DeepCycle reconstructs a cyclic cell cycle trajectory from unsegmented cell images using convolutional neural...

Area of Science:

• Cell Biology
• Computational Biology
• Machine Learning

Background:

• Understanding cell cycle progression is crucial for cellular characterization and biological process analysis.
• Single-cell methods offer unprecedented detail in studying the cell cycle.

Purpose of the Study:

• To develop DeepCycle, a deep learning method for estimating cell cycle trajectories from unsegmented single-cell microscopy images.
• To validate DeepCycle's accuracy and ability to resolve the complete cell cycle, including division.

Main Methods:

• Deep learning model (DeepCycle) utilizing brightfield and nuclei-specific fluorescent signals.
• Evaluation on 2.6 million single-cell microscopy images of MDCKII cells with the FUCCI2 system.
• Validation against real-time live-cell imaging data.

Main Results:

• DeepCycle generated a latent representation of cell images, revealing a continuous and closed cell cycle trajectory.
• The model demonstrated a nearly perfect correlation with experimentally measured cell cycle progression.
• This is the first model to resolve a closed cell cycle trajectory solely from unsegmented microscopy data.

Conclusions:

• DeepCycle accurately estimates cell cycle trajectories from basic microscopy data.
• The method provides a powerful tool for in-depth cellular characterization and cell cycle research.
• DeepCycle's ability to resolve the complete cell cycle trajectory opens new avenues in biological studies.