Biology subjects

Horvath, P.

Publications and source records attributed to Horvath, P..

At least 37 records · Page 2Linked to original sources

bioRxiv2022

Show me your neighbour and I tell what you are: fisheye transformation for deep learning-based single-cell phenotyping

Recently we have concluded that image-based features derived from the microenvironment have an enormous impact on successfully determining the class of an object1. Here we demonstrate that deep learning-based phenotypic analysis of cells with a properly chosen microenvironment-size provides results comparable to our earlier neighbourhood-based methods that utilise hand-crafted image features. We hypothesised that treating cells with equal weight, regardless of their position within the cellular microenvironment, is suboptimal, and direct neighbours have a larger impact on the phenotype of the cell-of-interest than cells in its larger proximity. Hence we present a novel approach that (1) considers the fully featured view of the cell-of-interest, (2) includes the neighbourhood and (3) gives lesser weight to cells that are far from the cell. To achieve this, we present a transformation similar to those characteristic for fisheye cameras. Such a transformation satisfies all the above defined criteria, with a fast rate of transform for any images. Using the proposed transformation with proper settings we could significantly increase the accuracy of single-cell phenotyping, both in case of cell culture and tissue-based microscopy images. The range of potential applications of the proposed method goes beyond microscopy, as we present improved results on the iWildCam 2020 dataset containing images of wild animals.

bioinformatics↗

bioRxiv2022

Probe set selection for targeted spatial transcriptomics

Targeted spatial transcriptomics methods capture the topology of cell types and states in tissues at single cell- and subcellular resolution by measuring the expression of a predefined set of genes. The selection of an optimal set of probed genes is crucial for capturing and interpreting the spatial signals present in a tissue. However, current selections often rely on marker genes, precluding them from detecting continuous spatial signals or novel states. We present Spapros, an end-to-end probe set selection pipeline that optimizes both probe set specificity for cell type identification and within-cell-type expression variation to resolve spatially distinct populations while taking into account prior knowledge, as well as probe design and expression constraints. To facilitate data analysis and interpretation, Spapros also provides rules for cell type identification. We evaluated Spapros by selecting probes on 6 different data sets and built an evaluation pipeline with 12 quality metrics to find that Spapros outperforms other selection approaches in both cell type recovery and recovering expression variation beyond cell types. Furthermore, we used Spapros to design a SCRINSHOT experiment of adult lung tissue to demonstrate how probes selected with Spapros identify cell types of interest and detect spatial variation even within cell types. Spapros enables optimal probe set selection, probe set evaluation, and probe design, as a freely available Python package.

bioinformatics↗

bioRxiv2022

Learning representations for image-based profiling of perturbations

Measuring the phenotypic effect of treatments on cells through imaging assays is an efficient and powerful way of studying cell biology, and requires computational methods for transforming images into quantitative data that highlight phenotypic outcomes. Here, we present an optimized strategy for learning representations of treatment effects from high-throughput imaging data, which follows a causal framework for interpreting results and guiding performance improvements. We use weakly supervised learning (WSL) for modeling associations between images and treatments, and show that it encodes both confounding factors and phenotypic features in the learned representation. To facilitate their separation, we constructed a large training dataset with Cell Painting images from five different studies to maximize experimental diversity, following insights from our causal analysis. Training a WSL model with this dataset successfully improves downstream performance, and produces a reusable convolutional network for image-based profiling, which we call Cell Painting CNN-1. We conducted a comprehensive evaluation of our strategy on three publicly available Cell Painting datasets, discovering that representations obtained by the Cell Painting CNN-1 can improve performance in downstream analysis for biological matching up to 30% with respect to classical features, while also being more computationally efficient.

bioinformatics↗

bioRxiv2022

CometAnalyser: a user-friendly, open-source deep-learning microscopy tool for quantitative comet assay analysis.

