Every automated biofoundry screening run needs a way to prove the plate itself is trustworthy before trusting any hit on it. The standard metric, used across real HTS/biofoundry pipelines, is the Z′-factor:
Z' = 1 − ( 3·σ_p + 3·σ_n ) / |μ_p − μ_n|
σ_p, μ_p = std-dev / mean of the positive-control wells
σ_n, μ_n = std-dev / mean of the negative-control wells
Z′ compares the combined spread of the two control populations (3σ each side, ≈99.7% coverage) against how far apart their means sit. It only uses the plate's own control wells, so it is a property of the assay/robot combination, not of any single measurement:
- Z′ > 0.5 — excellent assay, controls are cleanly separated, safe to run at scale.
- 0 ≤ Z′ ≤ 0.5 — marginal; usable but hits near the noise floor are unreliable.
- Z′ < 0 — the control populations overlap; the plate cannot reliably tell a hit from noise.
Well noise (CV) sets the per-well coefficient of variation from liquid-handling and detector variance. Signal window is the raw separation an automated liquid handler achieves between its positive and negative control reagents. Edge evaporation drift reproduces a real, well-documented HTS artifact: outer wells lose volume to evaporation faster than inner wells, systematically biasing their signal and inflating σ — this is exactly why real biofoundries track edge-vs-center Δ as a QC signal alongside Z′ itself.
Wells more than 3σ from their expected population mean are flagged (highlighted) — the same rule an automated pipeline uses to gate a plate before its data ever reaches a scientist.
How this 2D companion differs — this reproduces the 3D source's generator, statistics and flagging rule exactly (same Box-Muller draws, same column convention, same edge-falloff and 3σ gate). No numerical bug was found in the 3D source's Z′/CV/edge-delta formulas — verified separately against the standard Zhang/Chung/Oldenburg Z′ definition with synthetic low-noise/high-noise runs, both of which produced the expected excellent/unusable verdicts. Only the presentation changes: the orbitable 3D plate becomes a draggable, zoomable top-down grid, so every well in a 384-well plate stays legible and hoverable at once instead of depending on camera distance.