Machine learning-based peptide-spectrum match rescoring opens up the immunopeptidome.

Area of Science:

• Mass spectrometry
• Immunology
• Bioinformatics

Background:

• Immunopeptidomics is vital for identifying immunotherapy and vaccine targets.
• Identifying immunopeptides is difficult due to non-tryptic nature and vast search spaces.
• Challenges include somatic mutations, pathogen genomes, and post-translational modifications.

Purpose of the Study:

• To review bioinformatics pipelines for peptide-spectrum match rescoring in immunopeptidomics.
• To discuss machine learning applications for improving immunopeptide identification.
• To provide insights into current and future ML solutions for immunopeptidomics data analysis.

Main Methods:

• Peptide-spectrum match rescoring using machine learning.
• Incorporation of predicted peptidoform properties (fragment ion intensities, retention time, collisional cross section).
• Analysis of diverse bioinformatics pipelines for immunopeptidomics data.

Main Results:

• Peptide-spectrum match rescoring improves accuracy and sensitivity in immunopeptide identification.
• Machine learning effectively addresses challenges in mass spectrometry-based immunopeptidomics data analysis.
• Utilizing predicted peptidoform properties enhances the distinction between correct and incorrect peptide-spectrum matches.

Conclusions:

• Machine learning-based rescoring is a powerful tool for overcoming immunopeptidomics identification challenges.
• Future ML solutions hold promise for further boosting immunopeptide discovery.
• This review highlights the importance of bioinformatics pipelines and ML in advancing immunopeptidomics research.