Comet assay provides an easy solution to estimate DNA damage in single cells through microscopy assessment. It is widely used in the analysis of genotoxic damages induced by radiotherapy or chemotherapeutic agents. DNA damage is quantified at the single-cell level by computing the displacement between the genetic material within the nucleus, typically called "comet head", and the genetic material in the surrounding part of the cell, considered as the "comet tail". Today, the number of works based on Comet Assay analyses is really impressive. In this work, besides revising the solutions available to obtain reproducible and reliable quantitative data, we developed an easy-to-use tool named CometAnalyser. It is designed for the analysis of both fluorescent and silver-stained wide-field microscopy images and allows to automatically segment and classify the comets, besides extracting Tail Moment and several other intensity/morphological features for performing statistical analysis. CometAnalyser is an open-source deep-learning tool. It works with Windows, Macintosh, and UNIX-based systems. Source code, standalone versions, user manual, sample images, video tutorial and further documentation are freely available at: https://sourceforge.net/p/cometanalyser. HIGHLIGHTSO_LIComet assay provides an easy solution to estimate DNA damage in single cells. C_LIO_LIToday, an impressive number of works are based on Comet Assay analyses, especially in the field of cancer research. C_LIO_LIComet assay was originally performed as a qualitative analysis. C_LIO_LINone of the free tools today available work on both fluorescent- and silver-stained images. C_LIO_LIWe developed CometAnalyser, an open-source deep-learning tool designed for easy segmentation and classification of comets in fluorescent- and silver-stained images. C_LI

bioengineering↗

bioRxiv2021

Fully Automatic Cell Segmentation with Fourier Descriptors

Cell segmentation is a fundamental problem in biology for which convolutional neural networks yield the best results nowadays. In this paper, we present FourierDist, a network, which is a modification of the popular StarDist and SplineDist architectures. While StarDist and SplineDist describe an object by the lengths of equiangular rays and control points respectively, our network utilizes Fourier descriptors, predicting a coefficient vector for every pixel on the image, which implicitly define the resulting segmentation. We evaluate our model on three different datasets, and show that Fourier descriptors can achieve a high level of accuracy with a small number of coefficients. FourierDist is also capable of accurately segmenting objects that are not star-shaped, a case where StarDist performs suboptimally according to our experiments.

bioinformatics↗

bioRxiv2021

Proteome-wide landscape of solubility limits in a bacterial cell

Proteins are prone to aggregate when they are expressed above their solubility limits, a phenomenon termed supersaturation. Aggregation may occur as proteins emerge from the ribosome or after they fold and accumulate in the cell, but the relative importance of these two routes remain poorly known. Here, we systematically probed the solubility limits of each Escherichia coli protein upon overexpression using an image-based screen coupled with machine learning. The analysis suggests that competition between folding and aggregation from the unfolded state governs the two aggregation routes. Remarkably, the majority (70%) of insoluble proteins have low supersaturation risks in their unfolded states and rather aggregate after folding. Furthermore, a substantial fraction ([~]35%) of the proteome remain soluble at concentrations much higher than those found naturally, indicating a large margin of safety to tolerate gene expression changes. We show that high disorder content and low surface stickiness are major determinants of high solubility and are favored in abundant bacterial proteins. Overall, our proteome-wide study provides empirical insights into the molecular determinants of protein aggregation routes in a bacterial cell.

molecular biology↗

bioRxiv2021

Multiparametric assessment of cellular lipid metabolism in hypercholesterolemia

Systematic insight into cellular dysfunctions can improve understanding of disease etiology, risk assessment and patient stratification. We present a multiparametric high-content imaging platform enabling quantification of low-density lipoprotein (LDL) uptake and lipid storage in cytoplasmic droplets of primary leukocyte subpopulations. We validated this platform with samples from 65 individuals with variable blood LDL-cholesterol (LDL-c) levels, including familial hypercholesterolemia (FH) and non-FH subjects. We integrated lipid storage data into a novel readout, lipid mobilization, measuring the efficiency with which cells deplete lipid reservoirs. Lipid mobilization correlated positively with LDL uptake and negatively with hypercholesterolemia and age, improving differentiation of individuals with normal and elevated LDL-c. Moreover, combination of cell-based readouts with a polygenic risk score for LDL-c explained hypercholesterolemia better than the genetic risk score alone. This platform provides functional insights into cellular lipid trafficking from a few mls of blood and is applicable to dissect metabolic disorders, such as hypercholesterolemia. MotivationWe have limited information on how cellular lipid uptake and processing differ between individuals and influence the development of metabolic diseases, such as hypercholesterolemia. Available assays are labor intensive, require skilled personnel and are difficult to scale to higher throughput, making it challenging to obtain systematic functional cell-based data from individuals. To overcome this problem, we established a scalable automated analysis pipeline enabling reliable quantification of multiple cellular readouts, including lipid uptake, storage and mobilization, from different white blood cell populations. This approach provides new personalized insights into the cellular basis of hypercholesterolemia and obesity. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=200 SRC="FIGDIR/small/440471v4_ufig1.gif" ALT="Figure 1"> View larger version (84K): org.highwire.dtl.DTLVardef@9222c1org.highwire.dtl.DTLVardef@27e289org.highwire.dtl.DTLVardef@89ba82org.highwire.dtl.DTLVardef@33c9cd_HPS_FORMAT_FIGEXP M_FIG C_FIG

cell biology↗

bioRxiv2021

Computationally prioritized drugs inhibit SARS-CoV-2 infection and syncytia formation

