosis and the formation of human oocytes with the number of chromosome sets.

genetics↗

bioRxiv2025

A perturbation proteomics-based foundation model for virtual cell construction

Building a virtual cell requires comprehensive understanding of protein network dynamics of a cell which necessitates large-scale perturbation proteome data and intelligent computational models learned from the proteome data corpus. Here, we generate a large-scale dataset of over 38 million perturbed protein measurements in breast cancer cell lines and develop a neural ordinary differential equation-based foundation model, namely ProteinTalks. During pretraining, ProteinTalks gains a fundamental understanding of cellular protein network dynamics. Our model encodes protein networks and exhibits consistently improved predictive accuracy across various downstream tasks, highlighting its generalization capabilities and adaptability. In cancer cells, ProteinTalks robustly predicts drug efficacy and synergy, identifies novel drug combinations, and, through its interpretability, uncovers resistance-associated proteins. When applied to more complex system, patient-derived tumor xenografts, ProteinTalks predicts potential responses to drugs. Its integration with clinical patient data enhances the prognosis prediction of breast cancer patients. Collectively, we present a foundational model based on proteome dynamics, offering potential for various downstream applications, including drug discovery, and providing a basis for developing virtual cells.

systems biology↗

bioRxiv2024

Cell types associated with human brain functional connectomes and their implications in psychiatric diseases

Cell types are fundamental to the functional organization of the human brain, yet the specific cell clusters contributing to functional connectomes remain unclear. Using human whole-brain single-cell RNA sequencing data, we investigated the relationship between cortical cell cluster distribution and functional connectomes. Our analysis identified dozens of cell clusters significantly associated with resting-state network connectivity, with excitatory neurons predominantly driving positive correlations and inhibitory neurons driving negative correlations. Many of these cell clusters are also conserved in macaques. Notably, functional network connectivity is predicted by cellular communication among these clusters. We further identified cell clusters linked to various neuropsychiatric disorders, with several clusters implicated in multiple conditions. Comparative analysis of schizophrenia and autism spectrum disorder revealed distinct expression patterns, highlighting disease-specific cellular mechanisms. These findings underscore the critical role of specific cell clusters in shaping functional connectomes and their implications for neuropsychiatric diseases.

neuroscience↗

bioRxiv2024

Dynamic interplay between RNA N6-methyladenosine modification and porcine reproductive and respiratory syndrome virus infection

N6-methyladenosine (m6A) has attracted significant attention for its role in various biological processes, including RNA stability, translation, and the immune response. Understanding the role of m6A in viral infections is crucial to deep the complex interaction between virus and host cells. Porcine Reproductive and Respiratory Syndrome Virus (PRRSV) is a significant pathogen affecting swine health worldwide. Here, we firstly identified the m6A peaks in the PRRSV genome by m6A RNA immunoprecipitation sequencing (m6A-seq). Seven m6A-enriched regions within the PRRSV genome were detected, with one located in the N protein-coding region and the others distributed across nonstructural protein-coding regions. Notably, the Nsp2-coding region contained the highest m6A peak, spanning approximately 178 nucleotides. Functional analyses demonstrated a positive correlation between m6A modification levels and PRRSV replication in porcine alveolar macrophages (PAMs), as modulating the expression of m6A methyltransferases and demethylases significantly influenced viral replication. Moreover, treatment with the universal methylation inhibitor 3-deazaadenosine (3-DAA) effectively suppressed PRRSV replication, suggesting its potential as a novel anti-PRRSV therapeutic. To further elucidate the role of m6A in PRRSV infection, we analyzed the m6A landscape in PAMs infected with pandemic and highly pathogenic PRRSV strains. Among the 4,677 transcripts exhibiting altered m6A modification levels, the MAPK14 gene as well as other genes in p38/MAPK signaling pathway potentially emerged as the preliminary targets of m6A-mediated epigenetic regulation during PRRSV infection. These findings provide new insights into the epigenetic mechanisms underlying PRRSV infection and may facilitate the development of targeted anti-PRRSV strategies and therapeutics. IMPORTANCEThe involvement of m6A in viral genome has broader implications for our understanding of virus evolution and pathogenicity. The ability of viruses to adapt to their hosts and exploit host cellular mechanisms is a critical factor in their success as pathogens. As m6A is a modification that can be manipulated by both host and virus, studying its role in viral infections may reveal new insights into viral evolution. Understanding how different viruses utilize m6A could inform strategies for developing antiviral therapies that target these interactions. Such therapies could disrupt the ability of viruses to exploit m6A modifications, potentially reducing their virulence and enhancing host defenses.

