2026-10-14T01:32:12Z https://www.scidb.cn/oai

10.57760/sciencedb.43079 2026-07-17T11:30:03Z

2026-07-17 Microscopy dataset and image analysis code from timelapse imaging of individual steps in the lytic bacteriophage life cycle in single bacterial cells doi:10.57760/sciencedb.43079 en IntroductionThis dataset is a companion to the work presented in the preprint “Single-cell imaging of the lytic phage life cycle in bacteria” [1]. If you use this dataset in your work, please cite our preprint. A full description of the scientific objectives, methods, results and analysis can be found in the preprint. Brief descriptions are included here, along with details of the dataset.This dataset contains timelapse microscopy images of Escherichia coli (E. coli) cells undergoing lytic phage infection by T7 phage and a modified version of a T7 phage. Two fluorescent labels are used: (1) to visualise the phage and its genome directly, SYTOX Orange (a DNA binding dye) is used, and (2) to visualise gene expression from the phage genome, we modified the wild type (WT) T7 to create a version of T7 with a yellow fluorescent protein (YFP) reporter of gene 10 (the capsid gene). The modified phage is denoted as TY009 or T7*. The dataset is split into two branches. The larger branch (repository_inj_ga_tx) contains raw data and analysis for eight experiments, and covers the adsorption, genome injection, growth arrest and phage gene expression. Experiments 1, 2, 3, 6 and 8 use the modified phage. Experiments 4, 5 and 7 use the wild type phage. The other branch (repository_perforation_lysis) contains data from one experiment using the wild type phage, where lysis events were captured with very high time resolution (~10 ms imaging interval). Files titled README.md and 00_read_me.txt can be found throughout the repository, providing descriptions of the repository layout and explanations of the analysis. The scientific value and reusability of this data fall into two broad categories. The first is in the raw and processed data itself. The dataset contains a large volume of fluorescence and phase contrast microscopy images depicting lytic bacteriophage infection. These images can be used by researchers who wish to conduct their own analyses on the images, or to benchmark or supplement experimental microscopy data from new studies. The fluorescent labelling strategies used allow adsorption, genome injection and phage gene expression to be studied in detail, and the high-frequency imaging used to capture phage-induced cell lysis provides data on the rapid breakdown of the cell envelope. The second category of scientific value is the data processing and analysis approach. The code and descriptions found in this dataset, along with information in the preprint, provide a framework for extracting insights from data of this kind. This includes preprocessing steps, the design of exclusion criteria, analysis methods and modelling. Timelapse micrographs of bacteria are a rich data source, and there are many questions that can be asked of this data which have not been answered in our preprint. We hope that this dataset can support future work in microscopy, image analysis and modelling, and that it becomes a useful resource for the phage science community. Data integrityChecksums have been provided to allow users to verify the integrity of their download. The checksums are generated using a SHA256 hash of each file. The checksums can be verified using the script 00_verify_checksums.py. To verify your download integrity (assuming you have Python installed), simply run the command “python 00_verify_checksums.py” from a terminal or Windows command prompt from the relevant folder. For example, to check your download integrity of your repository_inj_ga_tx folder, navigate to that folder in the terminal or command prompt window, then run “python 00_verify_checksums.py”. Note that the command may be slightly different depending on the operating system (Windows may require “py 00_verify_checksums.py”, while macOS may require “python3 00_verify_checksums.py”).When the repository is first downloaded, the analysis folders will be compressed into zip files. If the top-level script 00_verify_checksums.py runs and verifies all files correctly, then you can be confident the download has preserved the original integrity of the data. If anything does not match, there is some level of corruption. Lower level checksums are provided within each zipped analysis folder to allow the source of the corruption to be pinpointed.Experimental method and raw dataFull method details are described in our preprint. Briefly, E. coli cells from an overnight culture were loaded into a microfluidic “mother machine” device and grown in LB media with pluronic at 37 oC until steady state growth was achieved. Then the media was switched into a nutritionally equivalent media containing phage. Phage would diffuse into the cell-containing