New affordable therapeutic protocols for COVID-19 are urgently needed despite the increasing number of effective vaccines and monoclonal antibodies. To this end, there is increasing attention towards computational methods for drug repositioning and de novo drug design. Here, we systematically integrated multiple data-driven computational approaches to perform virtual screening and prioritize candidate drugs for the treatment of COVID-19. From the set of prioritized drugs, we selected a subset of representative candidates to test in human cells. Two compounds, 7-hydroxystaurosporine and bafetinib, showed synergistic antiviral effects in our in vitro experiments, and strongly inhibited viral-induced syncytia formation. Moreover, since existing drug repositioning methods provide limited usable information for de novo drug design, we extracted and prioritized the chemical substructures of the identified drugs, providing a chemical vocabulary that may help to design new effective drugs.

pharmacology and toxicology↗

bioRxiv2021

Candida albicans enhances the progression of oral squamous cell cancrinoma in vitro and in vivo

Oral squamous cell carcinoma (OSCC) is a serious health issue worldwide. OSCC is highly associated with oral candidiasis, although it is unclear whether the fungus promotes the genesis and progression of OSCC or cancer facilitates the growth of the fungus. Therefore, we investigated whether Candida could directly influence OSCC development and progression. Our in vitro results suggest that the presence of live C. albicans, but not C. parapsilosis, enhances the progression of OSCC by stimulating the production of matrix metalloproteinases, oncometabolites, pro-tumor signaling routes, and overexpression of prognostic marker genes associated with metastatic events. We also found that oral candidiasis triggered by C. albicans enhanced the progression of OSCC in vivo through the induction of inflammation and overexpression of metastatic genes and markers of epithelial-mesenchymal transition. Taken together, these results suggest that C. albicans actively participates in the complex process of OSCC progression.

microbiology↗

bioRxiv2021

AI-driven Deep Visual Proteomics defines cell identity and heterogeneity

The systems-wide analysis of biomolecules in time and space is key to our understanding of cellular function and heterogeneity in health and disease1. Remarkable technological progress in microscopy and multi-omics technologies enable increasingly data-rich descriptions of tissue heterogeneity2,3,4,5. Single cell sequencing, in particular, now routinely allows the mapping of cell types and states uncovering tremendous complexity6. Yet, an unaddressed challenge is the development of a method that would directly connect the visual dimension with the molecular phenotype and in particular with the unbiased characterization of proteomes, a close proxy for cellular function. Here we introduce Deep Visual Proteomics (DVP), which combines advances in artificial intelligence (AI)-driven image analysis of cellular phenotypes with automated single cell laser microdissection and ultra-high sensitivity mass spectrometry7. DVP links protein abundance to complex cellular or subcellular phenotypes while preserving spatial context. Individually excising nuclei from cell culture, we classified distinct cell states with proteomic profiles defined by known and novel proteins. AI also discovered rare cells with distinct morphology, whose potential function was revealed by proteomics. Applied to archival tissue of salivary gland carcinoma, our generic workflow characterized proteomic differences between normal-appearing and adjacent cancer cells, without admixture of background from unrelated cells or extracellular matrix. In melanoma, DVP revealed immune system and DNA replication related prognostic markers that appeared only in specific tumor regions. Thus, DVP provides unprecedented molecular insights into cell and disease biology while retaining spatial information.

systems biology↗

bioRxiv2021

Oral epithelial cells distinguish between Candida species with high or low pathogenic potential through miRNA regulation

Oral epithelial cells monitor microbiome composition and initiate immune response upon dysbiosis, as in case of Candida imbalances. Comparison of healthy oral epithelial cell responses revealed that the inability of C. parapsilosis to induce a robust antifungal response was due to activation of various inflammation-independent pathways, while C. albicans robustly activated inflammation cascades. Regarding posttranscriptional regulation, several miRNAs were altered by both species. For C. parapsilosis, the applied dose directly correlated with changes in transcriptomic responses. Carbohydrate metabolism, hypoxia- and cardiovascular development-related responses dominate after C. parapsilosis stimulus, whereas C. albicans altered inflammatory responses. Subsequent analyses of HIF1- and HSC-activation pathways predicted target genes through which miRNA-dependent regulation of yeast-specific functions may occur, supporting the observed responses. Thus, C. parapsilosis is recognized as a commensal at low doses by the oral epithelium; however, increased fungal burden activates different pathways, some of which overlap with inflammatory processes induced by C. albicans. Impact statementAltered miRNA regulation discriminates between C. albicans and C. parapsilosis in human oral epithelial cells

microbiology↗

bioRxiv2020

Regression plane concept: analysing continuous cellular processes with machine learning

