Using Dimensionality Reduction to Visualize Phenotypic Changes in High-Throughput Microscopy.

Area of Science:

• Cell biology
• Bioinformatics
• Computational imaging

Background:

• High-throughput microscopy generates vast amounts of cell phenotype data.
• Analyzing complex, high-dimensional cell phenotype data, including protein localization, remains challenging.
• Unbiased methods are needed to visualize and detect subtle changes in cellular phenotypes.

Purpose of the Study:

• To develop and present a computational pipeline for visualizing changes in cellular phenotypes.
• To demonstrate the utility of reduced dimensionality representations for analyzing image feature spaces.
• To identify alterations in protein subcellular localization using deep learning-based feature extraction.

Main Methods:

• Utilized deep learning to extract features from microscopy images.
• Developed a freely available analysis pipeline for phenotype visualization.
• Applied dimensionality reduction techniques to visualize feature spaces.
• Analyzed changes in protein localization in a yeast model system treated with hydroxyurea.

Main Results:

• Successfully visualized high-dimensional cell phenotype data in reduced dimensions.
• The pipeline effectively identified changes in subcellular protein localization.
• Demonstrated the application of the method on the yeast GFP collection treated with hydroxyurea.

Conclusions:

• Reduced dimensionality representations offer a powerful approach for visualizing cellular phenotype changes.
• The developed pipeline provides a valuable tool for analyzing large-scale microscopy data.
• This method facilitates the systematic identification of protein localization alterations in response to stimuli.