Prediction of cell cycle distribution after drug exposure by high content imaging analysis using low-toxic DNA...

Area of Science:

• Cell Biology
• Biotechnology
• Computational Biology

Background:

• Anticancer drugs often interfere with cell cycle progression.
• Accurate cell cycle phase prediction is crucial for understanding drug mechanisms and sensitivity.

Purpose of the Study:

• To develop a machine learning model for predicting cell cycle phases using high-content imaging data.
• To evaluate the model's performance and identify key predictive features.
• To explore the potential application of the model in ex vivo drug sensitivity diagnosis.

Main Methods:

• Utilized HeLa and MCF7 cancer cells expressing the fluorescent ubiquitination-based cell cycle indicator (Fucci).
• Performed high-content imaging analysis after 36-hour drug exposure, staining with SiR-DNA.
• Developed binary classification models (G1, early S, S/G2/M) using four supervised machine learning algorithms, selecting Random Forest.
• Validated models using 7500 training and 2500 validation samples.

Main Results:

• The Random Forest model achieved an accuracy of 75%-87% through 10-fold cross-validation.
• Biologically relevant features like signal intensity and nuclear size were highly ranked, confirming model validity.
• The model successfully predicted cell cycle phases in cancer cells from DNA-staining images.

Conclusions:

• A robust cell cycle prediction model was successfully developed using simple DNA-staining imaging and machine learning.
• The model demonstrates potential for future ex vivo drug sensitivity diagnosis.
• This approach offers a valuable tool for cancer research and therapeutic development.