microbiology↗

bioRxiv2024

STEP: Deciphering Spatial Atlas at Single-Cell Level with Whole-Transcriptome Coverage

Recent advances in spatial transcriptomics have revolutionized our understanding of tissue spatial architecture and biological processes. However, many of these technologies face significant challenges in achieving either single-cell resolution or comprehensive whole-transcriptome profiling, hindering their capacity to fully elucidate intercellular interactions and the tissue microenvironment. To address these limitations, we present STEP, a hybrid framework that synergistically integrates probabilistic models with deep learning techniques for spatial transcriptome analysis. Through innovations in model and algorithm design, STEP not only enhances sequencing-based spatial transcriptome data to single-cell resolution but accurately infers transcriptome-wide expression levels for image-based spatial transcriptomic. By leveraging the nuclear features extracted from histological images, STEP achieves precise predictions of cell type and gene expression and effectively diffuses the discriminative ability to serial sections for modeling solid tissue landscapes. The capability is particularly advantageous for analyzing cells with distinctive characteristics, such as cancer cells, enabling cross-sample inference. In addition, STEP simulates intercellular communication through a spatially resolved cell-cell interaction network, uncovering intrinsic biological processes. Overall, STEP equips researchers with a powerful tool for understanding biological functions and unveiling spatial gene expression patterns, paving the way for advancements in spatial transcriptomics research. Code is available at https://github.com/childishHU/STEP.

bioinformatics↗

bioRxiv2024

Gene Expression Prediction from Histology Images via Hypergraph Neural Networks

Spatial transcriptomics reveals the spatial distribution of genes in complex tissues, providing crucial insights into biological processes, disease mechanisms, and drug development. The prediction of gene expression based on cost-effective histology images is a promising yet challenging field of research. Existing methods for gene prediction from histology images exhibit two major limitations. First, they ignore the intricate relationship between cell morphological information and gene expression. Second, these methods do not fully utilize the different latent stages of features extracted from the images. To address these limitations, we propose a novel hypergraph neural network model, HGGEP, to predict gene expressions from histology images. HGGEP includes a gradient enhancement module to enhance the models perception of cell morphological information. A lightweight backbone network extracts multiple latent stage features from the image, followed by attention mechanisms to refine the representation of features at each latent stage and capture their relations with nearby features. To explore higher-order associations among multiple latent stage features, we stack them and feed into the hypergraph to establish associations among features at different scales. Experimental results on multiple datasets from disease samples including cancers and tumor disease, demonstrate the superior performance of our HGGEP model than existing methods. Key PointsWe develop a novel histology image-based gene prediction model named HGGEP, which demonstrates high accuracy and robust performance. To reveal the intricate relationship between cell morphology and gene expression in images, we propose a gradient enhancement module, which effectively improves the models capability in perceiving cell morphology in images. HGGEP includes a hypergraph module that efficiently models higher-order associations among features across multiple latent stages, resulting in significant performance improvement.

bioinformatics↗

bioRxiv2024

Load-dependent and load-independent effects on longitudinal motor training in human continuous hand movements

