DeLTA 2.0: A deep learning pipeline for quantifying single-cell spatial and temporal dynamics

Area of Science:

• Microbiology
• Cell Biology
• Bioinformatics

Background:

• Microscopy advancements accelerate image acquisition, creating an analysis bottleneck for quantitative single-cell data.
• Existing bacterial segmentation and tracking tools often require manual input, lack accuracy, or are specific to experimental setups.

Purpose of the Study:

• Introduce DeLTA 2.0, a Python workflow for rapid, accurate analysis of single-cell images on 2D surfaces.
• Quantify gene expression and cell growth using deep convolutional neural networks, eliminating the need for human input post-training.
• Extend single-cell analysis capabilities to 2D growth environments, enabling studies on co-cultures and multi-generational phenomena.

Main Methods:

• Utilized deep convolutional neural networks for single-cell information extraction from time-lapse microscopy images.
• Developed a purely Python workflow, DeLTA 2.0, for automated segmentation and tracking.
• Extended functionality from microfluidic devices to general 2D growth environments.

Main Results:

• DeLTA 2.0 achieves rapid analysis (under 10 minutes for complete movies) with high accuracy (around 1% error rate).
• Successfully analyzed mixed populations of antibiotic-resistant and susceptible cells in 2D environments.
• Tracked pole age and growth rate across multiple generations in 2D cultures.

Conclusions:

• DeLTA 2.0 provides a powerful, accessible, and efficient tool for analyzing time-lapse microscopy data in various 2D cell growth settings.
• The workflow's improvements, including broad file format compatibility and a Google Colab notebook, increase user accessibility.
• Automated analysis of single-cell dynamics in 2D environments facilitates complex biological studies, such as those involving antibiotic resistance and multi-generational effects.