PeakLink

PeakLink links corresponding peptide peaks across proteomics datasets by combining time-domain and frequency-domain features with wavelet denoising and a non-linear SVM classifier to improve peptide peak correspondence for comparative proteomics.

Key Features:

• Dual-Domain Analysis: Uses time-domain and frequency-domain inputs to a non-linear support vector machine (SVM) classifier to enhance identification of corresponding peptide peaks.
• Retention Time Alignment: Applies a threshold on a retention time (rt) likelihood-ratio score to exclude candidate peak pairs with excessive elution time shifts.
• Noise Reduction and Correlation Calculation: Employs wavelet transformation to reduce noise and converts retention time and peak-shape correlations into statistical scores.
• SVM Classifier Training and Application: Trains and applies a non-linear SVM classifier using the computed statistical scores to distinguish corresponding from non-corresponding peptide peaks.

Scientific Applications:

• Comparative proteomics: Enhances peptide peak linkage accuracy in large-scale proteomics studies for comparing protein expression profiles across numerous samples.
• Cross-condition and cross-platform analysis: Enables robust comparison of peptide peaks across samples from different disease states, instruments, and laboratories.

Methodology:

Computational steps include initial filtering of candidate peaks by a retention time (rt) likelihood-ratio score threshold to remove large elution shifts; wavelet transformation for noise reduction; calculation of retention time and peak-shape correlations and conversion into statistical scores; use of time-domain and frequency-domain features as inputs; and training and application of a non-linear SVM classifier to distinguish corresponding versus non-corresponding peptide peaks.