Target-Decoy FDR Control in Proteogenomics (2D)

Search Space

Protein database size

PTM search expansion

Spectra searched 4000

Acceptance Threshold

PSM score cutoff 55

Log-scale histogram

Target PSMs Decoy PSMs

Live Statistics

Estimated FDR = decoys / targets above cutoff

0.0%

Targets accepted

Decoys accepted

Estimated true positives (targets − decoys)

Threshold for 1% FDR

–

How it works

In a proteogenomic pipeline, mass-spectrometry spectra are searched against a target protein database built from RNA-seq-derived transcripts. Every spectrum gets a peptide-spectrum match (PSM) score, but some "hits" are pure chance — the search engine matched noise to some peptide simply because the database was big enough to contain a plausible impostor.

The target-decoy strategy measures this directly: the same spectra are also searched against a decoy database (reversed or shuffled sequences that cannot correspond to real proteins). Any decoy hit is by definition a false match, so decoy hits at a given score threshold estimate how many target hits at that same score are also false:

FDR(threshold) ≈ #decoy PSMs ≥ threshold
                  ────────────────────────
                  #target PSMs ≥ threshold

Database size — a bigger search space gives random noise more candidate peptides to coincidentally match well, shifting the decoy score distribution upward (extreme-value statistics of more trials).
PTM search expansion — allowing more post-translational modification variants multiplies the number of candidate peptide forms per spectrum; more candidates again means a better chance a wrong one scores high, so the decoy histogram grows a heavier high-score tail.
Score cutoff — drag directly on the histogram or scatter strip, or use the slider. Raising it keeps only the most confident matches but discards some true ones; lowering it recovers more true peptides at the cost of admitting decoys (and their statistically identical target-side false matches).
The middle panel plots FDR at every possible threshold at once — the acceptance cutoff is just one point sliding along that curve.
This exact target-decoy-controlled peptide list is what feeds neoantigen discovery: only PSMs surviving a strict FDR (commonly 1%) are trusted enough to call a genuinely novel, tumor-specific peptide worth validating.

Score histogram drag to set cutoff

FDR vs. threshold

PSM scatter strip drag to pan · wheel to zoom