Biology subjects

Guo, J.

Publications and source records attributed to Guo, J..

At least 145 records · Page 8Linked to original sources

bioRxiv2024

CellBinDB: A Large-Scale Multimodal Annotated Dataset for Cell Segmentation with Benchmarking of Universal Models

In recent years, cell segmentation techniques have played a critical role in the analysis of biological images, especially for quantitative studies. Deep learning-based cell segmentation models have demonstrated remarkable performance in segmenting cell and nucleus boundaries, however, they are typically tailored to specific modalities or require manual tuning of hyperparameters, limiting their generalizability to unseen data. Comprehensive datasets that support both the training of universal models and the evaluation of various segmentation techniques are essential for overcoming these limitations and promoting the development of more versatile cell segmentation solutions. Here, we present CellBinDB, a large-scale multimodal annotated dataset established for these purposes. CellBinDB contains more than 1,000 annotated images, each labeled to identify the boundaries of cells or nuclei, including 4,6-Diamidino-2-Phenylindole (DAPI), Single-stranded DNA (ssDNA), Hematoxylin and Eosin (H&E), and Multiplex Immunofluorescence (mIF) staining, covering over 30 normal and diseased tissue types from human and mouse samples. Based on CellBinDB, we benchmarked seven state-of-the-art and widely used cell segmentation technologies/methods, and further analyzed the effects of four cell morphology indicators and image gradient on the segmentation results.

cell biology↗

bioRxiv2024

Prospective evaluation of structure-based simulations reveal their ability to predict the impact of kinase mutations on inhibitor binding

Small molecule kinase inhibitors are critical in the modern treatment of cancers, evidenced by the existence of over 80 FDA-approved small-molecule kinase inhibitors. Unfortunately, intrinsic or acquired resistance, often causing therapy discontinuation, is frequently caused by mutations in the kinase therapeutic target. The advent of clinical tumor sequencing has opened additional opportunities for precision oncology to improve patient outcomes by pairing optimal therapies with tumor mutation profiles. However, modern precision oncology efforts are hindered by lack of sufficient biochemical or clinical evidence to classify each mutation as resistant or sensitive to existing inhibitors. Structure-based methods show promising accuracy in retrospective benchmarks at predicting whether a kinase mutation will perturb inhibitor binding, but comparisons are made by pooling disparate experimental measurements across different conditions. We present the first prospective benchmark of structure-based approaches on a blinded dataset of in-cell kinase inhibitor affinities to Abl kinase mutants using a NanoBRET reporter assay. We compare NanoBRET results to structure-based methods and their ability to estimate the impact of mutations on inhibitor binding (measured as {Delta}{Delta}G). Comparing physics-based simulations, Rosetta, and previous machine learning models, we find that structure-based methods accurately classify kinase mutations as inhibitor-resistant or inhibitor-sensitizing, and each approach has a similar degree of accuracy. We show that physics-based simulations are best suited to estimate {Delta}{Delta}G of mutations that are distal to the kinase active site. To probe modes of failure, we retrospectively investigate two clinically significant mutations poorly predicted by our methods, T315A and L298F, and find that starting configurations and protonation states significantly alter the accuracy of our predictions. Our experimental and computational measurements provide a benchmark for estimating the impact of mutations on inhibitor binding affinity for future methods and structure-based models. These structure-based methods have potential utility in identifying optimal therapies for tumor-specific mutations, predicting resistance mutations in the absence of clinical data, and identifying potential sensitizing mutations to established inhibitors.

