Biology subjects

Liu, Y.

Publications and source records attributed to Liu, Y..

At least 1,531 records · Page 85Linked to original sources

bioRxiv2023

Novel multifunctional glycan probes to elucidate specificities of diverse glycan recognition systems

Glycan microarrays of sequence-defined glycans are the most widely used approach for high-throughput studies of interactions of glycans with glycan binding proteins. Currently nitrocellulose (NC) coated or N-hydroxysuccinimide (NHS) functionalized glass slides are the two most commonly used surfaces for immobilizing glycan probes as noncovalent and covalent arrays, respectively. The mode of glycan presentation can have an influence on the microarray readouts and is an important consideration in the glycan recognition knowledgebase. Here we present the development of glycan probes, tagged with a novel type of tri-functional Fmoc-amino-azido (FAA) linker, which can be readily converted into lipid-tagged or amino-terminating glycan probes for generating neoglycolipid (NGL)-based noncovalent arrays as well as covalent arrays. The azido functionality in the FAA glycan probes provides a means to light up the glycans via biorthogonal Click chemistry so that they become scanner-readable on the covalent arrays, which represents an advance in array quality control. Here analyses were carried out with a diverse set of 36 glycan binding proteins (GBPs) to compare the performance of 44 glycans presented in a liposomal formulation as noncovalent arrays on NC coated slides and on two types of NHS slides as covalent arrays. With most of anti-glycan antibodies and plant lectins investigated, there were negligible or subtle differences in the binding detected in different arrays. However, there were some striking differences observed between the covalent and noncovalent arrays. These include binding of Lotus Tetragonolobus Lectin (LTL) on the covalent but not the noncovalent arrays, and a clear preference observed for glycans on the noncovalent array in the binding of adhesins (VP1 proteins) of two human polyomaviruses JCPyV and BKPyV, and the human immune lectins Siglecs 7 and 9. Subtle yet significant differences in the binding to low affinity glycan ligands were also observed with three bacterial toxins in different arrays. These results revealed interesting insights into the binding behaviour of different GBPs on the noncovalent and covalent arrays and highlight the importance to consider different array platforms in elucidating glycan mediated interactions. The FAA glycan probes can be readily rendered fluorescent via Click chemistry in solution, enabling the detection of GBPs at the surface of cells. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=149 SRC="FIGDIR/small/548266v1_ufig1.gif" ALT="Figure 1"> View larger version (66K): org.highwire.dtl.DTLVardef@8f2a49org.highwire.dtl.DTLVardef@1212409org.highwire.dtl.DTLVardef@c4ea39org.highwire.dtl.DTLVardef@1ed6a88_HPS_FORMAT_FIGEXP M_FIG C_FIG A new trifunctional linker can present glycans on different array surfaces for analyses of soluble glycan binding proteins or as fluorescent probes for binding by proteins at the surface of cells.

biochemistry↗

bioRxiv2023

Learning Complete Protein Representation by Deep Coupling of Sequence and Structure

Learning effective representations is crucial for understanding proteins and their biological functions. Recent advancements in language models and graph neural networks have enabled protein models to leverage primary or tertiary structure information to learn representations. However, the lack of practical methods to deeply co-model the relationships between protein sequences and structures has led to suboptimal embeddings. In this work, we propose CoupleNet, a network that couples protein sequence and structure to obtain informative protein representations. CoupleNet incorporates multiple levels of features in proteins, including the residue identities and positions for sequences, as well as geometric representations for tertiary structures. We construct two types of graphs to model the extracted sequential features and structural geometries, achieving completeness on these graphs, respectively, and perform convolution on nodes and edges simultaneously to obtain superior embeddings. Experimental results on a range of tasks, such as protein fold classification and function prediction, demonstrate that our proposed model outperforms the state-of-the-art methods by large margins.

bioinformatics↗

bioRxiv2023

Enhanced cellular longevity arising from environmental fluctuations

Cellular longevity is regulated by both genetic and environmental factors. However, the interactions of these factors in the context of aging remain largely unclear. Here, we formulate a mathematical model for dynamic glucose modulation of a core gene circuit in yeast aging, which not only guided the design of pro-longevity interventions, but also revealed the theoretical principles underlying these interventions. We introduce the dynamical systems theory to capture two general means for promoting longevity - the creation of a stable fixed point in the "healthy" state of the cell and the dynamic stabilization of the system around this healthy state through environmental oscillations. Guided by the model, we investigate how both of these can be experimentally realized by dynamically modulating environmental glucose levels. The results establish a paradigm for theoretically analyzing the trajectories and perturbations of aging that can be generalized to aging processes in diverse cell types and organisms.

