Biology subjects

Peng, J.

Publications and source records attributed to Peng, J..

At least 109 records · Page 6Linked to original sources

bioRxiv2024

Identifying potential risk genes for clear cell renal cell carcinoma with deep reinforcement learning

Clear cell renal cell carcinoma (ccRCC) is the most prevalent type of renal cell carcinoma. However, our understanding of ccRCC risk genes remains limited. This gap in knowledge poses significant challenges to the effective diagnosis and treatment of ccRCC. To address this problem, we propose a deep reinforcement learning-based computational approach named RL-GenRisk to identify ccRCC risk genes. Distinct from traditional supervised models, RL-GenRisk frames the identification of ccRCC risk genes as a Markov decision process, combining the graph convolutional network and Deep Q-Network for risk gene identification. Moreover, a well-designed data-driven reward is proposed for mitigating the lim-itation of scant known risk genes. The evaluation demonstrates that RL-GenRisk outperforms existing methods in ccRCC risk gene identification. Additionally, RL-GenRisk identifies ten novel ccRCC risk genes. We successfully validated epidermal growth factor receptor (EGFR), corroborated through independent datasets and biological experimentation. This approach may also be used for other diseases in the future.

bioinformatics↗

bioRxiv2024

DiffuST: a latent diffusion model for spatial transcriptomics denoising

Spatial transcriptomics technologies have enabled comprehensive measurements of gene expression profiles while retaining spatial information and matched pathology images. However, noise resulting from low RNA capture efficiency and experimental steps needed to keep spatial information may corrupt the biological signals and obstruct analyses. Here, we develop a latent diffusion model DiffuST to denoise spatial transcriptomics. DiffuST employs a graph autoencoder and a pre-trained model to extract different scale features from spatial information and pathology images. Then, a latent diffusion model is leveraged to map different scales of features to the same space for denoising. The evaluation based on various spatial transcriptomics datasets showed the superiority of DiffuST over existing denoising methods. Furthermore, the results demonstrated that DiffuST can enhance downstream analysis of spatial transcriptomics and yield significant biological insights.

genomics↗

bioRxiv2024

Cold seeps are hotspots of deep-sea nitrogen-loss driven by microorganisms across 21 phyla

Nitrogen bioavailability, governed by the balance of fixation and loss processes, is a key factor regulating oceanic productivity, ecosystem functions, and global biogeochemical cycles. The key nitrogen-loss organisms--denitrifiers and anaerobic ammonium-oxidizing (anammox) bacteria--are not well understood in marine seafloor environments, especially in deep-sea cold seeps. In this study, we combined geochemical measurements, 15N stable isotope tracer analysis, metagenomics, metatranscriptomics, and three-dimensional protein structural simulations to investigate the diversity of denitrifying and anammox microbial communities and their biogeochemical roles in these habitats. Geochemical evidence from 301 sediment samples shows significantly higher nitrogen-loss rates in cold seeps compared to typical deep-sea sediments, with an estimated annual nitrogen loss of 6.16 Tg from seafloor surface sediments. Examination of a total of 147 million non-redundant genes reveals a high prevalence and active expression of nitrogen-loss genes, including nitrous-oxide reductase (NosZ; 6.88 genes per million or GPM on average), nitric oxide dismutase (Nod; 1.29 GPM), and hydrazine synthase (HzsA; 3.35 GPM) in surface sediments. Analysis of 3,164 metagenome-assembled genomes from this habitat has expanded the known diversity of nitrous-oxide reducers to six phyla and nitric oxide-dismutating organisms to one phylum and two new orders, while ten phyla host anammox bacteria going beyond Planctomycetota. These microbes show diverse structural adaptations and complex gene cluster arrangements that potentially enable survival in the harsh conditions of cold seeps. These findings suggest that cold seeps, despite their low temperatures, are significant, previously underestimated hotspots of nitrogen loss, potentially contribute substantially to the global nitrogen cycle.