biophysics↗

bioRxiv2024

Connectomic Organization of the Suprachiasmatic Nucleus

The suprachiasmatic nucleus (SCN), the mammalian master clock, is a special structure dedicated to the time computation. However, a connectomic understanding of the nucleus as a whole is lacking. Using serial section electron microscopy, we reconstructed multiscale and multimodal SCN communication networks. Intra-SCN synaptic network consists of 9,566 morphologically similar neurons and 4.3 million synapses, organized into multiscale circuitries that interact with SCN-traversing axon fascicles. Strikingly, SCN neurons engage in unique soma-soma ephaptic interaction, forming 2,038 electrotonic integrative units with the largest overlapping the light-responsive area (LRA). SCNs paracrine modality contains 47,396 m3 dense core vesicles, with notable scarcity in the LRA, suggesting cross-modal coordination and functional integration. These distinct network features provide comprehensive insights into the system architecture of the mammalian circadian clock.

neuroscience↗

bioRxiv2024

Interaction between histone residue H3K56 and mismatch repair protein MutSβ drives CAG repeat expansion

DNA repair protein MutS{beta} promotes CAG*CTG repeat expansions, which cause [~]20 untreatable neurodegenerative disorders including Huntingtons disease (HD). However, how MutS{beta} is recruited onto chromatin to enable repeat expansion is unknown. Here, we show that H3K56 interacts with MutS{beta} and brings it to chromatin. The H3K56-MutS{beta} interaction is regulated by H3K56 acetylation (H3K56ac) conditions, with acetylation inhibiting but deacetylation facilitating MutS{beta} chromatin recruitment. Blocking the H3K56-MutS{beta} interaction by either disrupting the MutS{beta} aromatic packet or increasing the H3K56ac level by removing histone deacetylases stabilizes CAG repeats, while depleting histone acetyltransferase p300/CBP promotes the H3K56-MutS{beta} interaction and CAG*CTG repeat instability, including expansion of CAG repeats of the huntingtin gene. Therefore, our study suggests a novel mechanism, by which MutS{beta} induces triplet repeat expansion, and new strategies of blocking the H3K56-MutS{beta} interaction to prevent HD and other triplet repeat expansion-caused disorders.

molecular biology↗

bioRxiv2024

Astrocyte ezrin defines resilience to stress-induced depressive behaviours in mice

Astrocyte atrophy is the main histopathological hallmark of major depressive disorder (MDD) in humans and in animal models of depression. Here we demonstrated that manipulating with ezrin expression specifically in astrocytes significantly increases the resilience of mice to chronic unpredictable mild stress (CUMS). Overexpression of ezrin in astrocytes from prefrontal cortex (PFC) rescued depressive-like behaviours induced by CUMS, whereas down-regulation of ezrin in astrocytes from PFC increased mice susceptibility to CUMS and promoted depressive-like behaviours. These behavioural changes correlated with astrocytic morphology. Astrocytes from PFC of mice sensitive to CUMS demonstrated significant atrophy; similar atrophy was found in astrocytes from animals with down-regulated ezrin expression. To the contrary morphology remains unchanged astrocytes in animals resistant to CUMS and in animals with astrocytic overexpression of ezrin. Morphological changes also correlated with ezrin immunoreactivity which was low in mice with depressive-like behaviours and high in mice resistant to stress. We conclude that Ezrin-dependent morphological remodelling of astrocytes defines the sensitivity of mice to stress: high ezrin expression renders them stress resilient, whereas low ezrin expression promotes depressive-like behaviour in response to chronic stress.

neuroscience↗

bioRxiv2024

Integrated Dual-Channel Retrograde Signaling Directs Stress Responses by Degrading the HAT1/TPL/IMPalpha-9 Suppressor Complex and Activating CAMTA3

