DeepContact: High-throughput quantification of membrane contact sites based on electron microscopy imaging

Area of Science:

• Cell Biology
• Bioimaging
• Computational Biology

Background:

• Membrane contact sites (MCSs) are crucial for cellular functions.
• Quantitative analysis of MCSs is vital for understanding cellular responses in health and disease.
• Existing tools lack efficiency and flexibility for comprehensive MCS analysis.

Purpose of the Study:

• To develop an efficient deep-learning tool, DeepContact, for organelle segmentation and MCS analysis.
• To enable quantitative and statistical analysis of diverse MCS types.
• To explore the role of MCS dynamics in physiological and pathological conditions.

Main Methods:

• Developed DeepContact, a deep-learning protocol for label-free electron microscopy (EM).
• Optimized organelle segmentation and contact analysis.
• Enabled interactive visualization and accommodation of varied organelle morphologies and contact widths.

Main Results:

• DeepContact demonstrated high efficiency and flexibility in analyzing MCSs.
• The tool enabled statistical analysis of various MCS types across multiple biological systems.
• Unidentified coordinative rearrangements of MCS types were profiled in cultured cells.
• A dynamic wave of ER-mitochondrial entanglement in Sertoli cells was revealed.

Conclusions:

• DeepContact provides an efficient and versatile platform for quantitative MCS analysis.
• The tool facilitates the discovery of novel cellular dynamics and organelle interactions.
• DeepContact has the potential to bridge the understanding of MCS dynamics with physiological and pathological processes.