Edge Effects in Plate-Based Assays#
Colonies on the outer rows and columns of an agar plate consistently grow differently from those in the interior. This systematic bias — the “edge effect” — can confound quantitative comparisons between strains or conditions.
Causes#
Temperature gradients — the plate edge loses heat faster than the center, creating a radial temperature gradient that affects growth rate
Humidity gradients — agar near the plate edge dries faster, concentrating nutrients and changing water activity
Gas exchange — outer wells have greater access to atmospheric oxygen
Nutrient diffusion — edge colonies have fewer neighbors competing for nutrients
The magnitude of the edge effect depends on incubator design, plate type, and organism growth rate. It is typically strongest in 384-well high-density plates.
Detection#
Edge-affected colonies are usually larger or more intensely pigmented than interior colonies of the same strain. This is visible in the raw measurements as a ring of elevated values around the plate perimeter.
PhenoTypic’s EdgeCorrector identifies edge-affected wells using a
permutation test against interior wells at matching grid positions.
Correction#
EdgeCorrector operates on measurement DataFrames (not images). It:
Identifies interior colonies (not on the outer rows/columns)
Selects the top-n interior values as a reference baseline
Tests each edge colony against the interior distribution
Adjusts values that are statistically elevated
from phenotypic.analysis import EdgeCorrector
corrector = EdgeCorrector(
on="Area",
groupby=["Strain"],
nrows=8, ncols=12,
top_n=3, pvalue=0.05,
)
corrected_df = corrector.correct(measurements_df)
Experimental Mitigation#
Statistical correction is a post-hoc remedy. For critical experiments, consider:
Randomized plate layout — distribute strains randomly across the plate so edge effects average out across conditions
Border wells as controls — inoculate border wells with a reference strain and exclude them from analysis
Humidified incubation — reduces evaporation gradients
Multiple replicates — average over several plates with different strain positions