Predicting cell cycle stage from 3D single-cell nuclear-stained images.

Area of Science:

• Cell Biology
• Biomedical Research
• Machine Learning Applications

Background:

• Cell cycle dynamics are crucial for eukaryotic cell proliferation and are fundamental to biological and medical research.
• Existing cell cycle profiling methods often require complex experimental procedures that can complicate data interpretation.
• There is a need for simplified, cost-effective, and minimally invasive techniques for accurate cell cycle staging.

Purpose of the Study:

• To develop a machine learning (ML) workflow, CellCycleNet, for simplified cell cycle staging from fluorescent microscopy data.
• To enable accurate cell cycle phase prediction with minimal experimenter intervention and cost.
• To provide a community resource including data, models, and software for cell cycle analysis.

Main Methods:

• Developed CellCycleNet, a machine learning workflow for cell cycle staging.
• Utilized DAPI staining for nuclear visualization in fixed interphase cells.
• Trained 2D and 3D machine learning models (support vector machine and deep neural network) using the Fucci2a reporter system as ground truth.
• Evaluated model performance on two benchmarking image datasets, focusing on G1 vs. S/G2 phase classification.

Main Results:

• CellCycleNet accurately predicts cell cycle phase using only DAPI-stained nuclei.
• 3D CellCycleNet models demonstrated superior performance compared to 2D models and support vector machine approaches.
• Simultaneous training on multiple datasets yielded the highest classification accuracy (AUROC of 0.94-0.95).
• Incorporating 3D features significantly enhanced classification performance across all tested model architectures.

Conclusions:

• CellCycleNet offers a simplified and effective approach to cell cycle staging from microscopy images.
• The use of 3D features is critical for improving the accuracy of machine learning-based cell cycle profiling.
• The released data, models, and software facilitate broader adoption and advancement in cell cycle research.