microbiology↗

bioRxiv2024

CellAgent: An LLM-driven Multi-Agent Framework for Automated Single-cell Data Analysis

Single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics (ST) data analysis are pivotal for advancing biological research, enabling precise characterization of cellular heterogeneity. However, existing analysis approaches require extensive manual programming and tool manipulation, posing significant challenges for researchers. To address this, we introduce CellAgent, an autonomous, LLM-driven approach that performs end-to-end scRNA-seq and spatial transcriptomics data analysis through natural language interactions. CellAgent employs a multi-agent hierarchical decision-making framework, simulating a "deep-thinking" workflow to ensure that each analytical step remains consistent with the overall task objective. To further enhance its capabilities, we developed sc-Omni, a high-performance, expert-curated toolkit that consolidates essential tools for scRNA-seq and spatial transcriptomics analysis. Additionally, we introduce a self-reflective optimization mechanism, enabling automated, iterative refinement of results through specialized evaluation methods, effectively replacing traditional manual assessments. Benchmarking against human experts demonstrates that CellAgent achieves approximately 60% improvement in efficiency across multiple downstream applications. In terms of accuracy, it maintains performance comparable to existing approaches while preserving natural language interactions. By translating natural language interactions into optimized analytical workflows, CellAgent establishes a scalable paradigm for LLM-driven scientific discovery, bridging the gap between experimental biologists and complex data analytics. This framework minimizes reliance on manual coding and exhaustive deliberation, ushering in the era of the "AI Agent for Science."

bioinformatics↗

bioRxiv2024

Rapid metabolic regulation of a novel arginine methylation of KCa3.1 attenuates T cell exhaustion.

T cell receptor (TCR) engagement initiates the activation process, and this signaling event is regulated in multifaceted ways. Nutrient availability in the immediate niche is one such mode of regulation1-3. Here, we investigated how the availability of an essential amino acid methionine (Met) and TCR signaling might interplay in the earliest events of T cell activation to affect subsequent T cell fate and function. We found that limiting Met during only the initial 30 minutes of CD8+ T cell activation increased Ca2+ influx, Ca2+-mediated NFAT1 (Nfatc2) activation, NFAT1 promoter occupancy, and T cell exhaustion. We identified changes in the protein arginine methylome during the initial 30 min of TCR engagement and discovered a novel arginine methylation of a Ca2+-activated potassium transporter, KCa3.1, which regulates Ca2+-mediated NFAT1 signaling to ensure optimal activation. Ablation of arginine methylation in KCa3.1 led to increased NFAT1 activation, rendering T cells dysfunctional in murine tumour and infection models. Furthermore, acute Met supplementation at early stages reduced nuclear NFAT1 in tumour-infiltrating T cells and augmented their anti-tumour activity. Our findings identify a metabolic event occurring early after T cell activation that influences the subsequent fate of the cell.

immunology↗

bioRxiv2024

Linear ubiquitin chains remodel the proteome and influence the levels of hundreds of regulators in Drosophila

Ubiquitin controls many cellular processes via its post-translational conjugation onto substrates. Its use is highly variable due to its ability to form poly-ubiquitin with various topologies. Among them, linear chains have emerged as important regulators of immune responses and protein degradation. Previous studies in Drosophila melanogaster found that expression of linear poly-ubiquitin that cannot be dismantled into single moieties leads to their own ubiquitination and degradation or, alternatively, to their conjugation onto proteins. However, it remains largely unknown which proteins are sensitive to linear poly-ubiquitin. To address this question, here we expanded the toolkit to modulate linear chains and conducted ultra-deep coverage proteomics from flies that express non-cleavable, linear chains comprising 2, 4, or 6 moieties. We found that these chains regulate shared and distinct cellular processes in Drosophila by impacting hundreds of proteins. Our results provide key insight into the proteome subsets and cellular pathways that are influenced by linear poly-ubiquitin with distinct lengths and suggest that the ubiquitin system is exceedingly pliable.

