Multi-class alignment of LC-MS data using probabilistic-based mixture regression models.

Area of Science:

• Analytical Chemistry
• Computational Biology
• Biostatistics

Background:

• Liquid chromatography-mass spectrometry (LC-MS) is crucial for analyzing complex biological samples.
• Accurate alignment of LC-MS spectra is essential for reliable data comparison and analysis.
• Existing alignment methods like dynamic time warping (DTW) and continuous profile model (CPM) have limitations.

Purpose of the Study:

• To present a novel framework for multi-class alignment of LC-MS data using probabilistic-based mixture regression models (PMRM).
• To improve the accuracy of LC-MS spectral alignment by considering both time and measurement spaces.
• To evaluate the performance of the proposed PMRM framework against established alignment techniques.

Main Methods:

• Development of a probabilistic-based mixture regression model (PMRM) framework.
• Implementation of the expectation maximization (EM) algorithm for parameter estimation.
• Incorporation of spline-based mixture regression models and prior transformation densities to model data variability.
• Application to align replicate LC-MS spectra from lysed E.coli cell proteins.

Main Results:

• The PMRM framework successfully aligned LC-MS spectra in both time and measurement domains.
• The proposed method demonstrated robust performance in handling variability within LC-MS data.
• Comparative analysis indicated that PMRM potentially offers advantages over DTW and CPM for LC-MS alignment.

Conclusions:

• Probabilistic-based mixture regression models provide a powerful framework for LC-MS data alignment.
• The PMRM approach effectively accounts for temporal and measurement variability in LC-MS spectra.
• This method offers a promising alternative for enhancing the accuracy and reliability of LC-MS data analysis.