Unsupervised machine learning using an imaging mass spectrometry dataset automatically reassembles grey and white...

Area of Science:

• Neuroscience
• Analytical Chemistry
• Biochemistry

Background:

• Current methods struggle to map small biomolecules like lipids in the brain.
• Understanding lipid distribution is crucial for brain function and disease research.

Purpose of the Study:

• To develop and optimize imaging mass spectrometry (IMS) for visualizing small biomolecule distribution.
• To investigate the spatial distribution of lipids in rat brain tissue.
• To apply machine learning for brain structure reconstruction based on molecular data.

Main Methods:

• Utilized solarix XR, a high-resolution MALDI-FT-ICR-MS instrument for molecular detection.
• Employed imaging mass spectrometry (IMS) to map biomolecule distribution in rat brains.
• Applied unsupervised machine learning techniques, including principal component analysis (PCA) and hierarchical clustering.

Main Results:

• Demonstrated differential distribution of phospholipids in neuronal cell bodies and axons.
• Successfully reconstructed brain structures, including grey and white matter, using PCA without predefined anatomical labels.
• Classified 10 molecular groups based on distribution patterns, identifying phospholipids as key components in the cerebellar cortex.

Conclusions:

• IMS is a powerful tool for comprehensive biomolecule distribution analysis in the brain.
• Machine learning algorithms can effectively reconstruct brain architecture from molecular data.
• Phospholipids exhibit specific localization patterns within the rat brain, particularly in the cerebellar cortex.