Thousands Of Cells Research Articles

An Omni-Mesoscope for multiscale high-throughput quantitative phase imaging of cellular dynamics and high-content molecular characterization.

The mesoscope has emerged as a powerful imaging tool in biomedical research, yet its high cost and low resolution have limited its broader application. Here, we introduce the Omni-Mesoscope, a high-spatial-temporal and multimodal mesoscopic imaging platform built from cost-efficient off-the-shelf components. This system uniquely merges the capabilities of label-free quantitative phase microscopy to capture live-cell morphodynamics across thousands of cells with highly multiplexed fluorescence imaging for comprehensive molecular characterization. This Omni-Mesoscope offers a mesoscale field of view of ~5 square millimeters with a high spatial resolution down to 700 nanometers, enabling the capture of detailed subcellular features. We demonstrate its capability in delineating molecular characteristics underlying rare morphodynamic cellular phenomena, including cancer cell responses to chemotherapy and the emergence of polyploidy in drug-resistant cells. We also integrate expansion technique to enhance three-dimensional volumetric super-resolution imaging of thicker tissues, opening the avenues for biological exploration at unprecedented scales and resolutions.

Attention-Guided Residual U-Net with SE Connection and ASPP for Watershed-Based Cell Segmentation in Microscopy Images.

Time-lapse microscopy imaging is a crucial technique in biomedical studies for observing cellular behavior over time, providing essential data on cell numbers, sizes, shapes, and interactions. Manual analysis of hundreds or thousands of cells is impractical, necessitating the development of automated cell segmentation approaches. Traditional image processing methods have made significant progress in this area, but the advent of deep learning methods, particularly those using U-Net-based networks, has further enhanced performance in medical and microscopy image segmentation. However, challenges remain, particularly in accurately segmenting touching cells in images with low signal-to-noise ratios. Existing methods often struggle with effectively integrating features across different levels of abstraction. This can lead to model confusion, particularly when important contextual information is lost or the features are not adequately distinguished. The challenge lies in appropriately combining these features to preserve critical details while ensuring robust and accurate segmentation. To address these issues, we propose a novel framework called RA-SE-ASPP-Net, which incorporates Residual Blocks, Attention Mechanism, Squeeze-and-Excitation connection, and Atrous Spatial Pyramid Pooling to achieve precise and robust cell segmentation. We evaluate our proposed architecture using an induced pluripotent stem cell reprogramming dataset, a challenging dataset that has received limited attention in this field. Additionally, we compare our model with different ablation experiments to demonstrate its robustness. The proposed architecture outperforms the baseline models in all evaluated metrics, providing the most accurate semantic segmentation results. Finally, we applied the watershed method to the semantic segmentation results to obtain precise segmentations with specific information for each cell.

Single-Cell Recording: An Inside Look at a Brain Cell’s Work

The brain is made up of thousands of cells that make sense of what a person experiences. These cells, which are called neurons, all have different jobs depending on where they are in the brain. Scientists use multiple methods to collect information about how and when neurons react; this is sort of like taking a picture of the brain’s activity! To understand the jobs neurons do, researchers generally use methods that do not go directly into the brain. These approaches are commonly used because they are convenient; however, because they are “pictures” taken from outside the skull, the information they provide scientists may be incomplete or inaccurate. Fortunately, it is also possible for researchers to use a method that goes directly inside the brain tissue, called single-cell recording, to more accurately capture what a neuron is doing.

Unravelling a diversity of cellular structures and aggregation dynamics during the early development of Myxococcus xanthus.

