Biology subjects

Walter, T.

Publications and source records attributed to Walter, T..

At least 19 recordsLinked to original sources

bioRxiv2026

An optimized workflow for spatial transcriptomics across early development in Xenopus

How gene expression patterns change spatially as the embryo transitions from simple to complex structures remains a major developmental biology question. Recently developed imaging-based spatial transcriptomics (ST) enable mapping expression of multiple gene at a single-cell resolution. Although Xenopus is a key model in embryology there is no established ST pipeline, and commercially available techniques face many challenges (sample preparation, probe design, cell segmentation). Furthermore, the highly diverse cell shapes and sizes across developmental stages and between different tissues represent major hurdles to accurately defining cells. Here, we describe an optimized workflow for ST in blastula-to-tailbud-stage frog embryos using Merscope, commercial MERFISH (Multiplexed Error-Robust Fluorescence In Situ Hybridization) originally designed for standard mammalian tissues. With stringent quality control and tailored computational pipelines, we optimize this technology for robust, semi-quantitative profiling of spatial transcriptomic landscapes in non-mammalian embryos. Reliable tissue preservation and cell-segmentation enable high-resolution mapping of gene expression during the development of a complex multi-tissue organization. This versatile strategy applies broadly to various dynamic systems, from embryos of various model organisms to complex and heterogeneous organs in mammals. Summary statementThis Single-cell Spatial Transcriptomics pipeline and reference atlas in Xenopus - a model organism in embryology - overcome technical challenges and resolve dynamic changes in patterning during development.

developmental biology↗

bioRxiv2026

HEDeST: An Integrative Approach to Enhance Spatial Transcriptomic Deconvolution with Histology

Spatial organization of cells is essential for tissue function, yet sequencing-based spatial transcriptomics often lacks single-cell resolution. We present HEDeST, a weakly supervised framework that integrates histology-derived morphological features with deconvolution-derived spot-level proportions to assign cell types at single-cell resolution. HEDeST is robust to technical variability, adaptable to user-defined cell types, and compatible with any deconvolution method. Across simulated and semi-simulated datasets, HEDeST outperforms existing morphology-based approaches and reveals biologically meaningful microenvironments when applied to real cancer datasets, providing a scalable tool for high-resolution spatial tissue analysis.

bioinformatics↗

bioRxiv2025

Deep Learning Bridges Histology and Transcriptomics to Predict Molecular Subtypes and Outcomes in Muscle-Invasive Bladder Cancer

Muscle-Invasive Bladder Cancer (MIBC) is a heterogeneous disease with distinct molecular subtypes influencing prognosis and therapeutic response. However, molecular profiling through RNA sequencing remains costly, time-consuming and complicated by intratumoral heterogeneity. We developed a Deep Learning (DL) approach to infer molecular subtypes from routine histopathological slides and to evaluate its prognostic value in patients treated with neoadjuvant chemotherapy (NAC). We developed an DL-model predicting the expression of 848 subtype-associated genes from histological images of transurethral resection of bladder tumor, enabling spatial molecular subtyping at tile level. The model was trained on 297 NAC-treated patients from the VESPER clinical trial and evaluated on three independent cohorts (COBLAnCE, n=224; Saint-Louis, n=30 and TCGA, n=315), covering diverse staining protocols and scanner types. Spatial transcriptomics from six VESPER patients confirmed the spatial consistency of the inferred expression profiles. Our approach achieved a ROC AUC of 0.94 for molecular subtype prediction, with 95% of genes significantly predicted, demonstrating its ability to capture transcriptomic dysregulations from histological morphology. Predicted expression maps revealed spatially coherent patterns and intratumoral molecular heterogeneity. Importantly, tumors predicted with basal/squamous features (pure or mixed), were associated with significantly worse progression-free and overall survival after NAC (log-rank p=0.014 and 0.037, respectively). This DL-based framework enables accurate and spatially resolved inference of gene expression and molecular subtypes in MIBC without sequencing. These findings could improve patient stratification in clinical practice and support the design of more targeted clinical trials. Further validation in larger cohorts is needed before routine clinical implementation.