Motor learning involves complex interactions between the cognitive and sensorimotor systems, which is susceptible to different levels of task load. While the mechanism underlying load-dependent regulations in cognitive functions has been extensively investigated, their influence on downstream execution in motor skill learning remains less understood. The current study extends the understanding of how load levels affect motor learning by a longitudinal functional near-infrared spectroscopy (fNIRS) study in which 30 participants (15 females) engaged in extensive practice on a two-dimensional continuous hand tracking task with varying task difficulties. We propose the index of difficulty (ID) as a quantitative estimate of task difficulty, which is positively correlated with psychometric measure of subjective workload level. Results shows that as behavioral performance improved over time, participants adopted a direction-specific and load-independent (i.e., consistent across different load levels) control strategy, shifting from feedback-dominant to feedforward-dominant control in the vertical direction as training progressed. Crucially, we provide robust evidence of the learning-induced alteration in load-dependent cortical activation patterns, suggesting that effective motor skill learning may lead to shift towards an inverted-U relationship between activation and load level in the pre-motor and supplementary motor areas. In addition, brain-behavior relationship in the frontoparietal network was strengthened after training. Taken together, our findings provide new insights into the learning-induced plasticity in brain and behavior associated with load-dependent and load-independent contributions to motor skill learning.

neuroscience↗

bioRxiv2024

Prediction of single-cell chromatin compartments from single-cell chromosome structures by MaxComp

The genome is partitioned into distinct chromatin compartments with at least two main classes, a transcriptionally active A and an inactive B compartment, corresponding mostly to the segregation of euchromatin and heterochromatin. Chromatin within the same compartment has a higher tendency to interact with itself than with regions in opposing compartments. A/B compartments are traditionally derived from ensemble Hi-C contact matrices through principal component analysis of their covariance matrices. However, defining compartments in single cells from single-cell Hi-C maps is non trivial due to sparsity of the data and the fact that homologous copies are typically not resolved. Here we present an unsupervised approach, named MaxComp, to determine single-cell A/B compartments from geometric considerations in 3D chromosome structures, either from multiplexed FISH imaging or from models derived from Hi-C data. By representing each single-cell structure as an undirected graph with edge-weights encoding structural information, the problem of predicting chromosome compartments can be transformed to an alternative form of the Max-cut problem, a semidefinite graph programming method (SPD) to determine an optimal division of a chromosome structure graph into two structural compartments. Our results show that compartment annotations from principal component analysis of ensemble Hi-C data can be perfectly reproduced as population averages of our single-cell compartment predictions. We therefore prove that compartment predictions can be achieved from geometric considerations alone using 3D coordinates of chromatin regions together with information about their nuclear microenvironment. Our results reveal substantial cell-to-cell heterogeneity of compartments in a cell population, which substantially differs between individual genomic regions. Moreover, by applying our approach to multiplexed FISH tracing experiments, our method sheds light on the relationship between single-cell compartment annotations and gene transcriptional activity in single cells. Overall our approach provides new insights into single-cell chromatin condensation, relationship between population and single-cell chromatin compartmentalization, the cell-to-cell variations of chromatin compartments and its impact on gene transcription. Author SummaryChromosome conformation capture and imaging techniques revealed the segregation of genomic chromatin into at least two functional compartments. Hi-C contact frequency matrices show checkerboard-like patterns indicating that chromatin regions are divided into at least two states, possibly a result of phase separation. Chromatin regions in the same state have preferential interactions with each other, often over extended sequence distances, while interactions to regions in the opposing state are minimized. Principal component analysis (PCA) on ensemble Hi-C contact frequency matrices can identify these compartment states. However, because the compartment annotations are derived from a cell population, this method cannot provide information about compartments in single cells. Here in this study, we introduce an unsupervised method to predict single-cell compartments using graph-based programming, which utilizes only structural information in single cells. Our results demonstrate that PCA-based ensemble compartment annotations can be reproduced as population averages of our single-cell compartment predictions. Moreover, our results reveal the cell-to-cell heterogeneity of compartments in a cell population, which shows significant disparities among different chromatin regions. Moreover, by applying our approach to multiplexed FISH tracing experiments, our method reveals the relationship between single-cell compartment annotations and gene transcriptional activity in single cells. Finally, our approach also allows us to relate chromatin structural features in single cells with compartment properties. Comparison with other existing approaches showed that our method produces overall better compartmentalization scores in single cells.

bioinformatics↗