Aggregation underlies the collective dynamics of a diversity of organisms, enabling the formation of complex structures and emergent behaviours on interaction with the environment. Cellular aggregation constitutes one of the routes to collective motility and multicellular development. Myxococcus xanthus, a social bacterium, is a valuable model for studying the aggregative path to multicellularity, a major transition in the evolutionary history of life. While the collective developmental behaviour of M. xanthus has been largely studied in high cellular densities, there is a lack of understanding at low-density conditions that can be ecologically relevant. In this work, we study the early stages of emergent collective behaviour of M. xanthus under nutrient-poor and low-density conditions, uncovering the formation of diverse cellular structures with different shapes and sizes, ranging from individual cells to networks comprising thousands of cells. We study their motility patterns and their prevalence along development and discuss their cross-scale role on the population's exploratory dynamics. This work contributes to understanding key, yet largely understudied, aspects in the early stages of multicellular development in myxobacteria, shedding light on the dynamics underlying aggregative processes in this and other taxa and study systems.

TIRTL-seq: Deep, quantitative, and affordable paired TCR repertoire sequencing.

α/β T cells are key players in adaptive immunity. The specificity of T cells is determined by the sequences of the hypervariable T cell receptor (TCR) α and β chains. Although bulk TCR sequencing offers a cost-effective approach for in-depth TCR repertoire profiling, it does not provide chain pairings, which are essential for determining T cell specificity. In contrast, single-cell TCR sequencing technologies produce paired chain data, but are limited in throughput to thousands of cells and are cost-prohibitive for cohort-scale studies. Here, we present TIRTL-seq (Throughput-Intensive Rapid TCR Library sequencing), a novel approach that generates ready-to-sequence TCR libraries from live cells in less than 7 hours. The protocol is optimized for use with non-contact liquid handlers in an automation-friendly 384-well plate format. Reaction volume miniaturization reduces library preparation costs to <$0.50 per well. The core principle of TIRTL-seq is the parallel generation of hundreds of libraries providing multiple biological replicates from a single sample that allows precise inference of both frequencies of individual clones and TCR chain pairings from well-occurrence patterns. We demonstrate scalability of our approach up to 1 million unique paired αβTCR clonotypes corresponding to over 30 million T cells per sample at a cost of less than $2000. For a sample of 10 million cells the cost is ~$200. We benchmarked TIRTL-seq against state-of-the-art 5'RACE bulk TCR-seq and 10x Genomics Chromium technologies on longitudinal samples. We show that TIRTL-seq is able to quantitatively identify expanding and contracting clonotypes between timepoints while providing accurate TCR chain pairings, including distinct temporal dynamics of SARS-CoV-2-specific and EBV-specific CD8+ T cell responses after infection. While clonal expansion was followed by sharp contraction for SARS-CoV-2 specific TCRs, EBV-specific TCRs remained stable once established. The sequences of both α and β TCR chains are essential for determining T cell specificity. As the field moves towards greater applications in diagnostics and immunotherapy that rely on TCR specificity, we anticipate that our scalable paired TCR sequencing methodology will be instrumental for collecting large paired-chain datasets and ultimately extracting therapeutically relevant information from the TCR repertoire.

Unlocking the Complexity: Exploration of Acute Lymphoblastic Leukemia at the Single Cell Level.

Acute lymphoblastic leukemia (ALL) is the most common cancer in children. ALL originates from precursor lymphocytes that acquire multiple genomic changes over time, including chromosomal rearrangements and point mutations. While a large variety of genomic defects was identified and characterized in ALL over the past 30 years, it was only in recent years that the clonal heterogeneity was recognized. Thanks to the latest advancements in single-cell sequencing techniques, which have evolved from the analysis of a few hundred cells to the analysis of thousands of cells simultaneously, the study of tumor heterogeneity now becomes possible. Different modalities can be explored at the single-cell level: DNA, RNA, epigenetic modifications, and intracellular and cell surface proteins. In this review, we describe these techniques and highlight their advantages and limitations in the study of ALL biology. Moreover, multiomics technologies and the incorporation of the spatial dimension can provide insight into intercellular communication. We describe how the different single-cell sequencing technologies help to unravel the molecular complexity of ALL, shedding light on its development, its heterogeneity, its interaction with the leukemia microenvironment and possible relapse mechanisms.

