Deciphering anomalous heterogeneous intracellular transport with neural networks

Area of Science:

• Cell Biology
• Biophysics
• Computational Biology

Background:

• Intracellular transport is complex, showing heterogeneous and non-Brownian movement patterns over time and space.
• Traditional averaging methods often fail to capture the full extent of this movement heterogeneity.
• Accurate characterization is crucial for understanding cellular functions and organelle dynamics.

Purpose of the Study:

• To develop a novel, accurate, and efficient method for resolving heterogeneous intracellular transport.
• To enable robust analysis of short time series data for rapidly moving cellular structures.
• To interpret anomalous intracellular dynamics and compare behaviors of closely related organelles.

Main Methods:

• Developed a deep learning feedforward neural network trained on fractional Brownian motion.
• The method requires significantly fewer data points than established techniques.
• Enabled robust estimation of Hurst exponents for short time series data.

Main Results:

• Successfully resolved heterogeneous behavior of intracellular transport in space and time.
• Enabled direct, dynamic segmentation and analysis of experimental tracks (e.g., endosomes, lysosomes).
• Interpreted anomalous intracellular dynamics using fractional Brownian motion with a stochastic Hurst exponent.

Conclusions:

• The deep learning approach provides a powerful tool for analyzing complex intracellular transport.
• Revealed unexpected differences in dynamics between closely related endocytic organelles.
• This method advances the study of cellular transport and organelle function.