systems biology↗

bioRxiv2023

Nuclear Localization of Argonaute is affected by Cell Density and May Relieve Repression by microRNAs

Argonaute protein is associated with post-transcriptional control of cytoplasmic gene expression through miRNA-induced silencing complexes (miRISC). Specific cellular and environmental conditions can trigger AGO protein to accumulate in the nucleus. Localization of AGO is central to understanding miRNA action, yet the consequences of AGO being in the nucleus are undefined. We show nuclear enrichment of AGO2 in HCT116 cells grown in two-dimensional culture to high density, HCT116 cells grown in three-dimensional tumor spheroid culture, and human colon tumors. The shift in localization of AGO2 from cytoplasm to nucleus de-represses cytoplasmic AGO2-eCLIP targets that were candidates for canonical regulation by miRISC. Constitutive nuclear localization of AGO2 using an engineered nuclear localization signal increases cell migration. Critical RNAi factors also affect the localization of AGO2. Knocking out an enzyme essential for miRNA biogenesis, DROSHA, depletes mature miRNAs and restricts AGO2 localization to the cytoplasm, while knocking out the miRISC scaffolding protein, TNRC6, results in nuclear localization of AGO2. These data suggest that AGO2 localization and miRNA activity can be regulated depending on environmental conditions, expression of mature miRNAs, and expression of miRISC cofactors. Localization and expression of core miRISC protein machinery should be considered when investigating the roles of miRNAs.

biochemistry↗

bioRxiv2023

Malaria resistance-related biological adaptation and complex evolutionary footprints of Tai-Kadai people inferred from 796 genomes

Pathogen-host adaptative interaction and complex population demographical processes, including admixture, drift and Darwen selection, have considerably shaped the Neolithic-to-Modern Western Eurasian population structure and genetic susceptibility to modern human diseases. However, the genetic footprints of evolutionary events in East Asia keep unknown as the underrepresentation of genomic diversity and the design of large-scale population studies. We reported one aggregated database of genome-wide-SNP variations from 796 Tai-Kadai (TK) genomes, including Bouyei first reported here, to explore the genetic history, population structure and biological adaptative features of TK-speaking people from Southern China and Southeast Asia. We found geography-related population substructure among TK-speaking people using the state-of-the-art population genetic structure reconstruction techniques based on the allele frequency spectrum and haplotype-resolved phased fragments. We found that the Northern TK-speaking people from Guizhou harboured one TK-dominant ancestry maximised in Bouyei people, and the Southern one from Thailand obtained more influences from Southeast Asians and indigenous people. We reconstructed the fitted admixture models and demographic graphs, which showed that TK-speaking people received gene flow from ancient rice farmer-related lineages related to the Hmong-Mien and Austroasiatic people and Northern millet farmers associated with the Sino-Tibetan people. Biological adaptation focused on our identified unique TK lineages related to Bouyei showed many adaptive signatures conferring Malaria resistance and low-rate lipid metabolism. Further gene enrichment, the allele frequency distribution of derived alleles, and their correlation with the incidence of Malaria further confirmed that CR1 played an essential role in the resistance of Malaria in the ancient "Baiyue" tribes.

genetics↗

bioRxiv2023

Small extracellular vesicles promote stiffness-mediated metastasis

Tissue stiffness is a critical prognostic factor in breast cancer and is associated with metastatic progression. Here we show an alternative and complementary hypothesis of tumor progression whereby physiological matrix stiffness affects the quantity and protein cargo of small EVs produced by cancer cells, which in turn drive their metastasis. Primary patient breast tissue produces significantly more EVs from stiff tumor tissue than soft tumor adjacent tissue. EVs released by cancer cells on matrices that model human breast tumors (25 kPa; stiff EVs) feature increased adhesion molecule presentation (ITG2{beta}1, ITG6{beta}4, ITG6{beta}1, CD44) compared to EVs from softer normal tissue (0.5 kPa; soft EVs), which facilitates their binding to extracellular matrix (ECM) protein collagen IV, and a 3-fold increase in homing ability to distant organs in mice. In a zebrafish xenograft model, stiff EVs aid cancer cell dissemination through enhanced chemotaxis. Moreover, normal, resident lung fibroblasts treated with stiff and soft EVs change their gene expression profiles to adopt a cancer associated fibroblast (CAF) phenotype. These findings show that EV quantity, cargo, and function depend heavily on the mechanical properties of the extracellular microenvironment. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=200 SRC="FIGDIR/small/545937v3_ufig1.gif" ALT="Figure 1"> View larger version (42K): org.highwire.dtl.DTLVardef@1be1dbdorg.highwire.dtl.DTLVardef@928710org.highwire.dtl.DTLVardef@1e14e9borg.highwire.dtl.DTLVardef@efdb7f_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2023