Segmentation and characterization of macerated fibers and vessels using deep learning

PurposeWood comprises different cell types, such as fibers, tracheids and vessels, defining its properties. Studying cells’ shape, size, and arrangement in microscopy images is crucial for understanding wood characteristics. Typically, this involves macerating (soaking) samples in a solution to separate cells, then spreading them on slides for imaging with a microscope that covers a wide area, capturing thousands of cells. However, these cells often cluster and overlap in images, making the segmentation difficult and time-consuming using standard image-processing methods.ResultsIn this work, we developed an automatic deep learning segmentation approach that utilizes the one-stage YOLOv8 model for fast and accurate segmentation and characterization of macerated fiber and vessel form aspen trees in microscopy images. The model can analyze 32,640 x 25,920 pixels images and demonstrate effective cell detection and segmentation, achieving a mAP0.5-0.95\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\hbox {mAP}_{0.5-0.95}$$\\end{document} of 78 %. To assess the model’s robustness, we examined fibers from a genetically modified tree line known for longer fibers. The outcomes were comparable to previous manual measurements. Additionally, we created a user-friendly web application for image analysis and provided the code for use on Google Colab.ConclusionBy leveraging YOLOv8’s advances, this work provides a deep learning solution to enable efficient quantification and analysis of wood cells suitable for practical applications.

Comparison of Multiple Machine Learning Models for the Classification of Cell States Based on Impedance Features

Abstract Microfluidic impedance flow cytometry (IFC) enables high-throughput single-cell analysis in label-free manner. Tens of thousands of cells can be measured in several minutes under multiple frequencies, which give rise to impedance features with rich information ideal for machine learning (ML)-based cell classification. Conventional data processing approach for IFC typically exploits the scattered distribution of the measured cells which correlates the impedance features (e.g., the impedance amplitude and phase at different frequencies, the amplitude ratio between high to low frequencies) and exhibits resolved cell clusters in scatter plot. By manually gating on the distributed dots plot, the cell subgroups get mapped to different cell type or cellular states. ML-based data processing for IFC not only reduces the human workload, and more importantly, it also eliminates the human interference to manual gating strategy, and thus potentially leading to more concise and accurate cell classification results. Here, we demonstrate the ML-based classification of different cell states for tumor cells subject to anticancer drug treatment. IFC-measured impedance data of H1650 cells and Hela cells under drug-induced mitosis block state and apoptosis state have been applied for ML-based cell state identification. Three machine learning models, including the random forest (RF), support vector machine (SVM) and K-nearest neighbours (KNN) have been trained for impedance features extracted from cell signals at both 500 kHz and 10 MHz. In comparison, the RF model give rise to the highest classification accuracies among all trained models here. For H1650 cells, 84.01% and 85.96% accuracies have been respectively achieved for G1/S state vs. apoptosis and G2/M vs. apoptosis. For the classification between G2/M vs. apoptosis for the paclitaxel-treated Hela cells, the RF model produces high accuracy of 98.70%.

An Omni-Mesoscope for multiscale high-throughput quantitative phase imaging of cellular dynamics and high-content molecular characterization.

The mesoscope has emerged as a powerful imaging tool in biomedical research, yet its high cost and low resolution have limited its broader application. Here, we introduce the Omni-Mesoscope, a cost-effective high-spatial-temporal, multimodal, and multiplex mesoscopic imaging platform built from cost-efficient off-the-shelf components. This system uniquely merges the capabilities of quantitative phase microscopy to capture live-cell dynamics over a large cell population with highly multiplexed fluorescence imaging for comprehensive molecular characterization. This integration facilitates simultaneous tracking of live-cell morphodynamics across thousands of cells, alongside high-content molecular analysis at the single-cell level. Furthermore, the Omni-Mesoscope offers a mesoscale field of view of approximately 5 mm 2 with a high spatial resolution down to 700 nm, enabling the capture of information-rich images with detailed sub-cellular features. We demonstrate such capability in delineating molecular characteristics underlying rare dynamic cellular phenomena, such as cancer cell responses to chemotherapy and the emergence of polyploidy in drug-resistant cells. Moreover, the cost-effectiveness and the simplicity of our Omni-Mesoscope democratizes mesoscopic imaging, making it accessible across diverse biomedical research fields. To further demonstrate its versatility, we integrate expansion microscopy to enhance 3D volumetric super-resolution imaging of thicker tissues, opening new avenues for biological exploration at unprecedented scales and resolutions.

