Time lapse microscopy images of Saccharomyces cerevisiae's full life cycle
Data files
May 08, 2025 version files 31.98 GB
-
OAM_230223.zip
7.70 GB
-
OAM_SN_042723.zip
4.47 GB
-
OAM_TK_221102.zip
425.43 MB
-
Pos13_1_B.zip
81.44 MB
-
README.md
5.63 KB
-
SN_230511_A.zip
9.08 GB
-
SN_230511_B.zip
9.07 GB
-
TK_230417.zip
1.14 GB
Abstract
The life cycles of biomedical and agriculturally relevant eukaryotic microorganisms involve complex transitions between proliferative and non-proliferative states, such as dormancy, mating, meiosis, and cell division. New drugs, pesticides, and vaccines can be created by targeting specific life cycle stages of parasites and pathogens. However, defining the structure of a microbial life cycle often relies on partial observations that are theoretically assembled in an ideal life cycle path. To create a more quantitative approach to studying complete eukaryotic life cycles, we generated a microfluidic assay to record the complete sexual life cycle of the model eukaryote Saccharomyces cerevisiae for up to three sexual life cycles. This data set contains TIF image files from the time-lapse imaging of the complete S. cerevisiae life cycle. We envision that it will be used to benchmark new single-cell processing algorithms and as a starting point for quantitatively characterizing other single-cell eukaryotic life cycles that remain incompletely described. The data can be processed using any image analysis software. We also developed the Python package yeastvision, which provides a user interface that allows free access to our image processing and single-cell tracking algorithms for complete life cycle analyses.
https://doi.org/10.5061/dryad.3bk3j9kw0
Description of the data and file structure
Defining the structure of a microbial life cycle often relies on partial observations that are theoretically assembled in an ideal life cycle path. To create a more quantitative and direct approach to studying complete eukaryotic life cycles, we generated a deep learning-driven imaging framework to track microorganisms across sexually reproducing generations.
Our approach combines microfluidic culturing, life cycle stage-specific segmentation of microscopy images using convolutional neural networks, and a novel cell tracking algorithm, FIEST, based on enhancing the overlap of single cell masks in consecutive images through deep learning video frame interpolation.
As proof of principle, we used this approach to quantitatively image and compare cell growth and cell cycle regulation in Saccharomyces cerevisiae’s sexual life cycle. We developed a fluorescent reporter system based on a fluorescently labeled Whi5 protein, the yeast analog of mammalian Rb, and a new High-Cdk1 activity sensor, LiCHI, designed to report during DNA replication, mitosis, meiotic homologous recombination, meiosis I, and meiosis II. This data set corresponds to the phase contrast and fluorescent images depicting the entire life cycle of S. cerevisiae.* *
Our computational framework and the images in this data set will contribute to improving the quantitative characterization of incompletely described single-cell eukaryotic life cycles. Importantly, this data set could be used to develop tracking and segmentation algorithms, and train deep learning models or fine tune large language models.
Files and variables
Folder Organization and Directory Layout
Images are organized hierarchically according to the experiments and their respective fields of view. A field of view (FOV) in this context is an area that was recorded by the camera on the surface of a microfluidic device where cells were cultured. The folders’ names correspond to the experiment name starting with the creators’ initials followed by the date of acquisition.
Each experiment folder contains subfolders for regions of interest (ROIs), defined as a subarea within a FOV. ROI subfolders are named ‘Pos’, short for the position corresponding to the FOV, followed by an underscore and a number, corresponding to the ROI within the FOV. For example, subfolders named ‘Pos13_1’ ,’ Pos13_2’, and ‘Pos13_3’ all correspond to the FOV number 13, and ROI 1, 2, and 3, respectively.
Image File Naming Convention
The image name is composed of the following parts: the prefix “img_”; a nine digit number such as “000000001” that indicates the time point (time point one in this case); a channel name such as “ _505_mNG_” that indicates the fluorescent channel used to record fluorophore (mNeonGreen fluorescent channel in this case); and finally the generic ending “000.tif”. All images are .tif files in either 8-bit or 16-bit depth.
