Biology subjects

Chai, H.

Publications and source records attributed to Chai, H..

17 recordsLinked to original sources

bioRxiv2026

Isocaloric restriction of a high-fat diet under TAK-242 administration: metabolic and gut microbial responses in mice

High-fat feeding promotes obesity in mice largely by increasing energy intake, so diet composition and energy intake are hard to separate without pair-feeding. We previously showed that isocaloric restriction normalizes body weight, glycemia and plasma lipids but not visceral adiposity, adipose inflammation or the fecal microbial community. We hypothesized that, during administration of the TLR4 signaling inhibitor TAK-242, matching energy intake to that of low-fat controls would normalize adiposity, intestinal barrier gene expression and the fecal microbial community. Male C57BL/6 mice (n = 8 per group) received a low-fat diet, a high-fat diet ad libitum, or a high-fat diet pair-fed to controls from week 0 or week 5 for 12 weeks; TAK-242 (3 ug/g, three times weekly) was given to every group from week 0.4, with no vehicle-injected arm. Intake did not differ among groups during a three-week low-fat acclimation (P = .88); over the 12 weeks on the assigned diets the ad libitum high-fat group ate 5.5% more than controls (P = .50). Over the full trajectory the ad libitum high-fat group gained 0.48 g/week more than controls (group x week, P = .012); both pair-fed groups had lower fasting glucose than that group. Adipocyte diameter, visceral fat proportion, ileal barrier genes, CCK-induced satiation and microbial alpha diversity did not differ significantly among groups. Adipose Ccl2 expression differed (P = .043, exploratory), as did fecal community composition (PERMANOVA P = .022), although no pairwise contrast survived correction. Energy restriction did not return the microbial community toward the low-fat profile.

physiology↗

bioRxiv2026

DQHTFI: Dynamic-Query Hypergraph Transformer for Fine-Grained Drug-Target Interaction and Affinity Prediction

Drug-target interaction prediction and binding affinity prediction are two key tasks in drug discovery and drug repurposing. Although deep learning methods have made significant progress, existing models typically rely on global representations of drugs and proteins, making it difficult to adequately model fine-grained interactions between their local units. Fixed multimodal fusion strategies also struggle to dynamically adjust the contributions of different modalities for different drug-target combinations. To address these issues, we propose DQHTFI, a fine-grained interaction prediction framework for drug-target interaction classification and binding affinity regression. DQHTFI employs BRICS fragments and Pfam functional domains as the basic interaction units and jointly learns semantic and structural representations. We design a dynamic-query hypergraph Transformer framework in which hyperedges are constructed among the multimodal features of fragment-domain pairs. Dynamic queries are generated from the cross-conditioned features of fragment-domain pairs to adaptively adjust the contribution of each modality, thereby modeling higher-order interactions between local units. Our proposed model achieves competitive results on multiple benchmark datasets.

bioinformatics↗

bioRxiv2026

Cohesin extrudes chromatin loop unidirectionally through two modes of mechanisms in human cells

Cohesin is crucial for establishing 3D genome organization and functions. However, the in vivo mechanism of cohesin-mediated extrusion remains unsettled. Here, we investigated extrusion directionality using integrated genome mapping coupled with perturbation in human cells. By defining cohesin loading at NIPBL binding loci and anchoring at CTCF sites, we systematically characterized the aggregated and single-molecule trajectories of extrusion "stripes" and contact "dots" from and between individual cohesin loading and anchoring points. Our analyses delineated two dynamic modes of in vivo mechanism in which NIPBL/cohesin initiates extrusion at its loading sites in a one-sided manner; after encountering CTCF, CTCF/cohesin switches its extrusion direction backward and robustly extrude DNA unidirectionally until completing the convergent CTCF loop. Surprisingly, depleting CTCF resulted substantial increase of NIPBL/cohesin-mediated extrusion activity, which is further validated by in silico perturbation. These findings reveal novel mechanistic insights into how cohesin and CTCF orchestrate cohesin-mediated loop extrusion in vivo.

genomics↗

bioRxiv2025

GenoME: a MoE-based generative model for individualized, multimodal prediction and perturbation of genomic profiles