Engineering transposon-associated TnpB-1 ωRNA system for efficient gene editing and disease treatment in mouse

Transposon-associated ribonucleoprotein TnpB is known to be the ancestry endonuclease of diverse Cas12 effector proteins from type-V CRISPR system. Given its small size (409 aa), it is of interest to examine whether engineered TnpB could be used for efficient mammalian genome editing. Here, we showed that the gene editing activity of native TnpB in mouse embryos was already higher than previously identified small-sized Cas12f1. Further stepwise engineering of noncoding RNA ({omega}RNA or reRNA) component of TnpB significantly elevated the nuclease activity of TnpB. Notably, an optimized TnpB-{omega}RNA system could be efficiently delivered in vivo with single adeno-associated virus (AAV) and prevented the disease phenotype in a tyrosinaemia mouse model. Thus, the engineered miniature TnpB system represents a new addition to the current genome editing toolbox, with the unique feature of the smallest effector size that facilitate efficient AAV delivery for editing of cells and tissues.

molecular biology↗

bioRxiv2023

Proteomic profiling identifies Serpin B9 as mediator of resistance to CAR T-cell and bispecific antibody treatment in B-cell lymphoma

Although T-cell-engaging therapies are highly effective in patients with relapsed and/or refractory B-cell non-Hodgkin lymphoma (B-NHL), responses are often not durable. To identify tumor-intrinsic drivers of resistance, we quantified in-vitro response to CD19-directed chimeric antigen receptor T-cells (CD19-CAR) and bispecific antibodies (BsAb) across 46 B-NHL cell lines and measured their proteomic profiles at baseline. Among the proteins associated with poor in-vitro response was Serpin B9, an endogenous granzyme B inhibitor. Knock-out of SERPINB9 in cell lines with high intrinsic expression rendered them more susceptible to CD19-CAR and CD19-BsAb. Overexpression in cell lines with low intrinsic expression attenuated responses. Polatuzumab, vorinostat, lenalidomide, or checkpoint inhibitors improved response to CD19-CAR, although independently of Serpin B9 expression. Besides providing an important resource of therapy response and proteomic profiles, this study refines our understanding of resistance in T-cell engaging therapies, and suggests clinically relevant combination regimes.

cancer biology↗

bioRxiv2023

Targeting Ovarian Cancer Stem Cells by Dual Inhibition of the Long Noncoding RNA HOTAIR and Lysine Methyltransferase EZH2

Persistence of cancer stem cells (CSC) is believed to contribute to resistance to platinum-based chemotherapy and disease relapse in ovarian cancer (OC), the fifth leading cause of cancer- related death among US women. HOXC transcript antisense RNA (HOTAIR) is a long noncoding RNA (lncRNA) overexpressed in high-grade serous OC (HGSOC) and linked to chemoresistance. However, HOTAIR impacts chromatin dynamics in OCSC and how this oncogenic lncRNA contributes to drug resistant disease are incompletely understood. Here we generated HOTAIR knock-out (KO) HGSOC cell lines using paired CRISPR guide RNA design to investigate the function of HOTAIR. We show that loss of HOTAIR function re-sensitized OC cells to platinum treatment and decreased the population of OCSC. Furthermore, HOTAIR KO inhibited the development of stemness-related phenotypes, including spheroid formation ability, as well as expression of key stemness-associated genes ALDH1A1, Notch3, Sox9, and PROM1. HOTAIR KO altered both the cellular transcriptome and chromatin accessibility landscape of multiple oncogenic-associated genes and pathways, including the NF-kB pathway. HOTAIR functions as an oncogene by recruiting enhancer of zeste 2 (EZH2) to catalyze H3K27 tri-methylation to suppress downstream tumor suppressor genes, and it was of interest to inhibit both HOTAIR and EZH2. In vivo, combining a HOTAIR inhibitor with an EZH2 inhibitor and platinum chemotherapy decreased tumor formation and increased survival. These results suggest a key role for HOTAIR in OCSC and malignant potential. Targeting HOTAIR in combination with epigenetic therapies may represent a therapeutic strategy to ameliorate OC progression and resistance to platinum-based chemotherapy.

