Spatially Informed Nonnegative Matrix Trifactorization for Coclustering Mass Spectrometry Data

Area of Science:

• Computational biology
• Biophysics
• Biochemistry

Background:

• Mass spectrometry imaging (MSI) provides high-resolution molecular data from tissues, crucial for understanding biological systems.
• Analyzing large MSI datasets requires efficient computational methods to extract meaningful biological insights.
• Identifying molecules with similar functions based on spatial abundance patterns is a key challenge.

Purpose of the Study:

• To develop a computational method for analyzing mass spectrometry imaging data.
• To identify molecules with similar functions by analyzing their spatial expression patterns.
• To improve the efficiency and scalability of MSI data analysis.

Main Methods:

• TRIFASE, a semi-nonnegative matrix trifactorization technique, was developed.
• An estimation algorithm was proposed to solve the matrix trifactorization problem.
• Two heuristic approximations were introduced to enhance scalability and computational efficiency.

Main Results:

• TRIFASE effectively performs coclustering of molecules based on spatial expression patterns.
• The method accounts for spatial correlation in mass spectrometry imaging data.
• Simulations and a mouse brain tissue analysis validated TRIFASE's ability to extract localized expression patterns and identify functionally linked protein blocks.

Conclusions:

• TRIFASE offers a powerful tool for analyzing mass spectrometry imaging data.
• The method facilitates the discovery of molecular mechanisms by linking protein activation to specific biological processes.
• TRIFASE enhances the understanding of molecular roles in biological systems through spatial analysis.