Solution to Statistical Challenges in Proteomics Is More Statistics, Not Less.

Area of Science:

• Proteomics
• Bioinformatics
• Statistical analysis

Background:

• High-throughput scientific studies require accurate estimation of incorrect findings, known as the false discovery rate (FDR).
• In proteomics, there's a common misconception conflating protein FDR with the protein-level target-decoy method for its estimation.
• This confusion has led to arguments against protein-level FDR analysis, questioning its validity.

Purpose of the Study:

• To clarify the invariant nature of protein-level false discovery rate (FDR) in proteomics.
• To demonstrate the significant errors arising from neglecting protein FDR and multiple testing corrections in large-scale proteomic analyses.
• To refute the notion that protein FDR is a flawed concept.

Main Methods:

• Utilized simple, accurate simulations to model proteomic data analysis.
• Assessed the impact of ignoring protein-level FDR and multiple testing corrections.
• Analyzed the behavior of protein identification under large dataset conditions.

Main Results:

• Confirmed that protein-level FDR is a well-defined and invariant quantity.
• Simulations showed that failure to account for protein FDR can lead to substantial, unrecognized errors.
• Large datasets without proper FDR control can result in the incorrect identification of thousands of absent proteins.

Conclusions:

• Protein-level FDR is a valid and essential metric in proteomics.
• Neglecting protein FDR control in high-throughput proteomics can lead to widespread false discoveries.
• Accurate FDR estimation is critical for reliable interpretation of proteomic study results.