cancer biology↗

bioRxiv2025

Cut-Detector: A Tool for Automated Temporal Analysis of Late Cytokinetic Events

Cytokinesis is the final step of cell division, resulting in the physical separation of the two daughter cells. Despite its fundamental importance in cell biology, biologists currently lack automatic tools for large-scale profiling of cytokinesis dynamics. In particular, the timing of the first microtubule cut in the intercellular bridge connecting the daughter cells is crucial, as it marks a critical step preceding abscission. Here, we introduce Cut-Detector, an open-source tool for the automatic analysis of late cytokinesis timing from time-lapse microscopy movies. Cut-Detector employs an AI approach to carry out the multiple tasks required to monitor cytokinesis: cell segmentation and tracking, detection of cell division events, localization of the intercellular bridge, and detection of the microtubule cuts. Cut-Detector will facilitate large-scale analyses to uncover new cytokinetic genes, a task that would be impractical without automation.

bioinformatics↗

bioRxiv2025

A Deep Learning approach for time-consistent cell cycle phase prediction from microscopy data

The cell cycle consists of four phases and impacts most cellular processes. In imaging assays, the cycle phase can be identified using dedicated cell-cycle markers. However, such markers occupy fluorescent channels that may be needed for other reporters. Here, we propose to address this limitation by inferring the phase from a widely used fluorescent reporter: SiR-DNA. Our method is based on a variational auto-encoder, enhanced with two auxiliary tasks: predicting the intensity of phase-specific markers and enforcing the latent space temporal consistency. Our model is freely available, along with a new dataset comprising over 600,000 annotated HeLa Kyoto nuclear images.

bioinformatics↗

bioRxiv2025

RNA2seg: a generalist model for cell segmentation in image-based spatial transcriptomics

Imaging-based spatial transcriptomics (IST) enables high-resolution spatial mapping of RNA species. A key challenge in IST is accurate cell segmentation to assign each RNA molecule to the right cell. Here, we present RNA2seg, a novel segmentation algorithm trained on over 4 million cells from MERFISH and CosMx datasets across seven organs using a teacher-student training scheme. RNA2seg integrates RNA point clouds and all available membrane and nuclear stainings. Validation on manually annotated data shows superior performance including in zero-shot and few-shot settings. The method is available as a documented pip package: https://github.com/fish-quant/rna2seg.

bioinformatics↗

bioRxiv2024

autoFISH - a modular toolbox for sequential smFISH experiments

Fluorescence in situ hybridization (FISH) allows for spatial and quantitative profiling of gene expression by visualizing individual RNA molecules. Here, we introduce automated FISH (autoFISH), a comprehensive toolbox to conduct automated single molecule FISH (smFISH) experiments that is both cost-effective and versatile. This includes detailed plans for constructing the necessary equipment, open-source software for control, reliable experimental protocols, and analysis workflows based on our FISH-quant analysis package. Validation experiments with both cell lines and tissue samples confirmed the systems robustness. We demonstrate standard and amplified smFISH, along with a modified protocol for tissue clearing that enhances nuclear retention while preserving background reduction efficiency.

molecular biology↗

bioRxiv2024

sCellST: a Multiple Instance Learning approach to predict single-cell gene expression from H&E images using spatial transcriptomics

Advancing our understanding of tissue organization and its disruptions in disease remains a key focus in biomedical research. Histological slides stained with Hematoxylin and Eosin (H&E) provide an abundant source of morphological information, while Spatial Transcriptomics (ST) enables detailed, spatiallyresolved gene expression (GE) analysis, though at a high cost and with limited clinical accessibility. Predicting GE directly from H&E images using ST as a reference has thus become an attractive objective; however, current patch-based approaches lack single-cell resolution. Here, we present sCellST, a multipleinstance learning model that predicts GE by leveraging cell morphology alone, achieving remarkable predictive accuracy. When tested on a pancreatic ductal adenocarcinoma dataset, sCellST outperformed traditional methods, underscoring the value of basing predictions on single-cell images rather than tissue patches. Additionally, we demonstrate that sCellST can detect subtle morphological differences among cell types by utilizing marker genes in ovarian cancer samples. Our findings suggest that this approach could enable single-cell level GE predictions across large cohorts of H&E-stained slides, providing an innovative means to valorize this abundant resource in biomedical research.

bioinformatics↗

bioRxiv2024

Contrastive learning for cell division detection and tracking in live cell imaging data

