A machine learning approach for single cell interphase cell cycle staging.

Area of Science:

• Cell Biology
• Biotechnology
• Computational Biology

Background:

• Nuclear architecture is crucial for cell function, with dynamic changes during the cell cycle.
• Altered nuclear morphology is a key indicator in diseases like cancer.
• Reliable automated tools for single-cell cycle staging from images are currently limited.

Purpose of the Study:

• To develop a supervised machine learning method for accurate interphase cell cycle staging using in situ fluorescence images.
• To enable bona fide classification of cell cycle phases at the individual cell level.
• To provide a robust tool for research and clinical applications.

Main Methods:

• Developed a Support Vector Machine (SVM) classifier utilizing normalized nuclear features from DAPI-stained nuclei.
• Employed fluorescent ubiquitination-based cell cycle indicator (Fucci) technology for molecular ground truth labeling.
• Trained and validated the model on over 3500 DAPI-stained nuclei across distinct cell types.

Main Results:

• Achieved an average F1-Score of 87.7% for cell cycle staging.
• Demonstrated high recall values exceeding 89% upon validation on different cell types.
• Successfully identified individual cells in G1 or S/G2 phases.

Conclusions:

• The developed supervised machine learning method offers a robust approach for cell cycle staging.
• This tool enhances the ability to classify cell cycle phases at the single-cell level using bioimaging and image analysis.
• The method has significant implications for advancing cell biology research and clinical diagnostics.