Automatic selection of preprocessing methods for improving predictions on mass spectrometry protein profiles

Area of Science:

• Biomedical Engineering
• Analytical Chemistry
• Bioinformatics

Background:

• Mass spectrometry proteomic profiling shows promise for clinical screening.
• Standardized preprocessing of noisy raw mass spectrometry data is a significant challenge.
• Current methods require expert knowledge for optimal data processing.

Purpose of the Study:

• To develop an automated system for selecting optimal preprocessing methods for mass spectrometry data.
• To address the obstacle of data standardization in clinical proteomic profiling.
• To reduce the need for specialized expertise in data preprocessing.

Main Methods:

• Developed a system for automatic selection of preprocessing methods.
• Introduced novel metrics to balance noise correction and information preservation.
• Evaluated the system on Surface-Enhanced Laser Desorption/Ionization (SELDI) and Matrix-Assisted Laser Desorption/Ionization (MALDI) datasets.

Main Results:

• The automated system successfully determined optimal preprocessing pipelines.
• The developed metrics effectively balanced noise reduction and data integrity.
• Preprocessing using the automated system led to improved downstream classification performance.

Conclusions:

• Automated preprocessing enhances the utility of mass spectrometry for clinical screening.
• The system offers a standardized and accessible approach to data analysis.
• Improved data preprocessing directly translates to more accurate diagnostic or prognostic models.