Background noise

Cleaning the signal before anything is interpreted, one locus at a time.

Cleaning the signal: background noise

Before interpreting anything, SAgA removes non-specific background, one HLA locus at a time. Non-specific binding can lift every bead of a locus, so that beads carrying no real antibody still read above zero — a recipe for false positives.

SAgA handles this in two steps:

Take the lowest bead of the locus and treat that minimum as the locus background level.
Subtract that background from every bead of the locus. Beads with no real antibody drop to zero; genuine reactivities keep only their true excess.

The before / after below uses one A-locus panel: the three A2 beads sit at 5600, but background has lifted every other bead to 600.

Raw panel: taken at face value, the background produces three false positives at 600 — Taken at face value, SAgA would return four antibodies — the real anti-A2 at 5600, plus *three false positives* (A1, A3, A11 at 600) produced by the background, not by real antibodies.

After subtracting the locus minimum of 600, only the true anti-A2 signal survives — SAgA reads the locus minimum (600) as background and subtracts it from every bead: the negatives drop to zero and the A2 beads keep only their true excess (5600 − 600 = 5000). A single anti-A2 antibody survives; the three false positives are gone.

The safeguard

If at least one bead of the locus already reads zero, the minimum is zero and SAgA can no longer assume a uniform background — so it does not subtract. Lifted beads are then kept as-is and reported as real positives. It is up to you to judge them.

When one bead reads zero, subtraction is suspended and lifted beads are kept — Here one bead is genuinely at zero, so the locus minimum is zero. Subtraction is not performed and the beads at 600 come out as real positives (A3, A11) next to the true anti-A2.

Cases for this part