The non-coding genome operates through a complex, multiscale regulatory system where regulated gene expressions are closely associated with cell-type-specific histone modifications, transcription factor binding and 3D conformation. Developing computational models that can integrate these patterns to predict and interpret the regulatory system remains challenging. Here, we present GenoME, a Mixture of Experts (MoE)-based generative model that uses DNA sequence and cell-type-specific ATAC-seq signals to predict a unified genomic profile encompassing epigenomics, transcriptomics, and chromatin architecture at base-pair to kilobase resolutions. GenoME enables multiscale predictions for held-out genomic regions and, critically, generalizes to predict the full regulatory landscape of unseen or individualized cell types from a single ATAC-seq input. We equip GenoME with an in silico perturbation framework that accurately forecasts the multimodal consequences of genetic perturbations and identifies functional enhancer-promoter connections, outperforming specialized models like Activity-by-Contact. These predictions can also be used to decipher the transcription factor grammar of cell-type-specific enhancers. GenoME thus provides a versatile, all-in-one platform for generative modeling, cross-cell-type generalization, and causal mechanistic investigation of the multiscale regulatory genome.

bioinformatics↗

bioRxiv2025

ChromMovie: A Molecular Dynamics Approach for Simultaneous Modeling of Chromatin Conformation Changes from Multiple Single-Cell Hi-C Maps

The development of 3C-based techniques for analyzing three-dimensional chromatin structure dynamics has driven significant interest in computational methods for 3D chromatin reconstruction. In particular, models based on Hi-C and its single-cell variants, such as scHi-C, have gained widespread popularity. Current approaches for reconstructing the chromatin structure from scHi-C data typically operate by processing one scHi-C map at a time, generating a corresponding 3D chromatin structure as output. Here, we introduce an alternative approach to the whole genome 3D chromatin structure reconstruction that builds upon existing methods while incorporating the broader context of dynamic cellular processes, such as the cell cycle or cell maturation. Our approach integrates scHi-C contact data with single-cell trajectory information and is based on applying simultaneous modeling of a number of cells ordered along the progression of a given cellular process. The approach is able to successfully recreate known nuclear structures while simultaneously achieving smooth, continuous changes in chromatin structure throughout the cell cycle trajectory. Although both Hi-C-based chromatin reconstruction and cellular trajectory inference are well-developed fields, little effort has been made to bridge the gap between them. To address this, we present ChromMovie, a comprehensive molecular dynamics framework for modeling 3D chromatin structure changes in the context of cellular trajectories. To our knowledge, no existing method effectively leverages both the variability of single-cell Hi-C data and explicit information from estimated cellular trajectories, such as cell cycle progression, to improve chromatin structure reconstruction.

bioinformatics↗

bioRxiv2024

Spatially informed phenotyping by cyclic-in-situ-hybridization identifies novel fibroblast populations and their pathogenic niches in systemic sclerosis

Spatially non-resolved transcriptomic data identified functionally distinct populations of fibroblasts in health and disease. However, in-depth transcriptional profiling in situ at single-cell resolution has not been possible so far. Here, we studied fibroblast populations in the skin of SSc patients and healthy individuals using cyclic in situ hybridization (cISH) as a novel approach for spatially-resolved transcriptional phenotyping with subcellular resolution. cISH deconvoluted the heterogeneity of 20,979 cells including 3,764 fibroblasts (FB). BANKSY-based spatially-informed clustering identified nine FB subpopulations, with SFRP2+ RetD FB and CCL19+ nonPV FB as novel subpopulations that reside in specific cellular niches and display unique gene expression profiles. SFRP2+ RetD FB and CCL19+ nonPV FB as well as COL8A1+ FB, display altered frequencies in SSc skin and play specific, disease-promoting roles for extracellular matrix release and leukocyte recruitment as revealed by their transcriptional profile, their cellular interactions and ligand-receptor analyses. The frequencies of COL8A1+ FB and their interactions with monocytic cells and B cells are associated with progression of skin fibrosis in SSc. In summary, our spatially-resolved transcriptomic approach identified novel fibroblast subpopulations deregulated in SSc skin with specific pathogenic roles, some of which may potentially serve as biomarkers for progression of skin fibrosis.

