Target-Decoy FDR in Peptide Mass Spectrometry (2D)

Ion Source

Accelerating voltage 4000 V

Target-Decoy FDR

Allowed FDR 1%

Ions detected

Accepted PSMs

Empirical FDR

—

Score cutoff

—

Legend

Target ion in flight Decoy ion in flight Accepted target (true ID) Accepted decoy (false ID) Rejected below cutoff

How it works

A precursor peptide ion of mass-to-charge ratio m/z is accelerated through voltage V and drifts down a field-free flight tube of length L. Its kinetic energy equals the work done by the field, so its speed depends only on m/z and V:

zeV = ½mv²  →  v = sqrt(2·zeV / m) ∝ sqrt(V / (m/z))
Flight time: t = L / v

Lighter ions (low m/z) outrun heavier ones, so a single burst fans out along the tube and arrives at the detector spread out in time — this time-of-flight is how a real TOF instrument reads out mass. Raising the voltage speeds every ion up and compresses the spread. Drag the flight-tube panel to pan, scroll/pinch to zoom.

Each detected ion is matched against a sequence database that has been doubled with reversed "decoy" sequences of the same length and amino-acid composition. A real peptide (target) tends to score higher against the true database than a decoy does, but the two score distributions overlap — some decoys score high by chance, and some real matches score low. The bottom panel plots both live score distributions. The target-decoy strategy turns this overlap into a false-discovery-rate estimate:

FDR(s) ≈ #decoys scoring ≥ s  /  #(targets + decoys) scoring ≥ s

Sliding the FDR slider looks up the highest score cutoff whose cumulative FDR (monotonically enforced from the top score down) stays at or below your chosen threshold, then re-classifies every future hit against that cutoff live. A stricter FDR (1%) accepts fewer identifications but almost all of them true; a looser FDR (20%) accepts more, at the cost of visibly more red decoy bars mixed into the "real" spectrum.

DDA (data-dependent acquisition) selects a handful of the most intense precursors per cycle — fewer ions, cleaner, more separable scores.
DIA (data-independent acquisition) fragments every precursor in a wide isolation window at once — far more ions per cycle, higher throughput, but more co-fragmentation noise, so target and decoy scores overlap more and the same FDR threshold accepts proportionally fewer IDs.

Burst interval and ion count follow the acquisition mode

drag = pan · wheel = zoom (flight tube)