Understand the source comparison
Local versus global FDR
Global FDR measures error rate across your entire filtered dataset. If you report 1000 proteins at 1% FDR, approximately 10 are expected to be wrong. You do not know which 10, just that the overall proportion is about 1%. Local FDR, sometimes called posterior
No winner is assigned.
This page preserves a source comparison for education. It does not add a rating, recommendation or clinical judgment.
- Global FDR measures error rate across your entire filtered dataset. If you report 1000 proteins at 1% FDR, approximately 10 are expected to be wrong. You do not know which 10, just that the overall proportion is about 1%.
- Local FDR, sometimes called posterior error probability, estimates error probability for individual identifications. A peptide with 0.1% local FDR has a 0.1% chance of being wrong, while another peptide with 5% local FDR has a 5% chance of being wrong. Both might pass a 5% global FDR threshold but have different individual confidences.
- Local FDR provides richer information for downstream analysis. When prioritizing proteins for validation, prefer those with lower local FDR. When weighting quantitative data, account for identification uncertainty through local FDR. When combining evidence across experiments, local FDR enables principled integration.
- Calculating local FDR requires modeling score distributions for correct and incorrect matches, then applying Bayes theorem. Tools like Percolator estimate local FDR alongside global FDR. The computational overhead is modest compared to the additional information gained.