The intricate communication between plastids and the nucleus, shaping stress-responsive gene expression, has long intrigued researchers. This study combines genetics, biochemical analysis, cellular biology, and protein modeling to uncover how the plastidial metabolite MEcPP activates the stress-response regulatory hub known as the Rapid Stress Response Element (RSRE). Specifically, we identify the HAT1/TPL/IMP- 9 suppressor complex, where HAT1 directly binds to RSRE and its activator, CAMTA3, masking RSRE and sequestering the activator. Stress-induced MEcPP disrupts this complex, exposing RSRE and releasing CAMTA3, while enhancing Ca2+ influx and raising nuclear Ca2+levels crucial for CAMTA3 activation and the initiation of RSRE- containing gene transcription. This coordinated breakdown of the suppressor complex and activation of the activator highlights the dual-channel role of MEcPP in plastid-to- nucleus signaling. It further signifies how this metabolite transcends its expected biochemical role, emerging as a crucial initiator of harmonious signaling cascades essential for maintaining cellular homeostasis under stress. SummaryThis study uncovers how the stress-induced signaling metabolite MEcPP disrupts the HAT1/TPL/IMP-9 suppressor complex, liberating the activator CAMTA3 and enabling Ca2+ influx essential for CAMTA3 activation, thus orchestrating stress responses via repressor degradation and activator induction.

plant biology↗

bioRxiv2024

Low hemoglobin causes hematoma expansion and poor intracerebral hemorrhage outcomes

ObjectivesAlthough lower hemoglobin levels associate with worse intracerebral hemorrhage (ICH) outcomes, causal drivers for this relationship remain unclear. We investigated the hypothesis that lower hemoglobin relates to increased hematoma expansion (HE) risk and poor outcomes using human observational data and assessed causal relationships using a translational murine model of anemia and ICH. MethodsICH patients with baseline hemoglobin measurements and serial CT neuroimaging enrolled between 2010-2016 to a multicenter, prospective observational cohort study were studied. Patients with systemic evidence of coagulopathy were excluded. Separate regression models assessed relationships of baseline hemoglobin with HE ([≥]33% and/or [≥]6mL growth) and poor long-term neurological outcomes (modified Rankin Scale 4-6) after adjusting for relevant covariates. Using a murine collagenase ICH model with serial neuroimaging in anemic vs. non-anemic C57/BL6 mice, intergroup differences in ICH lesion volume, ICH volume changes, and early mortality were assessed. ResultsAmong 1190 ICH patients analyzed, lower baseline hemoglobin levels associated with increased odds of HE (adjusted OR per -1g/dL hemoglobin decrement: 1.10 [1.02-1.19]) and poor 3-month clinical outcomes (adjusted OR per -1g/dL hemoglobin decrement: 1.11 [1.03-1.21]). Similar relationships were seen with poor 6 and 12-month outcomes. In our animal model, anemic mice had significantly greater ICH lesion expansion, final lesion volumes, and greater mortality, as compared to non-anemic mice. ConclusionsThese results, in a human cohort and a mouse model, provide novel evidence suggesting that anemia has causal roles in HE and poor ICH outcomes. Additional studies are required to clarify whether correcting anemia can improve these outcomes.

pathology↗

bioRxiv2024

Allopolyploidy expanded gene content but not pangenomic variation in the hexaploid oilseed Camelina sativa

Ancient whole-genome duplications (WGDs) are believed to facilitate novelty and adaptation by providing the raw fuel for new genes. However, it is unclear how recent WGDs may contribute to evolvability within recent polyploids. Hybridization accompanying some WGDs may combine divergent gene content among diploid species. Some theory and evidence suggest that polyploids have a greater accumulation and tolerance of gene presence-absence and genomic structural variation, but it is unclear to what extent either is true. To test how recent polyploidy may influence pangenomic variation, we sequenced, assembled, and annotated twelve complete, chromosome-scale genomes of Camelina sativa, an allohexaploid biofuel crop with three distinct subgenomes. Using pangenomic comparative analyses, we characterized gene presence-absence and genomic structural variation both within and between the subgenomes. We found over 75% of ortholog gene clusters are core in Camelina sativa and <10% of sequence space was affected by genomic structural rearrangements. In contrast, 19% of gene clusters were unique to one subgenome, and the majority of these were Camelina-specific (no ortholog in Arabidopsis). We identified an inversion that may contribute to vernalization requirements in winter-type Camelina, and an enrichment of Camelina-specific genes with enzymatic processes related to seed oil quality and Camelinas unique glucosinolate profile. Genes related to these traits exhibited little presence-absence variation. Our results reveal minimal pangenomic variation in this species, and instead show how hybridization accompanied by WGD may benefit polyploids by merging diverged gene content of different species.

