Modeling Lower-Order Statistics to Enable Decoy-Free FDR Estimation in Proteomics

Area of Science:

• Proteomics
• Computational Biology
• Mass Spectrometry

Background:

• Accurate statistical validation of peptide-spectrum matches (PSMs) is crucial for proteomics.
• Current false discovery rate (FDR) estimation methods often rely on decoy databases, increasing computational cost and relying on unverified assumptions.
• The potential of using non-top-scoring PSMs for FDR estimation has been underexplored.

Purpose of the Study:

• To develop a novel decoy-free procedure for statistical validation of PSMs in proteomics.
• To establish a new method for false discovery rate (FDR) estimation using non-top-scoring target PSMs.
• To compare the performance of the proposed method against existing decoy-based and decoy-free FDR estimation techniques.

Main Methods:

• Proposed a decoy-free procedure for developing null models for top-scoring PSMs.
• Utilized the transformed e-value (TEV) score and distributions of non-top-scoring target PSMs.
• Leveraged a theoretically derivable relationship between TEV score statistics and empirical optimization for parameter fitting.

Main Results:

• The novel method demonstrates comparable performance to existing popular FDR estimation techniques.
• In some instances, the proposed method outperformed both decoy-free and decoy-based approaches.
• The framework was validated across multiple datasets and two different search engines.

Conclusions:

• The developed decoy-free method offers an efficient and effective alternative for FDR estimation in proteomics.
• This approach reduces computational burden and avoids assumptions associated with decoy strategies.
• The findings suggest a promising new direction for statistical validation in mass spectrometry-based proteomics.