immunology↗

bioRxiv2024

Lineage-aware evolutionary analysis of hepatitis C virus within-host dynamics

Analysis of viral genetic data has previously revealed distinct within-host population structures in both untreated and interferon-treated chronic hepatitis C virus (HCV) infections. While multiple subpopulations persisted during the infection, each subpopulation was observed only intermittently. However, it was unknown whether similar patterns were also present after Direct Acting Antiviral (DAA) treatment, where viral populations were often assumed to go through narrow bottlenecks. Here we tested for the maintenance of population structure after DAA treatment failure. We analysed whole-genome next-generation sequencing data generated from a randomised study using DAAs (the BOSON study). We focused on samples collected from patients (N=84) who did not achieve sustained virological response (i.e. treatment failure) and had sequenced virus from multiple timepoints. For each individual, we tracked concordance in nucleotide variant frequencies through time. Using a sliding window approach, we applied sequenced-based and tree-based clustering algorithms across the entire HCV genome. Finally, we reconstructed viral haplotypes and estimated lineage specific within-host divergence rates from the haplotype phylogenies. Distinct viral subpopulations were maintained among a high proportion of individuals post DAA treatment failure. Using maximum likelihood modelling and model comparison, we found an overdispersion of viral evolutionary rates among individuals, and significant differences in evolutionary rates between lineages within individuals. These results suggest the virus is compartmentalised within individuals, with the varying evolutionary rates due to different viral replication rates or different selection pressures. We propose lineage awareness in future analyses of HCV evolution and infections to avoid conflating patterns from distinct lineages, and to recognise the likely existence of unsampled subpopulations.

evolutionary biology↗

bioRxiv2024

An integrated view of the structure and function of the human 4D nucleome

The dynamic three-dimensional (3D) organization of the human genome (the "4D Nucleome") is closely linked to genome function. Here, we integrate a wide variety of genomic data generated by the 4D Nucleome Project to provide a detailed view of human 3D genome organization in widely used embryonic stem cells (H1-hESCs) and immortalized fibroblasts (HFFc6). We provide extensive benchmarking of 3D genome mapping assays and integrate these diverse datasets to annotate spatial genomic features across scales. The data reveal a rich complexity of chromatin domains and their sub-nuclear positions, and over one hundred thousand structural loops and promoter-enhancer interactions. We developed 3D models of population-based and individual cell-to-cell variation in genome structure, establishing connections between chromosome folding, nuclear organization, chromatin looping, gene transcription, and DNA replication. We demonstrate the use of computational methods to predict genome folding from DNA sequence, uncovering potential effects of genetic variants on genome structure and function. Together, this comprehensive analysis contributes insights into human genome organization and enhances our understanding of connections between the regulation of genome function and 3D genome organization in general.

genomics↗

bioRxiv2024

Tri-omic mapping revealed concerted dynamics of 3D epigenome and transcriptome in brain cells

Exploring the genomic basis of transcriptional programs has been a longstanding research focus. Here, we report a high-throughput single-cell tri-omic method to capture chromatin accessibility, interaction, and RNA simultaneously (ChAIR). After validating in cultured cells, we applied ChAIR to brain cells across mouse lifespan and delineated the concerted dynamics of 3D-epigenomic architecture and transcription during maturation and aging. Particularly, ultra-long chromatin megacontacts and promoter-associated 3D-epigenomic states are effective in defining cell identity and revealing spatially-resolved anatomic specificity. Importantly, we found that neurons in different brain regions and non-neuronal cells may undergo divergent genomic mechanisms during differentiation and aging. Our results demonstrated ChAIRs robustness of connecting chromatin folding architecture with cellular property and its potential applications to address complex questions in single-cell resolution and spatial specificity.