genomics↗

bioRxiv2024

Reconstitution of chromatin reorganization during mammalian oocyte development

SignificanceThe generation of developmentally competent eggs requires oocyte growth and maturation. When compared to meiotic maturation, relatively little is known about the key processes that happen in growing oocytes. Since the first visualization of oocyte chromatin under the electron microscope, we have known that growing oocytes undergo a fundamental transition from the non-surrounded nucleolus (NSN) to surrounded nucleolus (SN) chromatin configuration. NSN-to-SN transition is crucial for subsequent embryonic development after fertilization, but what drives NSN-to-SN transition during mammalian oocyte development is still unknown. Here we show that natural degradation of RNA polymerase II (RNAPII) by segregase and proteasome drives NSN-to-SN transition, challenging the conventional view that transcriptional silencing and NSN-to-SN transition are driven by independent epigenetic mechanisms in mammalian oocytes. The final step of oocyte growth, which reorganizes chromatin from the uncondensed, non-surrounded nucleolus (NSN) configuration to the condensed, surrounded nucleolus (SN) configuration, is essential for embryonic development after fertilization. However, the mechanism of NSN-to-SN transition remains unknown. Here, we identify RNA polymerase II (RNAPII) degradation as the key driver of this process. RNAPII inhibitors, but not nucleoside-based transcription inhibitors, swiftly induced RNAPII degradation and NSN-to-SN transition in oocytes. Strikingly, induced SN-like nuclei recapitulated epigenetic features, chromatin interactions and developmental potential in SN oocytes. Using segregase and proteasome inhibitors and the newly established scFv miniTrim-Away for nuclear proteins, we further demonstrate that RNAPII degradation is necessary and sufficient for NSN-to-SN transition in mouse and human. RNAPII degradation globally cleared RNAPII from transcriptionally active chromatin and increased chromatin dynamics, concomitantly leading to transcriptional silencing and chromatin rearrangement in oocytes. Our study elucidates the framework of chromatin reorganization in growing mammalian oocytes and provides an alternative source of fully grown oocyte nuclei, representing a milestone for the potential use of growing oocytes in reproductive medicine.

developmental biology↗

bioRxiv2024

Long-Range Enhancer Networks Gradually Emerge as Key Regulators During Human Cortical Neurogenesis

The chromatin in the human brain cortex shows more enhancer-enhancer contacts than in macaques and mice, yet the organization of these contacts across cellular states and their mechanisms remain unclear. Here, we developed SCOPE-C, a cost-effective technique designed for mapping DNase I Hypersensitive Sites and their spatial interactions, with the potential for scalability to single-cell resolution. Employing SCOPE-C, we generated a comprehensive chromatin map detailing cis-regulatory elements in four cell types during neurogenesis in human, macaque, and mouse fetal cortex. This map introduces a model of long-range enhancer networks that regulate gene transcription during human cortical neurogenesis. In this model, CTCF-mediated loop extrusion forms stripes of enhancer interaction hubs extending up to 10 Mb in human Excitatory Neurons. Enriched with human-specific enhancers and neuropsychiatric disorder-linked SNPs, these remarkably long-range networks predominantly control EN marker genes, suggesting their critical role in robust gene expression during human cortical neurogenesis.

