PeakLink: a new peptide peak linking method in LC-MS/MS using wavelet and SVM

Area of Science:

• Proteomics
• Analytical Chemistry
• Bioinformatics

Background:

• Accurate peptide peak linking in liquid chromatography-mass spectrometry/tandem mass spectrometry (LC-MS/MS) is essential for tracking protein expression changes between runs.
• Current methods relying on retention time, mass, or peak shape have limited accuracy, especially with complex samples from different conditions, hindering large-scale proteomics studies.
• Linking peptide peaks without tandem mass spectrometry identification in one run to their counterparts in another is a significant challenge.

Purpose of the Study:

• To develop and evaluate a novel algorithm, PeakLink (PL), for improved peptide peak linking in LC-MS/MS data.
• To enhance the accuracy of correlating peptide peaks across different LC-MS/MS runs, particularly when tandem mass spectrometry identification is not available for all peaks.
• To provide a robust solution for large-scale comparative proteomics studies.

Main Methods:

• The PeakLink (PL) algorithm utilizes both time and frequency domain information as input for a non-linear support vector machine (SVM) classifier.
• PL employs an rt likelihood ratio score to filter out peaks with excessive retention time shifts.
• Wavelet transformation is used for noise reduction before calculating peak shape correlation, which is then converted into statistical scores for SVM classification.

Main Results:

• PeakLink (PL) demonstrated significant improvements in linking accuracy across challenging LC-MS/MS datasets.
• The method was successfully tested on samples from diverse disease states, different instruments, and various laboratories.
• PL outperformed existing algorithms in accurately linking peptide peaks between LC-MS/MS runs.

Conclusions:

• PeakLink (PL) offers a more accurate and robust approach for peptide peak linking in LC-MS/MS experiments.
• The method effectively addresses limitations of existing techniques, enabling more reliable large-scale proteomics analyses.
• PL provides a valuable tool for comparative proteomics, facilitating the study of protein expression profiles across different conditions.