genomics↗

bioRxiv2024

Multifaceted roles of cohesin in regulating transcriptional loops

Cohesin is required for chromatin loop formation. However, its precise role in regulating gene transcription remains largely unknown. We investigated the relationship between cohesin and RNA Polymerase II (RNAPII) using single-molecule mapping and live-cell imaging methods in human cells. Cohesin-mediated transcriptional loops were highly correlated with those of RNAPII and followed the direction of gene transcription. Depleting RAD21, a subunit of cohesin, resulted in the loss of long-range (>100 kb) loops between distal (super-)enhancers and promoters of cell-type-specific genes. By contrast, the short-range (<50 kb) loops were insensitive to RAD21 depletion and connected genes that are mostly housekeeping. This result explains why only a small fraction of genes are affected by the loss of long-range chromatin interactions due to cohesin depletion. Remarkably, RAD21 depletion appeared to up-regulate genes located in early initiation zones (EIZ) of DNA replication, and the EIZ signals were amplified drastically without RAD21. Our results revealed new mechanistic insights of cohesins multifaceted roles in establishing transcriptional loops, preserving long-range chromatin interactions for cell-specific genes, and maintaining timely order of DNA replication.

genomics↗

bioRxiv2023

Pan-genotypic probe-based enrichment to improve efficiency of Hepatitis B virus sequencing

Hepatitis B Virus (HBV) genome sequencing can be used to provide more complete genetic information at the population and individual level to shed light on the limitations of current interventions, and inform new strategies for elimination. HBV sequencing is challenging due to the partially dsDNA genome, high diversity, low viral loads and presence of large amounts of host genetic material in clinical samples. Here we describe the design and use of a pan-genotypic panel of 74 HBV specific capture-probes and nuclease treatment in improving sequencing efficiency. We processed 20 plasma samples (viral loads 1.98 to 4.07 log10, genotypes A-E) and three positive controls (human total brain RNA and bacteriophage lambda DNA) in triplicate to compare DNAse vs. RNAse vs. no nuclease treatment. We prepared libraries using the Takara Bio SMARTer Stranded Total RNA-Seq Kit v3, split the library in two, enriching half with the custom-designed probe panel and xGen Hybridization and Wash Kit (IDT), the other half was not enriched. Both libraries were sequenced on the NovaSeq6000 platform with 2x150nt paired-end reads. Capture resulted in a 47,970 fold increase in the number of reads mapped to the HBV genome in the "no nuclease" arm (243 HBV reads per million reads sequenced in the capture pool vs. 5x10-3 reads per million in the no-capture pool). Out of 20 samples, only 1 without capture generated HBV reads (viral load 3.89 log10 IU/ml) vs. 19 samples with capture. HBV sequence yield was increased in the capture arm and resulted in 2.30 log10 (95% confidence interval 1.99 - 2.48 log10) increase in HBV reads (per million reads sequenced) per log10 increase in viral load. The proportion of HBV reads increased a median of 12 fold with RNAse treatment. We developed a targeted pan-genotypic sequencing method using a custom panel of biotinylated oligos that increases the sequencing efficacy of HBV. This method will allow us to gain a better insight into HBV diversity.

genomics↗

bioRxiv2022

An uncertainty-based interpretable deep learning framework for breast cancer outcomes prediction

Accurate prediction of breast cancer outcomes is important for selecting appropriate treatment, which can prolong the survival period of the patients and improve the life quality. Recently, different deep learning-based methods are carefully designed for cancer outcomes prediction. However, the applications of these methods are still challenging due to the model interpretability. In this study, we proposed a novel multi-task deep neural network UISNet to interpret the feature importance of the prediction model by an uncertainty-based integrated gradients algorithm. Additionally, UISNet improves the prediction accuracy by introducing the prior biological pathway knowledge and utilizing the patients heterogeneity information. By applications to seven breast cancer public datasets, the method was shown to outperform state-of-the-art methods by achieving a 5.79% higher C-index value on average. For the identified genes based on the interpretable model, 11 out of the top 20 genes have been proved to be associated with breast cancer by literature review. The comprehensive tests indicated that our proposed method is accurate and robust to predict breast cancer outcomes, and is an effective way to identify the prognosis-related genes. The method codes are available at: https://github.com/chh171/UISNet.

bioinformatics↗

bioRxiv2022

Strong and consistent associations of waterbird community composition with HPAI H5 occurrence in European wild birds