genomics↗

bioRxiv2024

BOOST: a robust ten-fold expansion method on hour-scale

Expansion microscopy (ExM) enhances the microscopy resolution by physically expanding biological specimens and improves the visualization of structural and molecular details. Numerous ExM techniques and labeling methods have been developed and refined over the past decade to cater to specific research needs. Nonetheless, a shared limitation among current protocols is the extensive time required for sample processing, particularly for challenging-to-expand biological specimens (e.g., formalin-fixed paraffin-embedded (FFPE) sections and large three-dimensional specimens). Here, we have developed a rapid and robust ExM workflow named BOOST, which leverages a series of novel microwave (MW)-accelerated ExM chemistry, resulting in a single-step linear expansion of [~]10x. Specifically, BOOST facilitates a [~]10-fold expansion of cultured cells, tissue sections, and even the challenging-to-expand FFPE sections under merely 90 minutes with heat and surfactant-based protein denaturation. Furthermore, BOOST employs microwave-assisted proteomic staining and immunostaining to facilitate high-resolution visualization of structural and molecular details with significantly enhanced throughput. Noteworthily, BOOST has pioneered a [~]10-fold expansion of large millimeter-sized three-dimensional specimens in approximately three hours. BOOST offers an easily adaptable workflow based on stable and common reagents, thus boosting the potential adoption of ExM methods in biological investigations.

bioengineering↗

bioRxiv2024

Strawberry COP9 signalosome FvCSN5 regulates plant development and fruit ripening by facilitating polyamine oxidase FvPAO5 degradation to control polyamine and H2O2 homeostasis

Polyamines (PAs), mainly including putrescine, spermidine, and spermine, are essential for plant growth and development. However, the post-translational regulation of PA metabolism remains unknown. Here, we report the COP9 signalosome subunit 5A (FvCSN5A)-mediated degradation of the PA oxidase FvPAO5, which catalyzes the conversion of spermidine/spermine to produce H2O2. FvCSN5A is localized in the cytoplasm and nucleus, ubiquitously expressed in strawberry plants, and rapidly increases during fruit ripening. FvCSN5A RNA interference (RNAi) transgenic strawberries exhibit pleiotropic effects on plant development, fertility, and fruit ripening by regulating PA and H2O2 homeostasis, similar to FvPAO5 overexpression transgenic lines. FvCSN5A interacts with FvPAO5 in vitro and in vivo. The ubiquitination and degradation of FvPAO5 are impaired in FvCSN5A RNAi lines. Additionally, FvCSN5A interacts with cullin 1 (FvCUL1), a core component of the E3 ubiquitin-protein ligase complex. Transient genetic analysis in cultivated strawberry fruits showed that inhibiting FaPAO5 expression could partially rescue the ripening phenotype of FaCSN5A RNAi fruits. Collectively, the CSN5A-CUL1-PAO5 signaling pathway responsible for PA and H2O2 homeostasis is crucial to strawberry vegetative and reproductive growth and, final to fruit ripening. Our findings may open a promising avenue for improving crop yield and quality by manipulating the CSN5-PAO5 pair.

plant biology↗

bioRxiv2024

Targeting NOTCH2-JAG1 juxtacrine signaling reverses macrophage-mediated tumor resistance to taxol

Taxanes are widely used in chemotherapy, but primary and acquired resistance limit the clinical efficacy. Studies have shown tumor interaction with macrophages in the tumor microenvironment (TME) plays a significant role in taxane resistance, yet therapeutic strategies that directly deplete or repolarize macrophages are challenging and with considerable risk of side effects. Here we uncovered that tumor-macrophage interaction can be selectively targeted by inhibiting post-mitotic NOTCH2-JAG1 juxtacrine signaling in the TME, which strongly sensitizes paclitaxel response. Using translatome profiling, we found significant NOTCH2 upregulation during paclitaxel-induced prolonged mitosis. NOTCH2 was subsequently activated in the post-mitotic G1 phase by JAG1 expressed on the neighboring macrophages and tumor cells, which promoted tumor cell survival and upregulated cytokines that recruited JAG1-expressing macrophages, thus generating a positive feedback loop that further enhanced the pro-tumor NOTCH2 activity. By targeting this NOTCH2-JAG1 axis using NOTCH2 shRNA or a pan-NOTCH inhibitor, macrophage recruitment and paclitaxel resistance were significantly attenuated in multiple mouse tumor models of ovarian cancer. Clinical samples from paired primary and recurrent ovarian cancer patients also showed significant correlation of higher NOTCH2 expression with worse prognosis. Our results thus point to combining NOTCH2 inhibitor with taxane as an effective therapeutic strategy to selectively disrupt tumor-macrophage interaction in the TME and overcome macrophage-mediated taxane resistance in NOTCH2-positive tumors.