Biological processes are inherently continuous, and the chance of phenotypic discovery is significantly restricted by discretising them. Using multi-parametric active regression we introduce a novel concept to describe and explore biological data in a continuous manner. We have implemented Regression Plane (RP), the first user-friendly discovery tool enabling class-free phenotypic supervised machine learning.

bioinformatics↗

bioRxiv2020

Neuropilin-1 is a host factor for SARS-CoV-2 infection

SARS-CoV-2 is the causative agent of COVID-19, a coronavirus disease that has infected more than 6.6 million people and caused over 390,000 deaths worldwide1,2. The Spike (S) protein of the virus forms projections on the virion surface responsible for host cell attachment and penetration. This viral glycoprotein is synthesized as a precursor in infected cells and, to be active, must be cleaved to two associated polypeptides: S1 and S2(3,4). For SARS-CoV-2 the cleavage is catalysed by furin, a host cell protease, which cleaves the S protein precursor at a specific sequence motif that generates a polybasic Arg-Arg-Ala-Arg (RRAR) C-terminal sequence on S1. This sequence motif conforms to the C-end rule (CendR), which means that the C-terminal sequence may allow the protein to associate with cell surface neuropilin-1 (NRP1) and neuropilin-2 (NRP2) receptors5. Here we demonstrate using immunoprecipitation, site-specific mutagenesis, structural modelling, and antibody blockade that, in addition to engaging the known receptor ACE2, S1 can bind to NRP1 through the canonical CendR mechanism. This interaction enhances infection by SARS-CoV-2 in cell culture. NRP1 thus serves as a host factor for SARS-CoV-2 infection, and provides a therapeutic target for COVID-19.

cell biology↗

bioRxiv2020

Automatic deep learning driven label-free image guided patch clamp system for human and rodent in vitro slice physiology

Patch clamp recording of neurons is a labor-intensive and time-consuming procedure. We have developed a tool that fully automatically performs electrophysiological recordings in label-free tissue slices. The automation covers the detection of cells in label-free images, calibration of the micropipette movement, approach to the cell with the pipette, formation of the whole-cell configuration, and recording. The cell detection is based on deep learning. The model was trained on a new image database of neurons in unlabeled brain tissue slices. The pipette tip detection and approaching phase use image analysis techniques for precise movements. High-quality measurements were performed on hundreds of human and rodent neurons. We also demonstrate that further molecular and anatomical analysis can be performed on the recorded cells. The software has a diary module that automatically logs patch clamp events. Our tool can multiply the number of daily measurements to help brain research. ONE SENTENCE SUMMARYNovel deep learning and image analysis algorithms for automated patch clamp systems to reliably measure neurons in human and rodent brain slices.

neuroscience↗

bioRxiv2020

OpSeF IV: Open source Python framework for segmentation of biomedical images

Various pre-trained deep learning models for the segmentation of bioimages have been made available as developer-to-end-user solutions. They usually require neither knowledge of machine learning nor coding skills, are optimized for ease of use, and deployability on laptops. However, testing these tools individually is tedious and success is uncertain. Here, we present the Open Segmentation Framework (OpSeF), a Python framework for deep learning-based instance segmentation. OpSeF aims at facilitating the collaboration of biomedical users with experienced image analysts. It builds on the analysts knowledge in Python, machine learning, and workflow design to solve complex analysis tasks at any scale in a reproducible, well-documented way. OpSeF defines standard inputs and outputs, thereby facilitating modular workflow design and interoperability with other software. Users play an important role in problem definition, quality control, and manual refinement of results. All analyst tasks are optimized for deployment on Linux workstations or GPU clusters, all user tasks may be performed on any laptop in ImageJ. OpSeF semi-automates preprocessing, convolutional neural network (CNN)-based segmentation in 2D or 3D, and post-processing. It facilitates benchmarking of multiple models in parallel. OpSeF streamlines the optimization of parameters for pre- and post-processing such, that an available model may frequently be used without retraining. Even if sufficiently good results are not achievable with this approach, intermediate results can inform the analysts in the selection of the most promising CNN-architecture in which the biomedical user might invest the effort of manually labeling training data. We provide Jupyter notebooks that document sample workflows based on various image collections. Analysts may find these notebooks useful to illustrate common segmentation challenges, as they prepare the advanced user for gradually taking over some of their tasks and completing their projects independently. The notebooks may also be used to explore the analysis options available within OpSeF in an interactive way and to document and share final workflows. Currently, three mechanistically distinct CNN-based segmentation methods, the U-Net implementation used in Cellprofiler 3.0, StarDist, and Cellpose have been integrated within OpSeF. The addition of new networks requires little, the addition of new models requires no coding skills. Thus, OpSeF might soon become both an interactive model repository, in which pre-trained models might be shared, evaluated, and reused with ease.

