Microplate layout design — controls, references, and well assignment
How wells are assigned on a microplate directly affects data quality, downstream analysis, and how detectable systematic errors are once they have occurred. Layout design is rarely discussed explicitly during assay development — but changing a layout after data collection has started is expensive, because it breaks comparability with everything collected before.
Edge wells on microplates behave differently from interior wells. Temperature gradients, evaporation, and handling-induced vibration all contribute to higher signal variability at the plate perimeter. On a 96-well plate, the outer ring accounts for 36 of 96 wells. On a 384-well plate the proportion is smaller, but the effect is still measurable in sensitive assays.
The common response — placing controls in the outer columns so sample wells sit in the more stable interior — solves one problem and creates another. Controls in columns 1 and 2 of a 384-well plate are in the region most affected by the edge gradient. If the edge effect correlates along columns (which it often does, due to evaporation patterns), the controls used for normalization carry that systematic bias. Normalized data looks clean because controls and samples share the same systematic error. The bias disappears into the normalization, and no single plate looks wrong.
There is no layout that eliminates edge effects. The choice is which tradeoff to accept — and that decision should be made in full view of the assay statistics, not as a default.
Most liquid handlers dispense by column or row in a single pass. If the plate layout does not align with the dispensing direction, multi-step dispensing is required — adding time, tip changes, and potential for error.
A widely used convention for 384-well plates places controls in columns 1–2 and 23–24, with samples in columns 3–22. For 96-well plates, controls in columns 1–2 and samples in 3–12 follows the same logic. The convention is not universal — it varies by assay type and instrument configuration — but it avoids the most common dispensing conflicts and is a reasonable default for new assay development.
A plate layout only works if the analysis pipeline knows how to interpret it. This requires a formal plate map — explicit assignment of each well coordinate to a content type, compound identity, concentration, and replicate group — defined before data collection begins, not reconstructed from memory afterwards.
Common problems when this step is skipped: controls and samples confused during normalization, replicates not recognized as replicates, or well coordinate conventions differing between the robot software and the LIMS import format. Each of these is difficult to detect after the fact, particularly when the plates have already been discarded.
Plate layout decisions made before data collection begins are cheap to adjust. The same decisions revisited after three months of a screening campaign require either revalidation or a break in data comparability.
The tradeoffs — edge effects, dispensing alignment, control placement — are manageable when addressed explicitly at the start.
Design a control layout on the plate below and click Simulate to see Z′, signal-to-background, and the signal distribution for your layout. The edge effect heatmap below the plate shows which wells are most affected by evaporation and thermal gradients — adjust the sliders and watch Z′ change as the effect propagates into the control populations.
Click wells to assign types, or drag to select a rectangle. Click a row letter or column number to fill the entire row or column at once.
Parameter guide
σ neg / σ pos / σ compounds — Measurement noise for each well type as a standard deviation of the readout signal. Higher noise compresses the separation between neg and pos populations and directly reduces Z′.
Hit rate — The fraction of compound wells that are true positives. This does not affect Z′, which is calculated from the control populations only.
Cutoff (×σ) — The hit threshold is placed at μ(neg)+n×σ(neg). A factor of 3× is the standard default.
Z′ factor — Z′=1−(3σneg+3σpos)/|μpos−μneg|. Values ≥0.5 are excellent, ≥0 marginal, below 0 too narrow for reliable screening.
Edge reduction — Maximum signal loss at the outermost ring of wells. Set to 0 for an ideal plate.
Falloff radius — Distance from the plate edge at which the effect drops to zero, in wells.