cell biology↗

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↗

bioRxiv2024

A systems-biology approach connects aging mechanisms with Alzheimer's disease pathogenesis

Despite years of intense investigation, the mechanisms underlying neuronal death in Alzheimers disease, the most common neurodegenerative disorder, remain incompletely understood. To define relevant pathways, we integrated the results of an unbiased, genome-scale forward genetic screen for age-associated neurodegeneration in Drosophila with human and Drosophila Alzheimers disease-associated multi-omics. We measured proteomics, phosphoproteomics, and metabolomics in Drosophila models of Alzheimers disease and identified Alzheimers disease human genetic variants that modify expression in disease-vulnerable neurons. We used a network optimization approach to integrate these data with previously published Alzheimers disease multi-omic data. We computationally predicted and experimentally demonstrated how HNRNPA2B1 and MEPCE enhance tau-mediated neurotoxicity. Furthermore, we demonstrated that the screen hits CSNK2A1 and NOTCH1 regulate DNA damage in Drosophila and human iPSC-derived neural progenitor cells. Our work identifies candidate pathways that could be targeted to ameliorate neurodegeneration in Alzheimers disease.

systems biology↗

bioRxiv2024

Human brain aging heterogeneity observed from multi-region omics data reveals a subtype closely related to Alzheimer's disease

INTRODUCTION: The interconnection between brain aging and Alzheimers disease (AD) remain to be elucidated. METHODS: We investigated multi-omics (transcriptomics and proteomics) data from multiple brain regions (i.e., the hippocampus (HIPP), prefrontal cortex (PFC), and cerebellum (CRBL)) in cognitively normal individuals. RESULTS: We found that brain samples could be divided into ADL (AD-like) and NL (normal) subtypes which were correlated across brain regions. The differentially expressed genes in the ADL samples highly overlapped with AD gene signatures and the changes were consistent across brain regions (PFC and HIPP) in the multi-omics data. Intriguingly, the ADL subtype in PFC showed more differentially expressed genes than other brain regions, which could be explained by the baseline gene expression differences in the PFC NL samples. DISCUSSION: We conclude that brain aging heterogeneity widely exists, and our findings corroborate with the hypothesis that AD-related changes occur decades before the clinical manifestation of cognitive impairment in a sub-population.

neuroscience↗

bioRxiv2024

Deep sequencing of proteotoxicity modifier genes uncovers a Presenilin-2/beta-amyloid-actin genetic risk module shared among alpha-synucleinopathies

Whether neurodegenerative diseases linked to misfolding of the same protein share genetic risk drivers or whether different protein-aggregation pathologies in neurodegeneration are mechanistically related remains uncertain. Conventional genetic analyses are underpowered to address these questions. Through careful selection of patients based on protein aggregation phenotype (rather than clinical diagnosis) we can increase statistical power to detect associated variants in a targeted set of genes that modify proteotoxicities. Genetic modifiers of alpha-synuclein ([a]S) and beta-amyloid (A{beta}) cytotoxicity in yeast are enriched in risk factors for Parkinsons disease (PD) and Alzheimers disease (AD), respectively. Here, along with known AD/PD risk genes, we deeply sequenced exomes of 430 [a]S/A{beta} modifier genes in patients across alpha-synucleinopathies (PD, Lewy body dementia and multiple system atrophy). Beyond known PD genes GBA1 and LRRK2, rare variants AD genes (CD33, CR1 and PSEN2) and A{beta} toxicity modifiers involved in RhoA/actin cytoskeleton regulation (ARGHEF1, ARHGEF28, MICAL3, PASK, PKN2, PSEN2) were shared risk factors across synucleinopathies. Actin pathology occurred in iPSC synucleinopathy models and RhoA downregulation exacerbated [a]S pathology. Even in sporadic PD, the expression of these genes was altered across CNS cell types. Genome-wide CRISPR screens revealed the essentiality of PSEN2 in both human cortical and dopaminergic neurons, and PSEN2 mutation carriers exhibited diffuse brainstem and cortical synucleinopathy independent of AD pathology. PSEN2 contributes to a common-risk signal in PD GWAS and regulates [a]S expression in neurons. Our results identify convergent mechanisms across synucleinopathies, some shared with AD.