Since 2014, highly pathogenic avian influenza (HPAI) H5 viruses of clade 2.3.4.4 have been dominating the outbreaks across Europe, causing massive deaths among poultry and wild birds. However, the factors shaping its broad-scale outbreak patterns remain unclear. With extensive waterbird survey datasets of about 7,000 sites across Europe, we here demonstrated that H5N8 occurrence in wild birds in the 2016/17 and 2020/21 epidemics as well as H5N1 occurrence in 2005/06 epidemic were strongly associated with very similar waterbird community attributes, pointing to the possibility of similar interspecific transmission processes between different epidemics. A simple extrapolation of the model constructed from the 2016/17 epidemic can well predict the H5N8 pattern in wild birds in 2020/21 epidemic. We also found a dilution effect of phylogenetic diversity that was always negatively correlated with H5 occurrence in wild birds. In contrast, H5N8 occurrence in poultry was subject to different risk factors between the two epidemics. In general, waterbird community composition play a much more important role in determining the spatial pattern of H5N8 in wild birds than in poultry. Our work contributes to reveal the factors driving H5N8 patterns, and highlights the value of waterbird community factors in future HPAI surveillance and prediction.

ecology↗

bioRxiv2021

Defining the Characteristics of Type I Interferon Stimulated Genes: Insight from Expression Data and Machine Learning

A virus-infected cell triggers a signalling cascade resulting in the secretion of interferons (IFNs), which in turn induce the up-regulation of IFN-stimulated genes (ISGs) that play an important role in the inhibition of the viral infection and the return to cellular homeostasis. Here, we conduct detailed analyses on 7443 features relating to evolutionary conservation, nucleotide composition, gene expression, amino acid composition, and network properties to elucidate factors associated with the stimulation of genes in response to type I IFNs. Our results show that ISGs are less evolutionary conserved than genes that are not significantly stimulated in IFN experiments (non-ISGs). ISGs show significant depletion of GC-content in the coding region of their canonical transcripts, which leads to under-representation in the nucleotide compositions. Differences between ISGs and non-ISGs are also reflected in the properties of their coded amino acid sequence compositions. Network analyses show that ISG products tend to be involved in key paths but are away from hubs or bottlenecks of the human protein-protein interaction (PPI) network. Our analyses also show that interferon-repressed human genes (IRGs), which are down-regulated in the presence of IFNs, can have similar properties to ISGs, thus leading to false positives in ISG predictions. Based on these analyses, we design a machine learning framework integrating the usage of support vector machine (SVM) and feature selection algorithms. The ISG prediction achieves an area under the receiver operating characteristic curve (AUC) of 0.7455 and demonstrates the similarity between ISGs triggered by type I and III IFNs. Our machine learning model predicts a number of genes as potential ISGs that so far have shown no significant differential expression when stimulated with IFN in the cell types and tissue types compiled in the available IFN-related databases. A webserver implementing our method is accessible at http://isgpre.cvr.gla.ac.uk/. Author summaryInterferons (IFNs) are signalling proteins secreted from host cells. IFN-triggered signalling activates the host immune system in response to intra-cellular infection. It results in the stimulation of many genes that have anti-pathogen roles in host defenses. Interferon-stimulated genes (ISGs) have unique properties that make them different from those not significantly up-regulated in response to IFNs (non-ISGs). We find the down-regulated interferon-repressed genes (IRGs) have some shared properties with ISGs. This increases the difficulty of distinguishing ISGs from non-ISGs. The use of machine learning is a sensible strategy to provide high throughput classifications of putative ISGs, for investigation with in vivo or in vitro experiments. Machine learning can also be applied to human genes for which there are insufficient expression levels before and after IFN treatment in various experiments. Additionally, the interferon type has some impact on ISG predictability. We expect that our study will provide new insight into better understanding the inherent characteristics of human genes that are related to response in the presence of IFNs.

immunology↗

bioRxiv2021

scAdapt: Virtual adversarial domain adaptation network for single cell RNA-seq data classification across platforms and species