microfluidic channels (termed trenches in our preprint), where cells would become infected and then lyse. Throughout this process, images were captured with a Nikon Ti-2 Eclipse inverted epifluorescence microscope. Image metadata, describing the optical setup and other imaging parameters, can be found under repository_inj_ga_tx/raw_data/experiment_metadata for each of the eight experiments in the larger branch, and can be found under repository_perforation_lysis/trench_<trench_number>_images for the branch focusing on lysis. The raw image data is stored in compressed zip files, as there are a large number of images. For repository_inj_ga_tx, Python scripts in the corresponding analysis folders are provided which can unzip these correctly. Please note that unzipping these files will extract of the order of 10^4 to 10^6 individual image files. We highly recommend unsyncing the raw data folder from any cloud storage services, as syncing that many files is highly likely to stall or crash your cloud syncing service. Images for several of the experiments are stored as Zarr arrays (version 2). Zarr [2] is a useful data structure for microscopy image data, allowing images to be stored to disk as multidimensional arrays, loaded and indexed as though they were Numpy [3] arrays. Zarr is language agnostic, free and open-source, meaning it has broad compatibility. Other images are stored as zipped PNG files. The spatial resolution of the images for experiments 1, 2, 3, 4 and the lysis experiment is 0.065 μm/pixel. The spatial resolution of the images for experiments 5, 6, 7 and 8 is 0.1041 μm/pixel. The temporal resolution of the images for experiments 1 to 8 is 30 s. The temporal resolution of the images for the lysis experiment is 10 ms.For each experiment, exclusion criteria were applied to infection events for technical and scientific reasons, to ensure that the specific event could be analysed without interference from artefacts and that the data was relevant to the question being asked. These criteria are explained separately in our preprint for the genome injection (supplementary note 11), growth arrest and capsid production (supplementary note 14) and perforation and lysis (supplementary note 17). Data processing and analysisThe data processing steps vary depending on the scientific objectives. As with the experimental details, further reasoning and explanation can be found in our preprint for each analysis. In general, the analysis for each experiment is split into Python scripts which should be run in numerical order. Output CSV and JSON files from these steps are saved, with the prefix number corresponding to the script which was used to generate the data. These outputs are often the input to the next script, forming a set of linear-modular pipelines. A number of Python packages were used to assist with the analysis [3, 4, 5]. All the analysis code is contained in this dataset alongside the data, and the companion Github code repositories also contain all the analysis code, but without the data [6, 7].For the adsorption and genome injection analysis (repository_inj_ga_tx/analysis_inj/), cell segmentation is not used. However, an automated method to assist in finding the locations of stained phage particles called Trackpy (v0.6.1) [8] is used. For most experiments, only the output (tables of temporal and spatial coordinates for different particles, along with metadata describing the Trackpy parameters used) is given as the code requires a stack of PNG images, which would require replicating raw data to run out-of-the-box. However, the code is given for experiment 7 (repository_inj_ga_tx/analysis_inj/analysis_inj_exp7/00_particle_tracking.py), which can be used as a template to apply Trackpy to the other experiments. Following this, the coordinates are adjusted and extended by a variety of methods, all involving some degree of manual image inspection, to find the coordinates of the phage and the corresponding location of phage adsorption on the cell surface both before and after the phage is visible. As this step involved inspection, only the results (which can then be used to run the rest of the analysis deterministically) are given. Using the obtained coordinates, the signal intensity is found by finding the total or mean pixel intensity of the brightest fraction of pixels in a box around the phage location, and subtracting the corresponding total or mean from an offset array which serves as a control, effectively performing a background subtraction. This signal intensity is then cleaned. The cleaned signal is then used to calculate the genome injection duration, by finding when the signal intensity falls below a specified threshold. Further details of the data processing are found in a 00_read_me.txt file in the injection analysis folder for each experiment.For the growth arrest and capsid expression analysis, data processing begins with