genomics↗

bioRxiv2024

HKDC1 Promotes Liver Cancer Stemness Under Hypoxia via Stabilizing β-Catenin

Background and AimsHexokinases (HKs), a group of enzymes catalyzing the first step of glycolysis, have been shown to play important roles in liver metabolism and tumorigenesis. Our recent studies identified HKDC1 as a top candidate associated with liver cancer metastasis. We aimed to compare its cell-type specificity with other HKs upregulated in liver cancer and investigate the molecular mechanisms underlying its involvement in liver cancer metastasis. Approach and ResultsWe found that, compared to HK1 and HK2, the other two commonly upregulated HKs in liver cancer, HKDC1 was most strongly associated with the metastasis potential of tumors and organoids derived from two liver cancer mouse models we previously established. RNA in situ hybridization and single-cell RNA-seq analysis revealed that HKDC1 was specifically upregulated in malignant cells in hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA) patient tumors, whereas HK1 and HK2 were widespread across various tumor microenvironment lineages. An unbiased metabolomic profiling demonstrated that HKDC1 overexpression in HCC cells led to metabolic alterations distinct from those from HK1 and HK2 overexpression, with HKDC1 particularly impacting the tricarboxylic acid (TCA) cycle. HKDC1 was prometastatic in HCC orthotopic and tail vein injection mouse models and, molecularly, HKDC1 was induced by hypoxia and bound to glycogen synthase kinase 3{beta} to stabilize {beta}-catenin, leading to enhanced stemness of HCC cells. ConclusionsOverall, our findings underscore HKDC1 as a prometastatic HK specifically expressed in the malignant compartment of primary liver tumors, thereby providing a mechanistic basis for targeting this enzyme in advanced liver cancer.

cancer biology↗

bioRxiv2024

Genetic Modulation of Protein Expression in Rat Brain

Genetic variations in protein expression are implicated in a broad spectrum of common diseases and complex traits. However, the fundamental genetic architecture and variation of protein expression have received comparatively less attention than either mRNA or classical phenotypes. In this study, we systematically quantified proteins in the brains of a large family of rats using tandem mass tag (TMT)-based quantitative mass-spectrometry (MS) technology. We identified and quantified a comprehensive proteome of 8,119 proteins from Spontaneously Hypertensive (SHR/Olalpcv), Brown Norway with polydactyly-luxate (BN-Lx/Cub), and 29 of their fully inbred HXB/BXH progeny. Differential expression (DE) analysis identified 597 proteins with significant differences in expression between the parental strains (fold change > 2 and FDR < 0.01). We characterized 95 variant peptides by proteogenomics approach and discovered 464 proteins linked to strong cis-acting quantitative trait loci (pQTLs, FDR < 0.05). We also explored the linkage of pQTLs with behavioral phenotypes in rats and examined the sex-specific pQTLs to reveal both distinct and shared cis-pQTLs between sexes. Furthermore, by creating a novel view of the rat pangenome, we improved the ability to pinpoint candidate genes underlying pQTL. Finally, we explored the connection between the pQTLs in rat and human disorders, underscoring the translational potential of our findings. Collectively, this work demonstrates the value of large and systematic proteo-genetic datasets in understanding protein modulation in the brain and its functional linkage to complex central nervous system (CNS) traits.

genetics↗

bioRxiv2024

Sex differences in immune function and disease risk are not easily explained by an evolutionary mismatch