Control of Modular Tissue Flows Shaping the Embryo in Avian Gastrulation.

Avian gastrulation requires coordinated flows of thousands of cells to form the body plan. We quantified these flows using their fundamental kinematic units: one attractor and two repellers constituting its Dynamic Morphoskeleton (DM). We have also elucidated the mechanistic origin of the attractor, marking the primitive streak (PS), and controlled its shape, inducing gastrulation flows in the chick embryo that are typical of other vertebrates. However, the origins of repellers and dynamic embryo shape remain unclear. Here, we address these questions using active matter physics and experiments. Repeller 1, separating the embryo proper (EP) from extraembryonic (EE) tissues, arises from the tug-of-war between EE epiboly and EP isotropic myosin-induced active stress. Repeller 2, bisecting the anterior and posterior PS and associated with embryo shape change, arises from anisotropic myosin-induced active intercalation in the mesendoderm. Combining mechanical confinement with inhibition of mesendoderm induction, we eliminated either one or both repellers, as predicted by our model. Our results reveal a remarkable modularity of avian gastrulation flows delineated by the DM, uncovering the mechanistic roles of EE epiboly, EP active constriction, mesendoderm intercalation and ingression. These findings offer a new perspective for deconstructing morphogenetic flows, uncovering their modular origin, and aiding synthetic morphogenesis.

Influenza A viral burst size from thousands of infected single cells using droplet quantitative PCR (dqPCR).

An important aspect of how viruses spread and infect is the viral burst size, or the number of new viruses produced by each infected cell. Surprisingly, this value remains poorly characterized for influenza A virus (IAV), commonly known as the flu. In this study, we screened tens of thousands of cells using a microfluidic method called droplet quantitative PCR (dqPCR). The high-throughput capability of dqPCR enabled the measurement of a large population of infected cells producing progeny virus. By measuring the fully assembled and successfully released viruses from these infected cells, we discover that the viral burst sizes for both the seasonal H3N2 and the 2009 pandemic H1N1 strains vary significantly, with H3N2 ranging from 101 to 104 viruses per cell, and H1N1 ranging from 101 to 103 viruses per cell. Some infected cells produce average numbers of new viruses, while others generate extensive number of viruses. In fact, we find that only 10% of the single-cell infections are responsible for creating a significant portion of all the viruses. This small fraction produced approximately 60% of new viruses for H3N2 and 40% for H1N1. On average, each infected cell of the H3N2 flu strain produced 709 new viruses, whereas for H1N1, each infected cell produced 358 viruses. This novel method reveals insights into the flu virus and can lead to improved strategies for managing and preventing the spread of viruses.

Liam tackles complex multimodal single-cell data integration challenges.

Multi-omics characterization of single cells holds outstanding potential for profiling the dynamics and relations of gene regulatory states of thousands of cells. How to integrate multimodal data is an open problem, especially when aiming to combine data from multiple sources or conditions containing both biological and technical variation. We introduce liam, a flexible model for the simultaneous horizontal and vertical integration of paired single-cell multimodal data and mosaic integration of paired with unimodal data. Liam learns a joint low-dimensional representation of the measured modalities, which proves beneficial when the information content or quality of the modalities differ. Its integration accounts for complex batch effects using a tunable combination of conditional and adversarial training, which can be optimized using replicate information while retaining selected biological variation. We demonstrate liam's superior performance on multiple paired multimodal data types, including Multiome and CITE-seq data, and in mosaic integration scenarios. Our detailed benchmarking experiments illustrate the complexities and challenges remaining for integration and the meaningful assessment of its success.