cell segmentation. This process generates accurate cell masks which are used as a basis of quantifying bacterial physiology. The segmentation process is described in detail in the methods and supplementary note 4 of our preprint. Briefly, synthetic training data is generated using SyMBac [9], and model training and segmentation use Omnipose [10]. Only the output masks are given here to increase the accessibility of the dataset, as the training and segmentation are computationally intensive and require GPU architecture to run in a reasonable timeframe. The output masks have a different array shape to the raw images. This is because only a truncated version of the raw phase contrast images were segmented. This was done because the bright phase contrast “halo” near the trench exit in the images caused segmentation issues, so the halo was removed to achieve good segmentation quality in the rest of the image. To obtain time series cell properties of the cells using the provided masks, the scikit-image [5] regionprops function can be used. Once cell masks were obtained, cells were lineage tracked [11] to find the full lineage of the first cell to lyse in each trench. Details of the lineage tracking are found in supplementary note 5 of our preprint. The outputs of the lineage tracking are stored in, for example, repository_inj_ga_tx/analysis_ga_tx/analysis_ga_tx_exp1/time_series_data. These data are processed as described in repository_inj_ga_tx/analysis_ga_tx/00_read_me.txt, in order to obtain the growth arrest and capsid production start time points. For the perforation and lysis analysis, Fiji [12] was used to obtain static masks and extract time series intensity data for each cell. The description in repository_perforation_lysis/analysis/00_read_me.txt explains the data processing steps performed by each script.  LicensingThe code used for this project, contained both in the main dataset on ScienceDB and separately on Github, is licensed under an MIT license (see LICENSE-CODE.txt). The code was developed to operate on the dataset stored on ScienceDB. The image data, metadata, tables and other data stored in the dataset on ScienceDB are licensed separately under CC-BY 4.0 (see LICENSE-DATA.txt). Therefore this work consists of two components under separate licenses: Data (images, metadata, tables and other data):Copyright 2026 Charlie Wedd, Ruizhe Li, Temur Yunusov, Aaron Smith, Georgeos Hardo, Michael Hunter, Aparna Kaaraal Mohan, Racha Majed, Diana Fusco, Somenath Bakshi.Licensed under CC BY 4.0 - see LICENSE-DATA.txt, https://creativecommons.org/licenses/by/4.0/ Code (analysis scripts):Copyright 2026 Charlie Wedd, Ruizhe Li, Temur Yunusov, Aaron Smith, Georgeos Hardo, Michael Hunter, Aparna Kaaraal Mohan, Racha Majed, Diana Fusco, Somenath Bakshi.Licensed under MIT - see LICENSE-CODE.txt Unless otherwise indicated, documentation describing the dataset, imaging methodology, and experimental design is covered under the CC BY 4.0 data license; documentation describing code usage, installation, and software structure is covered under the MIT code license. References[1] Wedd, C. et al. Single-cell imaging of the lytic phage life cycle in bacteria (2025). https://www.biorxiv.org/content/10.1101/2024.04.11.588870v4 [2] Zarr-Python. https://zarr.readthedocs.io/en/stable/[3] Harris, C.R., Millman, K.J., van der Walt, S.J. et al. Array programming with NumPy. Nature 585, 357–362 (2020). https://doi.org/10.1038/s41586-020-2649-2 [4] Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nature Methods 17, 261–272 (2020). URL https://www.nature.com/articles/s41592-019-0686-2 [5] Walt, S. v. d. et al. scikit-image: image processing in Python. PeerJ 2, e453 (2014). https://peerj.com/articles/453 [6] phage_injection_growth_arrest_transcription_v1.0.0. https://zenodo.org/records/21385525 [7] phage_perforation_lysis_v1.0.0. https://zenodo.org/records/21385531[8] Trackpy v0.6.1. https://zenodo.org/records/7670439[9] Hardo, G., Noka, M. & Bakshi, S. Synthetic Micrographs of Bacteria (SyMBac) allows accurate segmentation of bacterial cells using deep neural networks. BMC Biology 20, 263 (2022). https://doi.org/10.1186/s12915-022-01453-6 [10] Cutler, K. J. et al. Omnipose: a high-precision morphology-independent solution for bacterial cell segmentation. Nature Methods 19, 1438–1448 (2022). https://www.nature.com/articles/s41592-022-01639-4 [11] MMLineageTracking v1.0.0. https://zenodo.org/records/20646295 [12] Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676–682 (2012). https://www.nature.com/articles/nmeth.2019 Phage; single-cell; microscopy; life cycle; adsorption; genome injection; gene expression; lysis Charlie Wedd Ruizhe Li Temur Yunusov Aaron Smith Georgeos Hardo Michael Hunter Aparna Kaaraal Mohan Racha Majed Diana Fusco Somenath Bakshi EMBARGO https://creativecommons.org/licenses/by/4.0/ dataset http://www.doi.org/10.1101/2024.04.11.588870 Science Data Bank