The "Pregnancy Compensation Hypothesis" (PCH) posits that sex differences in the mammalian immune system reflect the effects of selection on female immunity during pregnancy, during which increased immunomodulation is required to reduce immune responses to the fetus, while maintaining the ability to respond to pathogens. In humans, fewer pathogens and lower parity in urban, industrialized environments has been suggested to leave female immune systems under-stimulated, creating an "evolutionary mismatch" which exacerbates sex differences in immunity and increases female autoimmune disease risk. Yet, robust tests of this mismatch hypothesis have not been conducted. Here, we first confirmed a sex-bias in autoimmune disease prevalence in a large dataset of individuals from the United Kingdom (UK Biobank). Second, we asked whether sex differences in immune function are affected by shifts toward urban, lower-parity lifestyles in a single population--the Turkana people of northwest Kenya. We found that lifestyle alters sex differences in immune cell type proportions, but not in gene expression levels. Contrary to expectations from the PCH, parity did not predict immune physiology. We then tested for generalizable trends across natural fertility mammalian populations versus urban humans (data from Turkana, NHANES, GTEx, Batwa and Bakiga, yellow baboons, and macaques). We did not find consistent relationships between lifestyle and sex-biases in immune biomarkers or between parity and the same immune outcomes. Indeed, we found that parity predicts autoimmune disease risk in the opposite direction than expected from the PCH, with higher parity associated with higher autoimmune disease risk while adjusting for relevant socioeconomic variables in the UK Biobank. Taken together, our work suggests that while sex clearly influences immune physiology and disease risk in urban settings, these effects are not easily explained by an evolutionary mismatch. Future work addressing how sex interacts with lifestyle change to generate disease is needed.

evolutionary biology↗

bioRxiv2024

Proteomic profiling of intracranial atherosclerotic plaque in the human brain

BackgroundIntracranial atherosclerotic disease (ICAD) is one of the major causes of ischemic stroke and associated with high risk of stroke recurrence. There are no reliable and specific fluid biomarkers for ICAD, and little is known about the proteomic profiling of ICAD. In this study we aimed to explore the feasibility of applying proteomics technology to profile intracranial atherosclerotic plaques extracted from postmortem human brain arteries. MethodsEighteen segments (5-10mm in length) of major arteries from 10 postmortem brains were collected from the Mount Sinai Neuropathology Brain Bank. Among these segments, 5 had no evidence of atherosclerotic disease, and 13 had wall thickening or visible plaques with various degree of stenosis. Proteins were extracted from the vessel segments, quantified, and digested into peptides. Subsequently, the peptides underwent tandem mass tag (TMT) labeling, pooling, and analysis using two-dimensional liquid chromatography-tandem mass spectrometry (LC/LC-MS/MS). Protein identification and quantification were performed using the JUMP software. Differentially expressed proteins (DEPs) were defined as proteins with p.adj < 0.05 and absolute log2 (fold change) > log2 (1.2). ResultsA total of 7,492 unique proteins were detected, and 6,726 quantifiable proteins were retained for further analysis. Among these, 265 DEPs, spanning on 252 unique gene, were found to be associated with ICAD by comparing the arterial segments with vs those without atherosclerotic disease. The top 4 most significant DEPs include LONP1, RPS19, MRPL12 and SNU13. Among the top 50 DEPs, FADD, AIFM1 and PGK1 were associated with atherosclerotic disease or cardiovascular events in previous studies. Moreover, the previously reported proteins associated with atherosclerosis such as APCS, MMP12, CTSD were elevated in arterial segments with atherosclerotic changes. Furthermore, the up-regulation of APOE and LPL, the ICAD GWAS risk genes, was shown to be associated with the plaque severity. Finally, gene set enrichment analysis revealed the DEP signature is enriched for biological pathways such as chromatin structure, plasma lipoprotein, nucleosome, and protein-DNA complex, peroxide catabolic and metabolic processes, critical in ICAD pathology. ConclusionsDirect proteomic profiling of fresh-frozen intracranial artery samples by MS-based proteomic technology is a feasible approach to identify ICAD-associated proteins, which can be potential biomarker candidates for ICAD. Further plaque proteomic study in a larger sample size is warranted to uncover mechanistic insights into ICAD and discover novel biomarkers that may help to improve diagnosis and risk stratification in ICAD.