Validation and Implementation in the Cloud of Cell-Level Models of a Digital Twin Simulation Platform

The monitoring and modelling of the Li-Ion Batteries behaviour is still a major technical challenge due to the non-linearities and coupled phenomena that determine their operation. These Li-Ion Batteries can consist of thousands of cells with series/parallel connections, which suffer from operating and degradation deviations. The pursuit of new, increasingly intelligent and heavier state estimation algorithms requires a significant amount of data and computational power, which can be challenging to deploy in current Battery Management System solutions. To solve this problem, this paper proposes a Digital Twin Simulation Platform that considers all the individual cells based on the Cloud to extend the computational power and data storage capacity. This work presents validated cell models, a module-level modelling approach, and an experimental validation platform is suggested. In addition, the first results obtained when implementing the Digital Twin Simulation Platform in the Cloud are presented.

In Vivo Microrheology Reveals Local Elastic and Plastic Responses Inside 3D Bacterial Biofilms.

Bacterial biofilms are highly abundant 3D living materials capable of performing complex biomechanical and biochemical functions, including programmable growth, self-repair, filtration, and bioproduction. Methods to measure internal mechanical properties of biofilms in vivo with spatial resolution on the cellular scale have been lacking. Here, thousands of cells are tracked inside living 3D biofilms of the bacterium Vibrio cholerae during and after the application of shear stress, for a wide range of stress amplitudes, periods, and biofilm sizes, which revealed anisotropic elastic and plastic responses of both cell displacements and cell reorientations. Using cellular tracking to infer parameters of a general mechanical model, spatially-resolved measurements of the elastic modulus inside the biofilm are obtained, which correlate with the spatial distribution of the polysaccharides within the biofilm matrix. The noninvasive microrheology and force-inference approach introduced here provides a general framework for studying mechanical properties with high spatial resolution in living materials.

Mapping genotypes to chromatin accessibility profiles in single cells.

In somatic tissue differentiation, chromatin accessibility changes govern priming and precursor commitment towards cellular fates1-3. Therefore, somatic mutations are likely to alter chromatin accessibility patterns, as they disrupt differentiation topologies leading to abnormal clonal outgrowth. However, defining the impact of somatic mutations on the epigenome in human samples is challenging due to admixed mutated and wild-type cells. Here, to chart how somatic mutations disrupt epigenetic landscapes in human clonal outgrowths, we developed genotyping of targeted loci with single-cell chromatin accessibility (GoT-ChA). This high-throughput platform links genotypes to chromatin accessibility at single-cell resolution across thousands of cells within a single assay. We applied GoT-ChA to CD34+ cells from patients with myeloproliferative neoplasms with JAK2V617F-mutated haematopoiesis. Differential accessibility analysis between wild-type and JAK2V617F-mutant progenitors revealed both cell-intrinsic and cell-state-specific shifts within mutant haematopoietic precursors, including cell-intrinsic pro-inflammatory signatures in haematopoietic stem cells, and a distinct profibrotic inflammatory chromatin landscape in megakaryocytic progenitors. Integration of mitochondrial genome profiling and cell-surface protein expression measurementallowed expansion of genotyping onto DOGMA-seq through imputation, enabling single-cell capture of genotypes, chromatin accessibility, RNA expression and cell-surface protein expression. Collectively, we show that the JAK2V617F mutation leads to epigenetic rewiring in a cell-intrinsic and cell type-specific manner, influencing inflammation states and differentiation trajectories. We envision that GoT-ChA will empower broad future investigations of the critical link between somatic mutations and epigenetic alterations across clonal populations in malignant and non-malignant contexts.

