MSPLIT

MSPLIT identifies mixture tandem mass spectra by matching experimental MS/MS spectra against spectral libraries to determine up to two constituent peptides and their relative abundances.

Key Features:

• Identification of up to two peptides: Identifies MS/MS spectra that originate from mixtures of up to two peptides.
• Use of spectral libraries: Matches experimental spectra to spectral libraries containing single-peptide spectra for identification.
• Quantitative abundance handling: Employs a quantitative approach that identifies up to 98% of mixture spectra for equally abundant peptides and accommodates abundance ratios up to 10:1.
• Theoretical bounds on similarity: Applies theoretical bounds on spectral similarity to avoid exhaustive comparison of all peptide combinations.
• Computational speed: Achieves speed improvements reported up to five orders of magnitude and can identify mixture spectra in seconds using proteome-scale libraries.
• Generality: Underlying methods are applicable to other types of spectral libraries and mixture spectra beyond peptide MS/MS.

Scientific Applications:

• Proteomic analysis: Enables more accurate identification of peptide mixtures in complex proteomic samples.
• Biomarker discovery: Supports detection of peptide signatures relevant to biomarker studies.
• Disease diagnostics: Facilitates analysis of complex peptide mixtures relevant to diagnostic workflows.
• Study of protein interactions and modifications: Assists investigation of co-occurring peptides and post-translational modifications in mixture spectra.
• Complex biological systems: Aids interpretation of proteomic data from systems with high sample complexity.
• Personalized medicine: Contributes to analyses that inform individualized proteomic profiling.

Methodology:

Matches experimental MS/MS spectra to spectral libraries of single-peptide spectra, identifies up to two constituent peptides using a quantitative scoring approach that handles varying abundance ratios (including up to 10:1 and ~98% recovery for equal abundances), and uses theoretical bounds on spectral similarity to prune candidate comparisons and avoid exhaustive pairwise searches, yielding large speedups (up to five orders of magnitude) for proteome-scale libraries.