A rapid machine learning-based method to analyze cell dimensions in fission yeast microscope images
Start Date
29-4-2022 3:45 PM
Location
Alter Hall Poster Session 2 - 2nd floor
Abstract
The dimensions of cells are linked to its cell cycle progression and environmental conditions. Cells in different environmental conditions can display various characteristics that show us whether that environment is ideal or not. Fission yeast (Schizosaccharomyces pombe) is a rod-shaped ascomycete fungus where cell cycle progression can be tracked by the changes in cell length, area and width. Although microscopy analysis of fission yeast allows for quick determination of life cycle progression under different genetic and environmental conditions, the task of measuring and subsequently analyzing size metrics statistically places considerable burdens on time and effort. Moreover, outlining cells either manually or using plugin packages for commonly used image analysis software results in data affected by user bias or processing artifacts, respectively. In this work, we report a machine learning-based methodology that measures cell dimensions in fission yeast in an unbiased, automated manner and streamlines workflow from image acquisition to statistical analysis. Using this approach, we were able to efficiently train and modify the image-processing algorithm to our experimental needs. We processed image data for five different experiments in a time-efficient manner without the need of extensive computing power. To these processes, we coupled a downstream statistical routine that is simple to implement and interpret. Our findings suggest that with this method fission yeast researchers will be able to readily assess cell size dynamics under different conditions in ways that clearly highlight underlying genetic or environmental alterations. The increased efficiency in processing time will also enable examination of large sample sizes that reveal unique cell phenotypes with relevant biological, not just simply statistical significance.
A rapid machine learning-based method to analyze cell dimensions in fission yeast microscope images
Alter Hall Poster Session 2 - 2nd floor
The dimensions of cells are linked to its cell cycle progression and environmental conditions. Cells in different environmental conditions can display various characteristics that show us whether that environment is ideal or not. Fission yeast (Schizosaccharomyces pombe) is a rod-shaped ascomycete fungus where cell cycle progression can be tracked by the changes in cell length, area and width. Although microscopy analysis of fission yeast allows for quick determination of life cycle progression under different genetic and environmental conditions, the task of measuring and subsequently analyzing size metrics statistically places considerable burdens on time and effort. Moreover, outlining cells either manually or using plugin packages for commonly used image analysis software results in data affected by user bias or processing artifacts, respectively. In this work, we report a machine learning-based methodology that measures cell dimensions in fission yeast in an unbiased, automated manner and streamlines workflow from image acquisition to statistical analysis. Using this approach, we were able to efficiently train and modify the image-processing algorithm to our experimental needs. We processed image data for five different experiments in a time-efficient manner without the need of extensive computing power. To these processes, we coupled a downstream statistical routine that is simple to implement and interpret. Our findings suggest that with this method fission yeast researchers will be able to readily assess cell size dynamics under different conditions in ways that clearly highlight underlying genetic or environmental alterations. The increased efficiency in processing time will also enable examination of large sample sizes that reveal unique cell phenotypes with relevant biological, not just simply statistical significance.