Continuous distribution of cancer cells in the cell cycle unveiled by AI-segmented imaging of 37,000 HeLa FUCCI cells

Classification of live or fixed cells based on their unlabeled microscopic images would be a powerful tool for cell biology and pathology. For such software, the first step is the generation of a ground truth database that can be used for training and testing AI classification algorithms. The Application of cells expressing fluorescent reporter proteins allows the building of ground truth datasets in a straightforward way. In this study, we present an automated imaging pipeline utilizing the Cellpose algorithm for the precise cell segmentation and measurement of fluorescent cellular intensities across multiple channels. We analyzed the cell cycle of HeLa–FUCCI cells expressing fluorescent red and green reporter proteins at various levels depending on the cell cycle state. To build the dataset, 37,000 fixed cells were automatically scanned using a standard motorized microscope, capturing phase contrast and fluorescent red/green images. The fluorescent pixel intensity of each cell was integrated to calculate the total fluorescence of cells based on cell segmentation in the phase contrast channel. It resulted in a precise intensity value for each cell in both channels. Furthermore, we conducted a comparative analysis of Cellpose 1.0 and Cellpose 2.0 in cell segmentation performance. Cellpose 2.0 demonstrated notable improvements, achieving a significantly reduced false positive rate of 2.7 % and 1.4 % false negative. The cellular fluorescence was visualized in a 2D plot (map) based on the red and green intensities of the FUCCI construct revealing the continuous distribution of cells in the cell cycle. This 2D map enables the selection and potential isolation of single cells in a specific phase. In the corresponding heatmap, two clusters appeared representing cells in the red and green states. Our pipeline allows the high-throughput and accurate measurement of cellular fluorescence providing extensive statistical information on thousands of cells with potential applications in developmental and cancer biology. Furthermore, our method can be used to build ground truth datasets automatically for training and testing AI cell classification. Our automated pipeline can be used to analyze thousands of cells within 2 h after putting the sample onto the microscope.

Decoupling cell size homeostasis in diatoms from the geometrical constraints of the silica cell wall.

Unicellular organisms are known to exert tight control over their cell size. In the case of diatoms, abundant eukaryotic microalgae, two opposing notions are widely accepted. On the one hand, the rigid silica cell wall that forms inside the parental cell is thought to enforce geometrical reduction of the cell size. On the other hand, numerous exceptions cast doubt on the generality of this model. Here, we monitored clonal cultures of the diatom Stephanopyxis turris for up to 2 yr, recording the sizes of thousands of cells, in order to follow the distribution of cell sizes in the population. Our results show that S. turris cultures above a certain size threshold undergo a gradual size reduction, in accordance with the postulated geometrical driving force. However, once the cell size reaches a lower threshold, it fluctuates around a constant size using the inherent elasticity of cell wall elements. These results reconcile the disparate observations on cell size regulation in diatoms by showing two distinct behaviors, reduction and homeostasis. The geometrical size reduction is the dominant driving force for large cells, but smaller cells have the flexibility to re-adjust the size of their new cell walls.

Abstract LB333: Improved spatially resolved single-cell transcriptomic imaging in archival tissues with MERSCOPE

