msCRUSH: Fast Tandem Mass Spectral Clustering Using Locality Sensitive Hashing.

Area of Science:

• Proteomics
• Bioinformatics
• Computational Biology

Background:

• Large-scale proteomics generates massive, redundant tandem mass spectra (MS/MS).
• Peptide identification via database searching is a bottleneck in proteomic data analysis.
• Spectral clustering aims to reduce data redundancy while maintaining peptide identification sensitivity.

Purpose of the Study:

• To present msCRUSH, a novel software for spectral clustering in proteomics.
• To address the challenge of reducing MS/MS spectra redundancy while preserving peptide identification sensitivity.
• To improve the efficiency and accuracy of proteomic data analysis.

Main Methods:

• Implementation of a novel spectral clustering algorithm based on locality-sensitive hashing.
• Testing msCRUSH on a large dataset of 23.6 million tandem mass spectra.
• Comparison of msCRUSH performance against state-of-the-art spectral clustering software (PRIDE Cluster).

Main Results:

• msCRUSH demonstrated 6.9-11.3 times faster performance than PRIDE Cluster.
• Achieved higher clustering sensitivity and comparable accuracy to existing methods.
• Consensus spectra from msCRUSH enabled MSGF+ and Mascot to identify 3% and 1% more unique peptides, respectively, at 1% FDR.

Conclusions:

• msCRUSH offers a significant speed improvement for spectral clustering in large-scale proteomics.
• The software enhances peptide identification rates by improving spectral data quality.
• msCRUSH is an open-source C++ tool, facilitating broader adoption in proteomic data analysis.