Self-supervised classification of subcellular morphometric phenotypes reveals extracellular matrix-specific...

Area of Science:

• Cell Biology
• Bioinformatics
• Machine Learning

Background:

• Cell morphology is significantly influenced by microenvironmental factors, particularly the extracellular matrix (ECM).
• Distinct cell morphologies on specific ECM suggest ECM-dependent signaling states, but analysis is challenging due to morphological complexity.

Purpose of the Study:

• To develop an automated quantitative analysis pipeline for classifying cell morphometric phenotypes.
• To leverage multi-channel fluorescence microscopy and deep learning for analyzing cell shape complexity.
• To establish a framework for classifying cell shapes into distinct, cell-type, and ECM-specific morphological signatures.

Main Methods:

• Development of a deep learning-based analysis pipeline named SE-RNN (residual neural network with squeeze-and-excite blocks).
• Utilized multi-channel fluorescence microscopy for robust molecular specificity and biological interpretation.
• Applied SE-RNN to classify distinct morphological signatures of fibroblasts and epithelial cells on different ECM substrates.

Main Results:

• Successfully demonstrated SE-RNN's capability in classifying distinct morphological signatures.
• Showcased cell-type and ECM-specific classification of cell shapes.
• Underscored that cell shape variations are non-random and contain significant biological information.

Conclusions:

• The developed SE-RNN pipeline provides a robust framework for automated quantitative analysis of cell morphology.
• Cellular responses to ECM are specific and can be classified using deep learning approaches.
• This method enables a deeper understanding of cell signaling states through morphological phenotyping.