Abstract The advent of spatial transcriptomics has enabled a revolution in how complex tissues are studied. However, samples with lower quality RNA due to degradation, protein crosslinking, or high RNase content remain challenging for spatial transcriptomic measurement. In particular, formalin fixed, paraffin embedded (FFPE) tissues are the most widely used sample types in clinical and molecular diagnosis, yet they are notoriously difficult for single-cell transcriptomic analysis. To accurately profile the gene expression in FFPE samples in situ, a spatial transcriptomics technique with high detection efficiency and single molecule resolution is required. The Vizgen® MERSCOPE® Platform for spatial genomics is built on Multiplexed Error Robust in situ Hybridization (MERFISH) technology and directly profiles the transcriptome of intact tissues with high sensitivity in high-quality samples. Here we present an updated workflow to perform MERFISH in low and high-quality samples. We demonstrated its application in more than 5 FFPE sample types from mouse and human, including archival samples. For each tissue type, hundreds of thousands of cells were captured using the updated MERSCOPE Platform workflow, generating 100s million counts and their spatial information for profiled genes in each sample. The updated workflow involves streamlined sample preparation and chemistry optimization to improve sensitivity. MERSCOPE accurately profiled gene expression in situ and mapped cell types in archival human samples across a range of low and high RNA qualities. We compared the performance of MERSCOPE imaging using the updated protocol to the previous version and observed a significant increase in gene counts per 100 micron2 of tissue. We also demonstrated increased reproducibility between replicates with the streamlined workflow and chemistry. Furthermore, we demonstrated the updated workflow is compatible with simultaneous protein-based cell boundary staining. Finally, we constructed a spatially resolved single-cell atlas across low-quality archival breast and lung tumor types, mapped and cataloged different cell types within the tumor microenvironment, and systematically characterized the gene expression among cells. Spatially resolved transcriptomic profiling of low-quality samples at single-cell level provides enormous opportunities in cancer research. These improvements will enable new genomic inquiries into previously intractable tissues like FFPE, leading to new biological insights into cancer progression. Citation Format: Jiang He, Bin Wang, Justin He, Renchao Chen, Benjamin Patterson, Sudhir Tattikota, Timothy Wiggin, Lizi Maziashvili, Peter Reinhold, Manisha Ray, George Emanuel. Improved spatially resolved single-cell transcriptomic imaging in archival tissues with MERSCOPE [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB333.

Identification of Interpretable Clusters and Associated Signatures in Breast Cancer Single-Cell Data: A Topic Modeling Approach.

Topic modeling is a popular technique in machine learning and natural language processing, where a corpus of text documents is classified into themes or topics using word frequency analysis. This approach has proven successful in various biological data analysis applications, such as predicting cancer subtypes with high accuracy and identifying genes, enhancers, and stable cell types simultaneously from sparse single-cell epigenomics data. The advantage of using a topic model is that it not only serves as a clustering algorithm, but it can also explain clustering results by providing word probability distributions over topics. Our study proposes a novel topic modeling approach for clustering single cells and detecting topics (gene signatures) in single-cell datasets that measure multiple omics simultaneously. We applied this approach to examine the transcriptional heterogeneity of luminal and triple-negative breast cancer cells using patient-derived xenograft models with acquired resistance to chemotherapy and targeted therapy. Through this approach, we identified protein-coding genes and long non-coding RNAs (lncRNAs) that group thousands of cells into biologically similar clusters, accurately distinguishing drug-sensitive and -resistant breast cancer types. In comparison to standard state-of-the-art clustering analyses, our approach offers an optimal partitioning of genes into topics and cells into clusters simultaneously, producing easily interpretable clustering outcomes. Additionally, we demonstrate that an integrative clustering approach, which combines the information from mRNAs and lncRNAs treated as disjoint omics layers, enhances the accuracy of cell classification.

EpiCarousel: memory- and time-efficient identification of metacells for atlas-level single-cell chromatin accessibility data.

Recent technical advancements in single-cell chromatin accessibility sequencing (scCAS) have brought new insights to the characterization of epigenetic heterogeneity. As single-cell genomics experiments scale up to hundreds of thousands of cells, the demand for computational resources for downstream analysis grows intractably large and exceeds the capabilities of most researchers. Here, we propose EpiCarousel, a tailored Python package based on lazy loading, parallel processing, and community detection for memory- and time-efficient identification of metacells, i.e. the emergence of homogenous cells, in large-scale scCAS data. Through comprehensive experiments on five datasets of various protocols, sample sizes, dimensions, number of cell types, and degrees of cell-type imbalance, EpiCarousel outperformed baseline methods in systematic evaluation of memory usage, computational time, and multiple downstream analyses including cell type identification. Moreover, EpiCarousel executes preprocessing and downstream cell clustering on the atlas-level dataset with 707043 cells and 1154611 peaks within 2 h consuming <75 GB of RAM and provides superior performance for characterizing cell heterogeneity than state-of-the-art methods. The EpiCarousel software is well-documented and freely available at https://github.com/biox-nku/epicarousel. It can be seamlessly interoperated with extensive scCAS analysis toolkits.