CPM

CPM aligns and normalizes time-series LC-MS serum proteomic data using Continuous Profile Models to detect differential protein signals for biomarker discovery and class prediction.

Key Features:

• Continuous Profile Models: Implements Continuous Profile Models for alignment and normalization of time-series LC-MS signals.
• Class-differential discovery: Detects protein signal differences between two classes of LC-MS serum proteomic data without requiring tandem mass spectrometry, gels, or labeling techniques.
• Compatibility with low-precision MS: Operates effectively on data from lower-precision mass spectrometers.
• Spike-in validation: Validates performance on controlled spike-in experiments that mimic serum biomarker discovery scenarios.
• Spike-in complexity metric: Introduces a novel method to evaluate the complexity of spike-in problems.
• Ground-truth evaluation: Uses experimentally derived ground truth and precision-recall curves for performance assessment.
• Replicate-based assessment: Demonstrates evaluation using seven replicates per class.
• Public benchmark data: Provides publicly available LC-MS benchmark data for method comparison and validation.

Scientific Applications:

• Biomarker discovery in serum proteomics: Identification of differential protein signals for candidate biomarkers from LC-MS serum data.
• Class prediction for diagnostic studies: Evaluation of class prediction performance in proteomic screening experiments.
• Monitoring physiological responses: Comparative analysis of proteomic time-series to monitor physiological changes.
• Drug action mechanism studies: Detection of proteomic differences relevant to drug mechanism investigations.
• Benchmarking LC-MS analysis methods: Providing experimentally validated datasets and metrics for benchmarking alignment, normalization, and differential-detection methods.

Methodology:

Alignment and normalization of time-series LC-MS signals using Continuous Profile Models; validation on controlled spike-in experiments with an introduced spike-in complexity metric; evaluation with experimentally derived ground truth via precision-recall curves; analysis of differential protein signals between two classes without MS/MS, gels, or labeling, using seven replicates per class.