Targeted Feature Detection for Data-Dependent Shotgun Proteomics.

Area of Science:

• Proteomics
• Computational Biology
• Biochemistry

Background:

• Label-free quantification in shotgun LC-MS/MS proteomics is crucial but computationally challenging.
• Accurate peptide identification and signal detection are vital for reliable quantification.
• Missing values due to imperfect feature detection limit proteomic analysis.

Purpose of the Study:

• To develop a novel feature detection algorithm for label-free quantitative proteomics.
• To minimize missing values and improve the robustness of peptide quantification.
• To provide confidence scores for detected features and estimate the false discovery rate (FDR).

Main Methods:

• Developed FeatureFinderIdentification (FFId) within the OpenMS framework, leveraging OpenSWATH algorithms.
• Implemented a targeted approach using an MS1 assay library based on identified peptides.
• Utilized a support vector machine (SVM) classifier trained on internal and external peptide identifications to score feature candidates.

Main Results:

• FFId achieved >99% quantification coverage for identified peptides at 1% peptide-spectrum match (PSM) FDR.
• The algorithm demonstrated competitive quantification accuracy and reproducibility across replicates.
• Average FDR for feature selection was low (1.5% per sample, 3% for externally inferred features).

Conclusions:

• FFId offers a robust and accurate solution for label-free quantitative proteomics.
• The targeted approach significantly reduces missing values and enhances quantification reliability.
• FFId is open-source, freely available within OpenMS, and provides reliable FDR estimation.