cancer biology↗

bioRxiv2023

Interaction molecular QTL mapping discovers cellular and environmental modifiers of genetic regulatory effects

Bulk tissue molecular quantitative trait loci (QTLs) have been the starting point for interpreting disease-associated variants, while context-specific QTLs show particular relevance for disease. Here, we present the results of mapping interaction QTLs (iQTLs) for cell type, age, and other phenotypic variables in multi-omic, longitudinal data from blood of individuals of diverse ancestries. By modeling the interaction between genotype and estimated cell type proportions, we demonstrate that cell type iQTLs could be considered as proxies for cell type-specific QTL effects. The interpretation of age iQTLs, however, warrants caution as the moderation effect of age on the genotype and molecular phenotype association may be mediated by changes in cell type composition. Finally, we show that cell type iQTLs contribute to cell type-specific enrichment of diseases that, in combination with additional functional data, may guide future functional studies. Overall, this study highlights iQTLs to gain insights into the context-specificity of regulatory effects.

genetics↗

bioRxiv2023

Effect of Acute Stress on Working Memory in Pilots: Investigating the Modulatory Role of Memory Load

The effect of acute stress on working memory (WM) has not been extensively investigated in demanding tasks, although many practitioners, such as pilots, frequently face varying working memory demands. The current study investigated how acute stress modulates pilots WM performance. Forty-two pilots were randomly assigned to the stress group or the control group. The stress group experienced acute stress induced by the modified Trier Social Stress Test (TSST), whereas the control group did not receive any stress induction. Both groups performed N-back tasks under varying degrees of memory load (0-back, 1-back, and 2-back). State-Trait Anxiety Inventory scores, heart rates, and salivary cortisol concentrations were measured in the experiment to determine the extent of induced stress. The results revealed that (1) The modified TSST effectively induced acute stress in the stress group. (2) As rising of memory load, the reaction time expectedly increased while accuracy decreased. (3) An interaction between the stress and control group was found for accuracy, indicating that acute stress might contribute to higher accuracy in moderate WM load (1-back). These suggestions support the emotion regulation hypothesis, in which acute stress induces heightened emotional states that impact WM performance. Furthermore, the current study explored the interplay between acute stress and memory load in a specific population, offering insights into potential pilots stress management strategies and training programs to improve pilots WM abilities. These applications would be beneficial for strengthening pilots coping mechanisms in emergencies and guaranteeing flight safety.

physiology↗

bioRxiv2023

Transcriptomic Enrichment of Ferroptosis-Related Gene ACSL4 in Advanced Hepatic Fibrosis/Cirrhosis: Bioinformatics Analysis and Experimental validation

BackgroundLiver fibrosis is a critical part of the clinical process of liver disease that progresses to cirrhosis and even liver cancer, and effective treatment and clinical biomarkers are urgently needed to manage liver fibrosis. Ferroptosis, a notable biological phenomenon that has received attention because of the role it performs in liver fibrosis. The objective of this research is in order to identify key ferroptosis genes related to advanced liver fibrosis/cirrhosis. MethodsGene expression differences were analyzed in liver fibrosis liver tissue of hepatitis B virus(HBV)infection patients, non-alcoholic steatohepatitis (NASH) patients and alcoholic hepatitis (AH) patients to obtain overlapping ferroptosis-related genes that are significantly up-regulated. A multigroup comparison was performed to evaluate the diagnostic clinical importance of ferroptosis-related genes of patients in differential degrees of liver fibrosis, and confirmed via gene expression trend analysis. The differential expression of candidate ferroptosis-related genes through classical carbon tetrachloride (CCl4) induced advanced liver fibrosis mice model were validated by real-time quantitative PCR (qPCR). Correlation analysis was conducted to tentatively identify the connections between hepatic ferroptosis-related genes and key genes participating in functional pathways relevant to liver fibrosis. ResultsWe screened and obtained 10 genes related to ferroptosis, all of which were significantly up-regulated in liver tissue from liver fibrosis patients of different etiologies, and identified acyl-CoA synthetase long chain family member 4 (ACSL4) was transcriptomic enriched in patients with HBV infection, NASH, AH-associated advanced liver fibrosis and cirrhotic tissue adjacent to hepatocellular carcinoma (HCC). In CCl4 induced advanced liver fibrosis mice model, the hepatic ACSL4 expression was significantly up-regulated when compared to normal controls. In our study, we also suggest a significant association between ACSL4 and representative genes in liver fibrosis-related pathway. ConclusionWe found that ACSL4 mRNA can effectively differentiate the severity of liver fibrosis, suggesting its potential clinical diagnostic value in patients with liver fibrosis regardless of its etiology. ACSL4 may be a promising biomarker, which deserves further research.