systems biology↗

bioRxiv2024

Phylogenetically and structurally diverse reductive dehalogenases link biogeochemical cycles in deep-sea cold seeps

Reductive dehalogenation is crucial for halogen cycling and environmental remediation, yet its ecological role is not completely understood, especially in deep-sea environments. To address this gap, we investigated the diversity of reductive dehalogenases (RDases) and ecophysiology of organohalide reducers in deep-sea cold seeps, which are environments rich in halogenated compounds. Through genome- resolved metagenomic analysis of 165 global cold seep sediment samples, we identified four types of RDases, namely prototypical respiratory, transmembrane respiratory, and cytosolic RDases, and one novel clade. Cold seeps were found to harbor a higher abundance of these RDase genes compared to other marine sediments, establishing them as unique hotspots for microbial reductive dehalogenation activity. These RDases are encoded by a wide range of microorganisms across four archaeal and 36 bacterial phyla, significantly expanding the known diversity of organohalide reducers. Halogen geochemistry, laboratory incubations with various halogenated compounds, metatranscriptomic data, and metabolomic profiling confirmed the presence of organohalides at concentrations of up to 18 mg/g in these sediments and demonstrated their active microbial reduction. This process is tightly linked to other biogeochemical cycles, including carbon, hydrogen, nitrogen, sulfur, and trace elements. RDases from cold seeps have diverse N-terminal structures across different gene groups, and reductive dehalogenase genes in these environments are mostly functionally constrained and conserved. These findings collectively suggest that reductive dehalogenation is a central process in deep-sea environments, mediated by a diverse array of microbes and novel enzymes.

microbiology↗

bioRxiv2024

Early evolution of the ecdysozoan body plan

Extant ecdysozoans (moulting animals) are represented by a great variety of vermiform or articulated organisms. However, controversies remain about the nature of their ancestral body plan although the vermiform hypothesis seems to prevail. We describe here Beretella spinosa gen et sp. nov. a tiny ecdysozoan from the early Cambrian, Yanjiahe Formation, South China, with an unusual sack-like appearance, single opening, and spiny ornament. Beretella has no equivalent among animals, except Saccorhytus from the basal Cambrian. Phylogenetic analyses resolve both forms as a sister group (Saccorhytida) to all known Ecdysozoa, thus suggesting that ancestral ecdysozoans may have been non-vermiform animals. Saccorhytids are likely to represent an early dead-end off-shot along the stem-line Ecdysozoa that possibly evolved through anatomical simplification (e.g. lack of anus). Although extinct during the Cambrian, this animal lineage provides precious insight into the early evolution of Ecdysozoa and the nature (possibly non-vermiform) of the earliest representatives of the group.

paleontology↗

bioRxiv2024

A multi-scale feature fusion model based on biological knowledge graph and transformer-encoder for drug-drug interaction prediction

Drug-Drug Interaction (DDI) refers to the combined effects that occur when a patient takes multiple medications simultaneously or within the same period. This interaction can either enhance the therapeutic effects of the drugs or inhibit their efficacy, and in severe cases, it can even lead to adverse drug reactions (ADRs). Thus, it is crucial to identify potential DDIs, as this information is significant for both biological research and clinical medicine. However, most existing works only consider the information of individual drugs or focus on the local correlation between a few medical entities, thus overlooking the global performance of the entire human medical system and the potential synergistic effects of multi-scale information. Consequently, these limitations hinder the predictive ability of models. In this paper, we propose an innovative multi-scale feature fusion model called ALG-DDI, which can comprehensively incorporate attribute information, local biological information, and global semantic information. To achieve this, we first employ the Attribute Masking method to obtain the embedding vector of the molecular graph. Next, ALG-DDI leverages heterogeneous graphs to capture the local biological information between drugs and several highly related biological entities. The global semantic information is also learned from the medicine-oriented large knowledge graphs. Finally, we employ a transformer encoder to fuse the multi-scale drug representations and feed the resulting drug pair vector into a fully connected neural network for prediction. Experimental evaluations on datasets of varying sizes and different classification tasks demonstrate that ALG-DDI outperforms other state-of-the-art models.

