Bayesian mass spectra peak alignment from mass charge ratios.

Area of Science:

• Biomedical research
• Analytical chemistry
• Computational biology

Background:

• Proteomics studies using mass spectrometry (MS) are vital for biomarker discovery and disease diagnosis.
• MS data often contains noise, necessitating robust preprocessing for accurate analysis.
• Peak alignment in MS spectra is challenging due to variations in peak locations.

Purpose of the Study:

• To develop a novel Bayesian algorithm for global mass spectrometry peak alignment.
• To improve sample classification using aligned peaks from MS data.
• To offer a parameter-light approach for MS data preprocessing.

Main Methods:

• A new Bayesian algorithm, random grafting-pruning Markov chain Monte Carlo (RGPMCMC), was developed.
• The RGPMCMC algorithm was applied to global MS peak alignment.
• A model-based criterion was used for sample classification based on aligned peaks.

Main Results:

• The RGPMCMC algorithm demonstrated effectiveness in simulation studies compared to existing methods.
• Application to ovarian cancer MALDI-MS data yielded a lower 10-fold cross-validation error rate.
• The method provides a simpler alternative to intensive parameter tuning for MS data analysis.

Conclusions:

• The RGPMCMC algorithm offers a significant advancement in MS data preprocessing for proteomics.
• This approach enhances the reliability of biomarker discovery and disease diagnosis using MS.
• The RGPMCMC method is effective for both peak alignment and sample classification in complex MS datasets.