Estimating the confidence of peptide identifications without decoy databases

Area of Science:

• Proteomics
• Computational Biology
• Mass Spectrometry

Background:

• Decoy databases are standard for peptide identification confidence in mass spectrometry proteomics.
• However, they increase runtime and are unsuitable for certain complex searches like error-tolerant, de novo, or cross-linked peptide identification.

Purpose of the Study:

• To develop a fast, simple, and robust mixture modeling approach for estimating peptide identification confidence.
• To provide an alternative to decoy database searches that is applicable to a wider range of proteomics problems.

Main Methods:

• A mixture modeling approach was developed to estimate confidence without requiring decoy databases.
• The method includes an automatic check for the fulfillment of its underlying assumptions.
• The approach was evaluated on 41 diverse Liquid Chromatography-Mass Spectrometry (LC/MS) datasets.

Main Results:

• The mixture modeling approach yielded results highly comparable to the decoy database strategy.
• The computational cost associated with this new method is negligible.
• The approach demonstrated applicability to standard protein identification and specialized problems where decoy databases are not feasible.

Conclusions:

• The proposed mixture modeling approach offers a computationally efficient and robust alternative for assessing peptide identification confidence in mass spectrometry.
• This method expands the scope of reliable peptide identification, particularly for complex search types and large-scale analyses where traditional decoy strategies are limited.