A Novel Machine Learning Post-processing Filter for Mass Spectrometry-Based Proteogenomics Leveraging Retention Time...

Area of Science:

• Biochemistry
• Computational Biology
• Proteomics

Background:

• Mass spectrometry-based immunopeptidomics is advancing due to technology and computation.
• Previous work developed a machine learning filter using retention time and peptide physicochemical properties to reduce false positives in shotgun proteomics.
• Mascot-based traditional shotgun proteomic workflows often yield false positives.

Purpose of the Study:

• To provide a practical, step-by-step guide for applying a machine learning model to analyze experimental immunopeptidomics data.
• To empower researchers to implement advanced data analysis techniques for improved peptide identification accuracy.
• To facilitate the use of machine learning filters in routine proteomic workflows.

Main Methods:

• Data preparation and organization for machine learning analysis.
• Application of a previously developed machine learning filter incorporating peptide retention time and predicted physicochemical properties.
• Utilizing Mascot-based shotgun proteomic data for analysis.

Main Results:

• A clear, actionable guide for researchers to analyze their own immunopeptidomics data.
• Demonstration of how to effectively apply a machine learning model to reduce false positives in peptide identification.
• Enhanced accuracy in identifying peptides through the integration of machine learning with traditional proteomic workflows.

Conclusions:

• The developed machine learning approach significantly improves the accuracy of peptide identification in mass spectrometry-based immunopeptidomics.
• This guide enables researchers to readily implement advanced computational tools for more reliable proteomic data analysis.
• The study facilitates the adoption of machine learning filters for more sensitive and accurate immunopeptidomics research.