bioinformatics↗

bioRxiv2023

Interpretable Online Network Dictionary Learning for Inferring Long-Range Chromatin Interactions

Dictionary learning (DL), implemented via matrix factorization (MF), is commonly used in computational biology to tackle ubiquitous clustering problems. The method is favored due to its conceptual simplicity and relatively low computational complexity. However, DL algorithms produce results that lack interpretability in terms of real biological data. Additionally, they are not optimized for graph-structured data and hence often fail to handle them in a scalable manner. In order to address these limitations, we propose a novel DL algorithm called online convex network dictionary learning (online cvxNDL). Unlike classical DL algorithms, online cvxNDL is implemented via MF and designed to handle extremely large datasets by virtue of its online nature. Importantly, it enables the interpretation of dictionary elements, which serve as cluster representatives, through convex combinations of real measurements. Moreover, the algorithm can be applied to data with a network structure by incorporating specialized subnetwork sampling techniques. To demonstrate the utility of our approach, we apply cvxNDL on 3D-genome RNAPII ChIA-Drop data with the goal of identifying important long-range interaction patterns (long-range dictionary elements). ChIA-Drop probes higher-order interactions, and produces data in the form of hypergraphs whose nodes represent genomic fragments. The hyperedges represent observed physical contacts. Our hypergraph model analysis has the objective of creating an interpretable dictionary of long-range interaction patterns that accurately represent global chromatin physical contact maps. Through the use of dictionary information, one can also associate the contact maps with RNA transcripts and infer cellular functions. To accomplish the task at hand, we focus on RNAPII-enriched ChIA-Drop data from Drosophila Melanogaster S2 cell lines. Our results offer two key insights. First, we demonstrate that online cvxNDL retains the accuracy of classical DL (MF) methods while simultaneously ensuring unique interpretability and scalability. Second, we identify distinct collections of proximal and distal interaction patterns involving chromatin elements shared by related processes across different chromosomes, as well as patterns unique to specific chromosomes. To associate the dictionary elements with biological properties of the corresponding chromatin regions, we employ Gene Ontology (GO) enrichment analysis and perform multiple RNA coexpression studies. Availability and ImplementationThe code and test datasets are available at:\ https://github.com/rana95vishal/chromatin_DL/ Author summaryWe introduce a novel method for dictionary learning termed online convex Network Dictionary Learning (online cvxNDL). The method operates in an online manner and utilizes representative subnetworks of a network dataset as dictionary elements. A key feature of online cvxNDL is its ability to work with graph-structured data and generate dictionary elements that represent convex combinations of real data points, thus ensuring interpretability. Online cvxNDL is used to investigate long-range chromatin interactions in S2 cell lines of Drosophila Melanogaster obtained through RNAPII ChIA-Drop measurements represented as hypergraphs. The results show that dictionary elements can accurately and efficiently reconstruct the original interactions present in the data, even when subjected to convexity constraints. To shed light on the biological relevance of the identified dictionaries, we perform Gene Ontology enrichment and RNA-seq coexpression analyses. These studies uncover multiple long-range interaction patterns that are chromosome-specific. Furthermore, the findings affirm the significance of convex dictionaries in representing TADs cross-validated by imaging methods (such as 3-color FISH (fluorescence in situ hybridization)).

genomics↗