Multi-component analysis: blind extraction of pure components mass spectra using sparse component analysis.

Area of Science:

• Analytical Chemistry
• Chemometrics
• Signal Processing

Background:

• Blind source separation (BSS) typically requires more mixture signals than independent sources.
• Decomposing mass spectra mixtures into pure components is crucial for chemical compound identification.
• Existing BSS methods often struggle when the number of mixtures is less than the number of pure components.

Purpose of the Study:

• To present a novel sparse component analysis (SCA)-based method for blind decomposition of mass spectra mixtures.
• To address the challenge of decomposing mixtures when the number of mixtures is less than the number of pure components.
• To demonstrate the capability of SCA in extracting pure mass spectra under such conditions.

Main Methods:

• Sparse Component Analysis (SCA) using l(1) norm minimization via linear programming.
• SCA implemented through multilayer hierarchical alternating least squares nonnegative matrix factorization with sparseness constraints.
• Robust data clustering algorithm combined with pure component concentration matrix for estimating the number of components.

Main Results:

• Successfully demonstrated the blind extraction of five pure component mass spectra from only two mixtures.
• Showcased the effectiveness of SCA-based methods in scenarios where mixtures are fewer than pure components.
• Validated that no a priori information on the number of pure components is required for decomposition.

Conclusions:

• The proposed SCA methodology offers a robust solution for blind decomposition of mass spectra mixtures, overcoming limitations of traditional BSS.
• The method is capable of accurately identifying pure components even with limited mixture data.
• This approach can be integrated into existing software for mass spectra analysis and chemical compound identification.