MotivationIn single cell analyses, cell types are conventionally identified based on known marker gene expressions. Such approaches are time-consuming and irreproducible. Therefore, many new supervised methods have been developed to identify cell types for target datasets using the rapid accumulation of public datasets. However, these approaches are sensitive to batch effects or biological variations since the data distributions are different in cross-platforms or species predictions. ResultsWe developed scAdapt, a virtual adversarial domain adaptation network to transfer cell labels between datasets with batch effects. scAdapt used both the labeled source and unlabeled target data to train an enhanced classifier, and aligned the labeled source centroid and pseudo-labeled target centroid to generate a joint embedding. We demonstrate that scAdapt outperforms existing methods for classification in simulated, cross-platforms, cross-species, and spatial transcriptomic datasets. Further quantitative evaluations and visualizations for the aligned embeddings confirm the superiority in cell mixing and preserving discriminative cluster structure present in the original datasets. Availabilityhttps://github.com/zhoux85/scAdapt. Contactangyd25@mail.sysu.edu.cn or luojinx5@mail.sysu.edu.cn

bioinformatics↗

bioRxiv2019

Integrating multi-omics data with deep learning for predicting cancer prognosis

BackgroundGenomic information is nowadays widely used for precise cancer treatments. Since the individual type of omics data only represents a single view that suffers from data noise and bias, multiple types of omics data are required for accurate cancer prognosis prediction. However, it is challenging to effectively integrate multi-omics data due to the large number of redundant variables but relatively small sample size. With the recent progress in deep learning techniques, Autoencoder was used to integrate multi-omics data for extracting representative features. Nevertheless, the generated model is fragile from data noises. Additionally, previous studies usually focused on individual cancer types without making comprehensive tests on pan-cancer. Here, we employed the denoising Autoencoder to get a robust representation of the multi-omics data, and then used the learned representative features to estimate patients risks. ResultsBy applying to 15 cancers from The Cancer Genome Atlas (TCGA), our method was shown to improve the C-index values over previous methods by 6.5% on average. Considering the difficulty to obtain multi-omics data in practice, we further used only mRNA data to fit the estimated risks by training XGboost models, and found the models could achieve an average C-index value of 0.627. As a case study, the breast cancer prognosis prediction model was independently tested on three datasets from the Gene Expression Omnibus (GEO), and shown able to significantly separate high-risk patients from low-risk ones (C-index>0.6, p-values<0.05). Based on the risk subgroups divided by our method, we identified nine prognostic markers highly associated with breast cancer, among which seven genes have been proved by literature review. ConclusionOur comprehensive tests indicated that we have constructed an accurate and robust framework to integrate multi-omics data for cancer prognosis prediction. Moreover, it is an effective way to discover cancer prognosis-related genes.

bioinformatics↗

bioRxiv2019

Imputing missing RNA-seq data from DNA methylation by using transfer learning based-deep neural network

BackgroundGene expression plays a key intermediate role in linking molecular features at DNA level and phenotype. However, due to various limitations in experiments, the RNA-seq data is missing in many samples while there exists high-quality of DNA methylation data. As DNA methylation is an important epigenetic modification to regulate gene expression, it can be used to predict RNA-seq data. For this purpose, many methods have been developed. A common limitation of these methods is that they mainly focus on single cancer dataset, and do not fully utilize information from large pan-cancer dataset. ResultsHere, we have developed a novel method to impute missing gene expression data from DNA methylation data through transfer learning-based neural network, namely TDimpute. In the method, the pan-cancer dataset from The Cancer Genome Atlas (TCGA) was utilized for training a general model, which was then fine-tuned on the specific cancer dataset. By testing on 16 cancer datasets, we found that our method significantly outperforms other state-of-the-art methods in imputation accuracy with 7%-11% increase under different missing rates. The imputed gene expression was further proved to be useful for downstream analyses, including the identification of both methylation-driving and prognosis-related genes, clustering analysis, and survival analysis on the TCGA dataset. More importantly, our method was indicated to be useful for general purpose by the independent test on the Wilms tumor dataset from the Therapeutically Applicable Research To Generate Effective Treatments (TARGET) project. ConclusionsTDimpute is an effective method for RNA-seq imputation with limited training samples.

bioinformatics↗