Fluorescent live-cell microscopy is essential for understanding cellular dynamics by imaging specific molecules, their interactions, and biochemical states in live samples. It is crucial for biological research and drug screening. However, live-cell imaging often requires balancing temporal resolution with cell viability (i.e. conditions enabling cells to thrive) because of photo-toxicity. Consequently, there is increasing interest in lowering temporal resolution to allow for extended observation periods and to study event sequences that uncover causal relationships and mechanistic insights. Tracking cells in video microscopy with low temporal resolution remains a challenge. We introduce a new integrated methodology that uses contrastive learning and graph-based approaches to improve cell division detection and tracking. Our approach employs contrastive learning models to create cell representations that facilitate the detection of cell divisions and enhance cell tracking. Of note, we propose a weakly-supervised constrastive learning approach to build robust temporal cell representations through time-based augmentations. Additionally, we introduce an innovative graph optimization technique to identify cell tracks based on these representations and observed division events. We evaluate our methods on an in-house dataset and public datasets from the Cell Tracking Challenge, achieving substantial performance improvements in both native and reduced temporal resolutions. Our methodology thus enhances adaptability to various temporal resolutions, improving precision and efficiency in live-cell microscopy analysis. This advancement is particularly beneficial for extended drug screening studies, ensuring cell viability and maintaining normal cell homeostasis, which is vital for therapeutic research.

bioinformatics↗

bioRxiv2024

The TissueTractor, a device for applying large strains to tissues and cells for simultaneous high-resolution live cell microscopy.

Mechanical strain substantially influences tissue shape and function in various contexts, from embryonic development to disease progression. Disruptions in these processes can result in congenital abnormalities and short-circuit mechanotransduction pathways. Manipulating strain in live tissues is crucial for understanding its impact on cellular and subcellular activities. Existing tools, such as optogenetic modulation of strain, are limited to small strain over limited distance and durations. Here, we introduce a high-strain stretcher system, the TissueTractor, designed for high-resolution spatiotemporal imaging of live tissues, enabling strain application varying from 0% to over 150%. This system is needed to unravel the intricate connections between mechanical forces and developmental processes. We demonstrated the stretcher with Xenopus laevis organotypic explants, human umbilical endothelial cells, and mouse neonatal cardiomyocytes to highlight the stretchers adaptability. These demonstrations underscore the potential of this stretcher to deepen our understanding of the mechanical cues governing tissue dynamics and morphogenesis. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=170 SRC="FIGDIR/small/600827v3_ufig1.gif" ALT="Figure 1"> View larger version (46K): org.highwire.dtl.DTLVardef@207163org.highwire.dtl.DTLVardef@1f424e7org.highwire.dtl.DTLVardef@1a2e9e1org.highwire.dtl.DTLVardef@a8d785_HPS_FORMAT_FIGEXP M_FIG C_FIG

biophysics↗

bioRxiv2024

Cell cycle-dependent mRNA localization in P-bodies

Understanding the dynamics of RNA targeting to membraneless organelles is essential to disentangle their functions. Here, we investigate how P-bodies (PBs) evolve during cell cycle progression. PB purification across the cell cycle uncovers widespread changes in their RNA content, which are partly uncoupled from cell cycle-dependent changes in RNA expression. Single molecule FISH shows various mRNA localization patterns in PBs peaking in G1, S, or G2, with examples illustrating the timely capture of mRNAs in PBs when their encoded protein becomes dispensable. Yet, rather than directly reflecting absence of translation, cyclic mRNA localization in PBs can be controlled by RBPs, such as HuR in G2, and by RNA features. Indeed, while PB mRNAs are AU-rich at all cell cycle phases, they are specifically longer in G1, possibly related to post-mitotic PB reassembly. Altogether, our study supports a model where PBs are more than a default location for excess untranslated mRNAs.

cell biology↗

bioRxiv2024

Self-supervised learning to predict intrahepatic cholangiocarcinoma transcriptomic classes on routine histology

ObjectiveThe transcriptomic classification of intrahepatic cholangiocarcinomas (iCCA) has been recently refined from two to five classes, associated with pathological features, targetable genetic alterations and survival. Despite its prognostic and therapeutic value, the classification is not routinely used in the clinic because of technical limitations, including insufficient tissue material or the cost of molecular analyses. Here, we assessed a self-supervised learning (SSL) model for predicting iCCA transcriptomic classes on whole-slide digital histological images (WSIs) DesignTranscriptomic classes defined from RNAseq data were available for all samples. The SSL method, called Giga-SSL, was used to train our model on a discovery set of 766 biopsy slides (n=137 cases) and surgical samples (n=109 cases) from 246 patients in a five-fold cross-validation scheme. The model was validated in The Cancer Genome Atlas (TCGA) (n= 29) and a French external validation set (n=32). ResultsOur model showed good to very good performance in predicting the four most frequent transcriptomic class in the discovery set (area under the curve [AUC]: 0.63-0.84), especially for the hepatic stem-like class (37% of cases, AUC 0.84). The model performed equally well in predicting these four transcriptomic classes in the two validation sets, with AUCs ranging from 0.76 to 0.80 in the TCGA set and 0.62 to 0.92 in the French external set. ConclusionWe developed and validated an SSL-based model for predicting iCCA transcriptomic classes on routine histological slides of biopsy and surgical samples, which may impact iCCA management by predicting prognosis and guiding the treatment strategy.

