Colocalization of fluorescence and Raman microscopic images for the identification of subcellular compartments: a...

Area of Science:

• Biophysics
• Cell Biology
• Spectroscopy

Background:

• Raman microscopy offers label-free cellular and subcellular structure recognition.
• Accurate annotation of Raman spectroscopic images requires identifying colocalization with fluorescence microscopy.
• Current methods rely on fluorescence labeling, limiting label-free applications.

Purpose of the Study:

• To develop a label-free method for resolving cellular compartments using Raman microscopy.
• To establish a supervised classifier for automatic identification of subcellular structures.
• To enable annotation of Raman spectroscopic images without fluorescence labeling.

Main Methods:

• A colocalization scheme correlating fluorescence channels with unsupervised clustering (hierarchical cluster analysis) was developed.
• This scheme was used to pre-select spectra for training a supervised classifier.
• A random forest classifier was trained using selected spectra to identify lipid droplets and the nucleus.

Main Results:

• The colocalization scheme identified statistically significant overlapping regions between spectral and fluorescence images.
• The trained random forest classifier accurately identified specific cellular components (lipid droplets, nucleus) in Raman spectral images.
• Validation confirmed the potential for label-free identification of subcellular structures.

Conclusions:

• The proposed method successfully enables label-free resolution and identification of cellular compartments in Raman spectral images.
• Combining colocalization with supervised machine learning provides a powerful tool for annotating Raman data.
• This approach advances the application of Raman microscopy in biological research by reducing reliance on fluorescence labeling.