cancer biology↗

bioRxiv2024

Human cancer genomes harbor the mutational signature of tobacco-specific nitrosamines NNN and NNK

Tobacco usage is linked to multiple cancer types and accounts for a quarter of all cancer-related deaths. Tobacco smoke contains various carcinogenic compounds, including polycyclic aromatic hydrocarbons (PAH), though the mutagenic potential of many tobacco-related chemicals remains largely unexplored. In particular, the highly carcinogenic tobacco-specific nitrosamines NNN and NNK form pre-mutagenic pyridyloxobutyl (POB) DNA adducts. In the study presented here, we identified genome-scale POB-induced mutational signatures in cell lines and rat tumors, while also investigating their role in human cancer. These signatures are characterized by T>N and C>T mutations forming from specific POB adducts damaging dT and dC residues. Analysis of 2,780 cancer genomes uncovered POB signatures in [~]180 tumors; from cancer types distinct from the ones linked to smoking-related signatures SBS4 and SBS92. This suggests that, unlike PAH compounds, the POB pathway may contribute uniquely to the mutational landscapes of certain hematological malignancies and cancers of the kidney, breast, prostate and pancreas.

genomics↗

bioRxiv2024

Linalool's Multifaceted Antimicrobial Potential: Unveiling its Antimicrobial Efficacy and Immunomodulatory Role Against Saprolegnia parasitica

Saprolegnia parasitica (S. parasitica) is one of the most virulent oomycete species in freshwater aquatic environments, causing severe saprolegniasis and leading to significant economic losses in the aquaculture industry. Thus far, the prevention and control of saprolegniasis face a shortage of medications. Linalool, a natural antibiotic alternative found in various essential oils, exhibits promising antimicrobial activity against a wide range of pathogens. In this study, the specific role of linalool in protecting S. parasitica infection at both in vitro and in vivo levels was investigated. Linalool showed multifaceted anti-oomycetes potential by both of antimicrobial efficacy and immunomodulatory efficacy. For in vitro test, linalool exhibited strong anti-oomycetes activity and mode of action included: (1) Linalool disrupted the cell membrane of the mycelium, causing the intracellular components leak out; (2) Linalool prohibited ribosome function, thereby inhibiting protein synthesis and ultimately affecting mycelium growth. Surprisingly, meanwhile we found the potential immune protective mechanism of linalool in the in vivo test: (1) Linalool enhanced the complement and coagulation system which in turn activated host immune defense and lysate S. parasitica cells; (2) Linalool promoted wound healing, tissue repair, and phagocytosis to cope with S. parasitica infection; (3) Linalool positively modulated the immune response by increasing the abundance of beneficial Actinobacteriota; (4) Linalool stimulated the production of inflammatory cytokines and chemokines to lyse S. parasitica cells. In all, our findings showed that linalool possessed multifaceted anti-oomycetes potential which would be a promising natural antibiotic alternative to cope with S. parasitica infection in the aquaculture industry.

microbiology↗

bioRxiv2024

Identification of BiP as a temperature sensor mediating temperature-induced germline sex reversal

