pValid: Validation Beyond the Target-Decoy Approach for Peptide Identification in Shotgun Proteomics

Area of Science:

• Proteomics
• Mass Spectrometry
• Bioinformatics

Background:

• The target-decoy approach is standard for mass spectrometry-based proteomics validation.
• It estimates false discovery rate but cannot identify incorrect identifications within the threshold.
• Improved validation methods are crucial for accurate proteomic data analysis.

Purpose of the Study:

• To characterize a framework for validating peptide identifications.
• To investigate and compare common and novel validation methods.
• To introduce and evaluate a new machine-learning-based validation tool, pValid.

Main Methods:

• Defined validation method accuracy using false-positive rate (FPR) and false-negative rate (FNR).
• Developed pValid incorporating open database search and theoretical spectrum prediction via machine learning.
• Compared pValid against four common methods and a synthetic peptide method on benchmark datasets.

Main Results:

• pValid demonstrated superior performance with an average FPR of 0.03% and FNR of 1.79% across three benchmark datasets.
• pValid outperformed the synthetic peptide validation method on a dedicated synthetic peptide dataset.
• On a large-scale human proteome dataset, pValid identified the most incorrect peptide identifications among all tested methods.

Conclusions:

• pValid offers significantly lower FPR and FNR compared to existing methods.
• The method shows high sensitivity and precision in peptide identification validation.
• pValid is a cost-effective and feasible tool for validating peptide identifications in mass spectrometry-based proteomics.