bioinformatics↗

bioRxiv2023

Leucine alleviates cytokine storm syndrome by regulating macrophage polarization via the mTORC1/LXRα signaling pathway

Cytokine storms are associated with severe pathological damage and death in some diseases. Excessive activation of M1 macrophages and the subsequent secretion of pro-inflammatory cytokines are a major cause of cytokine storms. Therefore, promoting the polarization of M2 macrophages to restore immune balance is a promising therapeutic strategy for treating cytokine storm syndrome (CSS). This study was aimed at investigating the potential protective effects of leucine on lipopolysaccharide (LPS)-induced CSS in mice and exploring the underlying mechanisms. CSS was induced by LPS administration in mice, which were concurrently administered leucine orally. In vitro, BMDMs were polarized to M1 and M2 phenotypes with LPS and IL-4, respectively, and treated with leucine. Leucine decreased mortality in mice treated with lethal doses of LPS. Specifically, leucine decreased M1 polarization and promoted M2 polarization, thus diminishing pro-inflammatory cytokine levels and ameliorating CSS in mice. Further studies revealed that leucine induced macrophage polarization through the mechanistic target of rapamycin complex 1 (mTORC1)/liver X receptor (LXR) pathway, which synergistically enhanced the expression of the IL-4-induced M2 marker Arg1 and subsequent M2 polarization. In summary, this study revealed that leucine ameliorates CSS in LPS mice by promoting M2 polarization through the mTORC1/LXR/Arg1 signaling pathway. Our findings indicate that a fundamental link between metabolism and immunity contributes to the resolution of inflammation and the repair of damaged tissues.

immunology↗

bioRxiv2023

Reduced Reverse Replay in Anxious Individuals Impairs Reward Prediction

Understanding how we learn about the value and structure of our environment is central to neurocognitive theories of many psychiatric and neurological disorders. Learning processes have been extensively studied during performance of behavioural tasks (online learning) but less so in relation to "resting" (offline) states. A candidate mechanism for such offline learning is "replay," the sequential neural reactivation of past experiences. Notably, value-based learning is especially tied to replay unfolding in reverse order relative to the original experience ("backward replay"). Here, we demonstrate the utility of EEG-based neural decoding for investigating offline learning, and relate it to trait anxiety. Participants were first required to infer sequential relationships among task objects by using a learned rule to reorganise their visual experiences into distinct sequences. Afterwards, they observed that the final object in one of the sequences was associated with a monetary reward and then entered a post-value resting state. During this rest, we found evidence of backward replay for reward-linked object sequences. The strength of such replay was negatively associated with trait anxiety and positively predicted an increased behavioural preference for reward-predictive stimuli. We also found that high-trait-anxiety individuals showed inefficient credit assignment irrespective of reward magnitude, indicating that this effect does not merely reflect reduced reward sensitivity. Together, these findings suggest a potential aberrant replay mechanism during offline learning in individuals with high trait anxiety. More broadly, our approach illustrates the potential of EEG for measuring structured neural representations in vivo, with implications for studying cognition across a range of neuropsychiatric and neurological disorders.

neuroscience↗

bioRxiv2023

A mean-field to capture asynchronous irregular dynamics of conductance-based networks of adaptive quadratic integrate-and-fire neuron models

Mean-field models are a class of models used in computational neuroscience to study the behaviour of large populations of neurons. These models are based on the idea of representing the activity of a large number of neurons as the average behaviour of "mean field" variables. This abstraction allows the study of large-scale neural dynamics in a computationally efficient and mathematically tractable manner. One of these methods, based on a semi-analytical approach, has previously been applied to different types of single-neuron models, but never to models based on a quadratic form. In this work, we adapted this method to quadratic integrate-and-fire neuron models with adaptation and conductance-based synaptic interactions. We validated the mean-field model by comparing it to the spiking network model. This mean-field model should be useful to model large-scale activity based on quadratic neurons interacting with conductance-based synapses.