Besides regulation by karyotype, sex determination is also modulated by environmental cues like temperature, but the involving temperature-transduction mechanism remains elusive. Moreover, while sex determination was traditionally seen as dictated exclusively by either karyotype or temperature, recent observations suggest these factors can co-regulate sex, posing a mechanistic mystery. Here, we discovered that certain wild-isolated and mutant C. elegans strains displayed genotypic-germline-sex-determination (GGSD) but with a temperature override. We found that ER chaperone BiP transduces temperature into germline-sex-governing signal and enables the co-existence of GGSD and temperature-dependent-germline-sex-determination (TGSD). Mechanistically, BiP availability is reduced at warmer temperatures through detecting increased ER-protein-folding burden, which promotes male-germline-fate through ERAD-mediated downregulation of the oocyte-fate driving factor, TRA-2. Remarkably, we can induce a switch between GGSD and TGSD by manipulating this newly-discovered process. Functionally, TGSD facilitates C. elegans hermaphrodites in maintaining brood size at warmer temperatures. Moreover, BiP also acts in germline-sex-determination in a dioecious nematode. Collectively, our findings identify thermosensitive BiP as a conserved temperature sensor in TGSD and provide mechanistic insights into the fascinating transition of GGSD and TGSD.

developmental biology↗

bioRxiv2024

Direcred Micro-Ecology Application System: On-site One-step Fermentation Facilitating Microbial Application through Theoretical and Equipment Innovation

Functional microbial agents play a crucial role in various fields such as agriculture, feed fermentation, aquaculture, and environmental protection. However, traditional microbial agents were confronted with critical challenges such as limited shelf-life, reduced activity, and inconsistent efficacy. In this case, we innovatively proposed the concept of Directed Micro-Ecology (DME) and developed its application system, including a core module named DME intelligent fermentor (DME25). Over 40 functional strains, including bacterial strains and fungus strains, were successfully cultured to 10[~]50 x108 CFU/mL within 20[~]48 h and maintained a relatively low contamination rate (<2.5%). Finally, the stability and effectiveness of these DME-fermented strains were validated in different application areas, all of which exhibited perfect functional characteristics. Firstly, the bacillus strains inhibited the progression of wilt disease and significantly improved the growth of tomatoes. Secondly, all tested lactobacillus strains improved the nutrition and quality of fermented feed, complying with feed industry standards. Lastly, the ammonia nitrogen concentration, nitrite concentration of aquaculture water and phosphate concentration, COD of aquaculture tail water were significantly reduced within 1[~]4 d. The successful application of the DME intelligent fermentor in different fields marks a pivotal breakthrough in technological innovation of microbial agents on-site one-step fermentation. This technological advancement opens new avenues for enhancing the stability and effectiveness of microbial agents, infusing powerful impetus to the development of microbial application.

microbiology↗

bioRxiv2024

Selective Labeling Meets Semi-Supervised Neuron Segmentation

Semi-supervised learning offers a cost-effective approach for neuron segmentation in electron microscopy (EM) volumes. This technique leverages extensive unlabeled data to regularize supervised training for more robust predictions of neuron boundaries. However, the distribution mismatch between labeled and unlabeled data, due to limited annotations and diverse neuronal structures, can cause unreliable predictions on unlabeled data, thus limiting the generalization of semi-supervised models. In this paper, we develop a distribution-aware pipeline to address the inherent mismatch issue and enhance semi-supervised neuron segmentation in EM volumes. At the data level, we propose an unsupervised heuristic to select valuable sub-volumes as labeled data based on distribution similarity in a pretrained feature space, ensuring a representative coverage of neuronal structures. At the model level, we introduce an axial-through mixing strategy into anisotropic neuron segmentation and integrate it into a semi-supervised framework. Building on this, we establish cross-view consistency constraints through intra- and inter-mixing of labeled and unlabeled data, which facilitates the learning of shared semantics across distributions. Extensive experiments on public EM datasets from multiple species, imaging modalities, and resolutions demonstrate the effectiveness of our method. Codes and demos are available at https://github.com/yanchaoz/SL-SSNS.

neuroscience↗