Machine Learning-Assisted Quantification of Organelle Abundance

Area of Science:

• Cell Biology
• Microscopy
• Computational Biology

Background:

• Organelle abundance is crucial for understanding organelle formation and function.
• Current quantification methods using confocal microscopy rely on user-defined intensity thresholds, introducing bias and variability.
• Automated analysis is needed to standardize organelle quantification.

Purpose of the Study:

• To develop and present a machine learning-assisted algorithm for automated organelle abundance quantification.
• To standardize threshold selection and minimize user intervention in microscopy-based organelle analysis.
• To demonstrate the algorithm's utility across various organelle types and sample organisms.

Main Methods:

• Utilized machine learning with the WEKA segmentation plugin on the open-source Fiji platform.
• Developed an automated algorithm to quantify organelle number, area, and density from fluorescence intensity.
• Trained the algorithm on sample datasets for application to images acquired with identical imaging parameters.

Main Results:

• Successfully automated the quantification of organelle number, area, and density.
• Demonstrated applicability to both membrane-bound (e.g., peroxisomes) and non-membrane-bound organelles (e.g., lipid droplets, stress granules).
• Validated the approach in human cells and whole fish samples.

Conclusions:

• The developed algorithm provides an automated, standardized, and user-independent method for organelle abundance quantification.
• This open-source tool enhances the reliability and reproducibility of microscopy-based organelle studies.
• The machine learning approach offers a robust solution for analyzing diverse cellular compartments.