Images ending in ‘Ph3_000.tif’ were acquired in the phase contrast channel; images ending in ‘mKOk_000.tif’ were acquired in the mKusabira-Orange kappa channel and depict the C-terminally tagged Whi5 protein; images ending in ‘mRuby3_000.tif’ were acquired in the mScarlet-I channel and also depict the C-terminally tagged Whi5 protein; images ending in ‘mNG_000.tif’ were acquired in the mNeonGreen channel and depict the new LiCHI sensor.
Example directory organization for ROI images collected at time-point 0, the first time-point in the series:
[Experiment Name Folder]
|- [Field of View_RO Folder]
|-|- img_000000000_505_mNG_000.tif,
|-|- img_000000000_555_mRuby3_000.tif and
|-|- img_000000000_Ph3_000.tif
Segmented ROI Toy Dataset
The toy data set folder “Pos13_1_B.zip” contains phase contrast image time series from an ROI with cells undergoing a full life cycle (sporulation, germination, mating, proliferation) and masks corresponding to the segmentation results from life cycle stage-specific cellpose models.
The mask names are composed of the phase contrast image name plus an ending that corresponds to the life cycle stage-specific segmentation model. The ending “_ART_masks.tif” corresponds to masks obtained with the model “ProSeg”; the ending “_MAT16_masks.tif” corresponds to masks obtained with the model “MatSeg”; the ending “_TET_masks.tif” corresponds to masks obtained with the model “SpoSeg”. Detailed information about the segmentation models is available here: https://www.biorxiv.org/content/10.1101/2024.04.25.591211v1 . This toy data set folder can be used to test the full life cycle tracking code using yeastvision or the source code script deposited here: https://github.com/MirandaLab/Life_Cycle_Tracking.
Code/software
This data set can be visualized using standard Python, MATLAB, or image processing applications. We recommend our python package: https://pypi.org/project/yeastvision/ to visualize, segment, and track single cells.
Access Information
Other publicly accessible locations of the data:
This data set was recorded using time-lapse microscopy and microfluidics devices (Y04C CellASIC plate with OniX controllers). Microfluidics experiments were performed on an automated Zeiss Axio Observer Z1 microscope controlled by ZEN pro software and with temperature control (Zeiss). Images were acquired at a 12 minute sampling rate using an 40X 1.3 NA oil Ph 3 M27 immersion objective. Image focus was controlled using Definite Focus 3.0. Images were recorded using an AxioCam 712 monochrome. An X-CITE XYLIS XT720S lamp (Excelitas Technologies) was used as a light source.
Fluorescent channel filter sets were tailored using dichroic mirrors and bandpass filters from Semrock. The mKOκ detection channel was designed using the excitation filter FF01-534/20-25, the dichroic FF552 Di02-25x36, and the emission filter FF01 563/9-25; the mRuby3 detection channel was designed using the excitation filter FF01-563/9-25, the dichroic FF573 Di01-25x36, and the emission filter FF01 598/25-25; and the mNG detection channel was designed using the excitation filter FF01-504/12-25, the dichroic FF518 Di01-25x36, and the emission filter FF01 530/11-25.
The original images were collected as uncompressed uint16 files. Images ending in ‘mKOk_000.tif’, ‘mRuby3_000.tif’, or ‘mNG_000.tif’ were turned into uint8 using MATLAB’s uint8(). For images ending in ‘Ph3_000.tif’, the following scaling in MATLAB was applied: the original uint16 was turned into a double using double(), scaled between 0 - 1 , and then multiplied by 255 to render the image grayscale values between 0 - 255, and then saved as uint8 using uint8(). Rescaling was done by dividing the subtraction of all pixel values minus the minimum value in the original image by the subtraction of the maximum value minus the minimum value in the original image. The resulting 0 - 1 scaled image was multiplied by 255 to achieve an 8-bit grayscale range of 0 - 255.