cancer biology↗

bioRxiv2023

Democratizing computational pathology: optimized Whole Slide Image representations for The Cancer Genome Atlas

Automatic analysis of hematoxylin and eosin (H&E) stained Whole Slide Images (WSI) bears great promise for computer assisted diagnosis and biomarker discovery. However, scarcity of annotated datasets leads to underperforming models. Furthermore, the size and complexity of the image data limit their integration into bioinformatic workflows and thus their adoption by the bioinformatics community. Here, we present Giga-SSL, a self-supervised method for learning WSI representations without any annotation. We show that applying a simple linear classifier on the Giga-SSL representations improves classification performance over the fully supervised alternative on five benchmarked tasks and across different datasets. Moreover, we observe a substantial performance increase for small datasets (average gain of 7 AUC point) and a doubling of the number of mutations predictable from WSIs in a pan-cancer setting (from 45 to 93). We make the WSI representations available, compressing the TCGA-FFPE images from 12TB to 23MB and enabling fast analysis on a laptop CPU. We hope this resource will facilitate multimodal data integration in order to analyze WSI in their genomic and transcriptomic context.

bioinformatics↗

bioRxiv2023

A point cloud segmentation framework for image-based spatial transcriptomics

Recent progress in image-based spatial RNA profiling enables to spatially resolve tens to hundreds of distinct RNA species with high spatial resolution. It hence presents new avenues for comprehending tissue organization. In this context, the ability to assign detected RNA transcripts to individual cells is crucial for downstream analyses, such as in-situ cell type calling. Yet, accurate cell segmentation can be challenging in tissue data, in particular in the absence of a high-quality membrane marker. To address this issue, we introduce ComSeg, a segmentation algorithm that operates directly on single RNA positions and that does not come with implicit or explicit priors on cell shape. ComSeg is thus applicable in complex tissues with arbitrary cell shapes. Through comprehensive evaluations on simulated datasets, we show that ComSeg outperforms existing state-of-the-art methods for in-situ single-cell RNA profiling and cell type calling. On experimental data, our method also demonstrates proficiency in estimating RNA profiles that align with established scRNA-seq datasets. Importantly, ComSeg exhibits a particular efficiency in handling complex tissue, positioning it as a valuable tool for the community.

bioinformatics↗

bioRxiv2023

pyHiM, a new open-source, multi-platform software package for spatial genomics based on multiplexed DNA-FISH imaging

The three-dimensional (3D) nuclear organization of chromatin in eukaryotes plays a crucial role in gene regulation, DNA replication, and DNA damage repair. While genome-wide ensemble methods have enhanced our understanding of chromatin organization, they lack the ability to capture single-cell heterogeneity and preserve spatial information. To overcome these limitations, a new family of imaging-based methods has emerged, giving rise to the field of spatial genomics. In this study, we present pyHiM, an open-source and modular software toolbox specifically designed for the robust, automatic analysis of sequential spatial genomics data. pyHiM enables the reconstruction of chromatin traces in individual cells from raw, multicolor images, offering novel, robust and validated algorithms, extensive documentation, and tutorials. Its user-friendly graphical interface and command-line interface allow for easy installation and execution on various hardware platforms. The software employs a modular architecture, allowing independent execution of analysis steps and customization according to sample specificity and computing resources. pyHiM supports preprocessing, spot detection, mask detection, and trace generation, generating human-readable reports and intermediate results for data validation and further analysis. Moreover, it offers additional features for data formatting, result display, and post-processing. pyHiMs scalability and parallelization capabilities enable the analysis of large, complex datasets in a reasonable time frame. Overall, pyHiM aims to facilitate the democratization and standardization of spatial genomics analysis, foster collaborative developments, and promote the growth of a user community to drive discoveries in the field of chromatin organization.