bioinformatics↗

bioRxiv2020

AnnotatorJ: an ImageJ plugin to ease hand-annotation of cellular compartments

AnnotatorJ combines single-cell identification with deep learning and manual annotation. Cellular analysis quality depends on accurate and reliable detection and segmentation of cells so that the subsequent steps of analyses e.g. expression measurements may be carried out precisely and without bias. Deep learning has recently become a popular way of segmenting cells, performing unimaginably better than conventional methods. However, such deep learning applications may be trained on a large amount of annotated data to be able to match the highest expectations. High-quality annotations are unfortunately expensive as they require field experts to create them, and often cannot be shared outside the lab due to medical regulations. We propose AnnotatorJ, an ImageJ plugin for the semi-automatic annotation of cells (or generally, objects of interest) on (not only) microscopy images in 2D that helps find the true contour of individual objects by applying U-Net-based pre-segmentation. The manual labour of hand-annotating cells can be significantly accelerated by using our tool. Thus, it enables users to create such datasets that could potentially increase the accuracy of state-of-the-art solutions, deep learning or otherwise, when used as training data.

bioinformatics↗

bioRxiv2019

Test-time augmentation for deep learning-based cell segmentation on microscopy images

Recent advancements in deep learning have revolutionized the way microscopy images of cells are processed. Deep learning network architectures have a large number of parameters, thus, in order to reach high accuracy, they require massive amount of annotated data. A common way of improving accuracy builds on the artificial increase of the training set by using different augmentation techniques. A less common way relies on test-time augmentation (TTA) which yields transformed versions of the image for prediction and the results are merged. In this paper we describe incorporating the test-time argumentation prediction method into two major segmentation approaches used in the single-cell analysis of microscopy images, namely semantic segmentation using U-Net and instance segmentation using Mask R-CNN models. Our findings show that even using only simple test-time augmentations, such as rotation or flipping and proper merging methods, will result in significant improvement of prediction accuracy. We utilized images of tissue and cell cultures from the Data Science Bowl (DSB) 2018 nuclei segmentation competition and other sources. Additionally, boosting the highest-scoring method of the DSB with TTA, we could further improve and our method has reached an ever-best score at the DSB.

bioinformatics↗

bioRxiv2019

Structure of the native supercoiled flagellar hook as a universal joint

Bacteria swim in viscous liquid environments by using the flagellum1-3. The flagellum is composed of about 30 different proteins and can be roughly divided into three parts: the basal body, the hook and the filament. The basal body acts as a rotary motor powered by ion motive force across the cytoplasmic membrane as well as a protein export apparatus to construct the axial structure of the flagellum. The filament is as a helical propeller, and it is a supercoiled form of a helical tubular assembly consisting of a few tens of thousands of flagellin molecules4. The hook is a relatively short axial segment working as a universal joint connecting the basal body and the filament for smooth transmission of motor torque to the filament5,6. The structure of hook has been studied by combining X-ray crystal structure of a core fragment of hook protein FlgE and electron cryomicroscopy (cryoEM) helical image analysis of the polyhook in the straight form and has given a deep insight into the universal joint mechanism7. However, the supercoiled structure of the hook was an approximate model based on the atomic model of the straight hook without its inner core domain7 and EM observations of supercoiled polyhook by freeze-dry and Pt/Pd shadow cast8. Here we report the native supercoiled hook structure at 3.1 [A] resolution by cryoEM single particle image analysis of the polyhook. The atomic model built on the three-dimensional (3D) density map show the actual changes in subunit conformation and intersubunit interactions upon compression and extension of the 11 protofilaments that occur during their smoke ring-like rotation and allow the hook to function as a dynamic molecular universal joint with high bending flexibility and twisting rigidity.

biophysics↗