bioinformatics↗

bioRxiv2023

Genome-wide identification of silencers in human cells

Transcriptional regulation is a complex process that is controlled by a variety of factors, including enhancers and silencers. Silencers, also known as repressor elements, play a crucial role in the fine-tuning of gene expression by inhibiting or suppressing transcription in the human genome. Although significant progresses have been made, genome-wide silencer research is still in its early stages. Here, we used a genome-wide method called massively parallel reporter assays (MPRAs) to identify silencers in three human cell lines: K562, LNCap, and HEK293T. We identified 739,434, 643,484, and 491,952 silencer regions in these cell lines, respectively. We found that most of the silencers we identified had inhibitory activity and significantly enriched inhibitory motifs. These results confirm that silencers are ubiquitous in the human genome and play an important role in regulating gene expression. Therefore, our study provides a general strategy for genome-wide functional identification of silencer elements. This information could be used to better understand the mechanisms of gene regulation and to develop new therapeutic strategies for diseases that are caused by dysregulation of gene expression.

genomics↗

bioRxiv2023

Spatiotemporal Control of Genomics and Epigenomics by Ultrasound

CRISPR (clustered regularly interspaced short palindromic repeats) is a revolutionary technology for genome editing. Its derived technologies such as CRISPR activation (CRISPRa) and CRISPR interference (CRISPRi) further allow transcriptional and epigenetic modulations. Focused ultrasound (FUS) can penetrate deep in biological tissues and induce mild hyperthermia in a confined region to activate heat-sensitive genes. Here we engineer a set of CRISPR(a/i) tools containing heat-sensitive genetic modules controllable by FUS for the regulation of genome and epigenome in live cells and animals. We demonstrated the capabilities of FUS-inducible CRISPRa, CRISPRi, and CRISPR (FUS-CRISPR(a/i)) to upregulate, repress, and knockout exogenous and/or endogenous genes, respectively, in different cell types. We further targeted FUS-CRISPR to telomeres in tumor cells to induce telomere disruption, inhibiting tumor growth and enhancing tumor susceptibility to killing by chimeric antigen receptor (CAR)-T cells. FUS-CRISPR-mediated telomere disruption for tumor priming combined with CAR-T therapy demonstrated synergistic therapeutic effects in xenograft mouse models. The FUS-CRISPR(a/i) toolbox allows the remote, noninvasive, and spatiotemporal control of genomic and epigenomic reprogramming in vivo, with extended applications in cancer treatment.

bioengineering↗

bioRxiv2023

CellStitch: 3D Cellular Anisotropic Image Segmentation via Optimal Transport

BackgroundSpatial mapping of transcriptional states provides valuable biological insights into cellular functions and interactions in the context of the tissue. Accurate 3D cell segmentation is a critical step in the analysis of this data towards understanding diseases and normal development in situ. Current approaches designed to automate 3D segmentation include stitching masks along one dimension, training a 3D neural network architecture from scratch, and reconstructing a 3D volume from 2D segmentations on all dimensions. However, the applicability of existing methods is hampered by inaccurate segmentations along the non-stitching dimensions, the lack of high-quality diverse 3D training data, and inhomogeneity among different dimensions; as a result, they have not been widely used in practice. MethodsTo address these challenges, we formulate the problem of finding cell correspondence across layers with a novel optimal transport (OT) approach. We propose CellStitch, a flexible pipeline that segments cells from 3D images without requiring large amounts of 3D training data. We further extend our method to interpolate internal slices from highly anisotropic cell images to recover isotropic cell morphology. ResultsWe evaluated the performance of CellStitch through eight 3D plant microscopic datasets with diverse anisotropic levels and cell shapes. CellStitch substantially outperforms the state-of-the art methods on anisotropic images, and achieves comparable segmentation quality against competing methods in isotropic setting. We benchmarked and reported 3D segmentation results of all the methods with instance-level precision, recall and average precision (AP) metrics. ConclusionThe proposed OT-based 3D segmentation pipeline outperformed the existing state-of-the-art methods on different datasets with nonzero anisotropy, providing high fidelity recovery of 3D cell morphology from microscopic images.

bioinformatics↗