genomics↗

bioRxiv2023

Cell Segmentation in Images Without Structural Fluorescent Labels

High-content screening (HCS) provides an excellent tool to understand the mechanism of action of drugs on disease-relevant model systems. Careful selection of fluorescent labels (FLs) is crucial for successful HCS assay development. HCS assays typically comprise (1) FLs containing biological information of interest, and (2) additional structural FLs enabling instance segmentation for downstream analysis. However, the limited number of available fluorescence microscopy imaging channels restricts the degree to which these FLs can be experimentally multiplexed. In this paper, we present a segmentation workflow that overcomes the dependency on structural FLs for image segmentation, typically freeing 2 fluorescence microscopy channels for biologically relevant FLs. It consists in extracting structural information encoded within readouts that are primarily biological, by fine-tuning pre-trained state-of-the-art generalist cell segmentation models for different combinations of individual FLs, and aggregating the respective segmentation results together. Using annotated datasets that we provide, we confirm our methodology offers improvements in performance and robustness across several segmentation aggregation strategies and image acquisition methods, over different cell lines and various FLs. It thus enables the biological information content of HCS assays to be maximized without compromising the robustness and accuracy of computational single-cell profiling. Impact StatementThis methodological paper describes a framework enabling cell segmentation for datasets without structural fluorescent labels to highlight cell organelles. Such capabilities favorably impact costs and possible discoveries in single-cell downstream analysis by improving our ability to incorporate more biological read-outs into a single assay. The perspective of computational and experimental biologist coauthors ensures a multidisciplinary viewpoint and accessibility for a wide readership.

bioinformatics↗

bioRxiv2022

Diagnosis with Confidence: Deep Learning for Reliable Classification of Squamous Lesions of the Upper Aerodigestive Tract

BackgroundDiagnosis of head and neck (HN) squamous dysplasias and carcinomas is critical for patient care cure and follow-up. It can be challenging, especially for grading intraepithelial lesions. Despite recent simplification in the last WHO grading system, the inter- and intra-observer variability remains substantial, particularly for non-specialized pathologists, exhibiting the need for new tools to support pathologists. MethodsIn this study we investigated the potential of deep learning to assist the pathologist with automatic and reliable classification of HN lesions following the 2022 WHO classification system. We created, for the first time, a large-scale database of histological samples (>2000 slides) intended for developing an automatic diagnostic tool. We developed and trained a weakly supervised model performing classification from whole slide images (WSI). We evaluated our model on both internal and external test sets and we defined and validated a new confidence score to assess the predictions which can be used to identify difficult cases. ResultsOur model demonstrated high classification accuracy across all lesion types on both internal and external test sets (respectively average AUC: 0.878 (95% CI:[0.834-0.918]) and 0.886 (95% CI: [0.813-0.947])) and the confidence score allowed for accurate differentiation between reliable and uncertain predictions. ConclusionsOur results demonstrate that the model, associated with confidence measurements, can help in the difficult task of classifying head and neck squamous lesions by limiting variability and detecting ambiguous cases, taking us one step closer to a wider adoption of AI-based assistive tools.

bioinformatics↗

bioRxiv2022

Streetlights affect moth orientation beyond flight-to-light behaviour

One of the most dramatic changes occurring on our planet in recent decades is the ever-increasing extensive use of artificial light at night, which drastically altered the environment nocturnal animals are adapted to 1,2. One nocturnal species group experiencing marked declines are moths, which are not only of great importance for species conservation, but also for their key role in food webs and in ecosystem services such as nocturnal plant pollination 3,4. Light pollution has been identified as a driver in the dramatic insect decline of the past years 5-7, yet little is known about its impact on natural insect orientation behaviour. Using harmonic radar tracking, we show that the orientation of several species of moths is significantly affected by streetlights, although only 4 % of individuals showed flight-to-light behaviour. We reveal a species-specific barrier effect of streetlights on lappet moths whenever the moon was not available as a natural celestial cue. Furthermore, streetlights increased the tortuosity of flight trajectories for both hawk moths and lappet moths. Our results provide the first spatially resolved experimental evidence for the fragmentation of landscapes by streetlights and demonstrate that light pollution affects movement patterns of moths beyond previously assumed extend, potentially affecting their reproductive success and hampering a vital ecosystem service.

animal behavior and cognition↗