In cytometry data analysis, a gate functions as a classification boundary, a decision rule that partitions the measurement space into populations of interest 1. From a decision-theoretic perspective, the geometry of a gate encodes specific assumptions about the statistical relationships among measured parameters 2. When these assumptions are not explicitly justified, the gate becomes an arbitrary aesthetic choice rather than a principled analytical decision. The choice of a gate's shape, size, and location depends on the analytical goal. Phenotypic analyses typically aim to characterize an entire population of interest and therefore favor inclusive gates, accepting the risk of capturing some irrelevant cells. Sorting gates, by contrast, prioritize specificity over sensitivity, knowingly excluding events at a population's periphery. Similarly, when analyzing antigen-specific responses via cytokine production, gates are often set conservatively to capture only confidently cytokine-positive events, with the understanding that cytokine-low events will be excluded. In practice, there is often no sharp boundary between “positive” and “negative” populations. Marker expression may span a continuum, and the distributions of label intensities that allow us to identify the markers may include outliers that encroach on neighboring populations. Gating, therefore, inevitably involves judgment, and proficiency testing has demonstrated that this subjectivity can lead to substantial inter-analyst variability 3, 4. When a gate is drawn on a bivariate plot, the analyst is not merely isolating a population but making an explicit claim about which cells count and which do not. Gate shape is therefore not a neutral or aesthetic choice; it encodes assumptions about how the measured markers relate to one another. When two markers are statistically independent (such that expression of one provides no information about expression of the other), the appropriate gate is rectangular. In this case, gating is equivalent to posing two independent yes/no questions: “Is this cell positive for marker X?” and “Is this cell positive for marker Y?”. Together, the independent thresholds define a rectangle in two-dimensional space. Even when fluorescence intensities are better treated as count data (e.g., Poisson or negative binomial), a principled gate can be understood as an acceptance region that retains a chosen fraction of events (e.g., 99%) under a background model. If X and Y are independent and the goal is per-marker thresholding, rectangular gates follow naturally. If instead the goal is a joint percentile boundary (an isoprobability region), the contour is typically “ellipse-like,” and one could estimate the population center and scatter and use a Mahalanobis-distance cutoff to draw an ellipse, which is the exact solution under a Gaussian model and a practical approximation for count data. By contrast, drawing a polygonal, or freehand gate, introduces a more complex assumption: that the two markers are meaningfully connected. Such a gate encodes a joint decision rule, implicitly modeling a bivariate distribution and specifying that cells with a given expression of marker X should fall within a particular range of marker Y expression or vice versa. The analyst is no longer applying independent criteria, but a coupled classification whose logic is embedded in the gate shape itself. This assumption is sometimes justified. Markers may covary because of real underlying biology, such as shared signaling pathways or co-regulation during differentiation. Apparent correlations may also arise from technical sources, including imperfect compensation or spectral unmixing, in which residual structure remains even after correction for spectral overlap. In addition, commonly used data transformations—such as biexponential and hyperbolic arcsine scaling—can distort population geometry if used with improper parameters, making otherwise uncorrelated populations appear tilted, curved, or compressed 5. Non-rectangular gates may also be required when suboptimal reagents are used, such as dim fluorochromes or inappropriately low antibody titers. While such gates may be necessary to salvage existing data, the appropriate remedy for future experiments is to optimize the panel rather than rely on complex gate geometry. Problems arise when non-rectangular gates are drawn without a clear understanding of why the markers appear correlated. This reflects visual pattern-matching rather than statistically or biologically motivated reasoning. The consequences are not trivial: valid cells may be excluded without justification, inappropriate events may be included, and results may be difficult or impossible for other cytometrists to reproduce, since the precise shape of a freehand gate is inherently subjective. Rectangular gates offer a useful default precisely because they are simple and encode minimal assumptions, specifically, that the two markers can be evaluated independently. In other words, they represent two independent decisions—one for each marker—and can be fully specified by reporting threshold values, making them straightforward to understand, justify, and reproduce. Non-rectangular gates are not inherently incorrect, but they encode more complex assumptions and therefore require explicit explanation. For OMIP and other methodological manuscripts, this distinction is particularly important. Non-rectangular gates are acceptable, but the assumptions underlying their shape must be clearly stated and justified. A related issue concerns the placement of lower boundaries for negative populations. This question is more nuanced than that of upper thresholds. If a population is truly negative for a marker, its measured signal should consist only of background autofluorescence and measurement noise, and there is no inherent reason to exclude events simply because they fall at very low intensity values. A clearly negative cell should not be excluded merely because it is very negative. There are, however, legitimate reasons to impose a different lower boundary. Although light-scatter-based gating mitigates this, events with very low fluorescence values may include debris, non-cellular material, or electronic noise. Compensation or spectral unmixing may also produce mathematically negative values when more signal is subtracted than was present, a consequence of noise and imperfect mixing model specification. Finally, measurements near the instrument's noise floor may be unreliable. Even in these cases, the same principle applies: any boundary that does not extend to the measurement limits represents an additional analytical decision and must be justified. For negative populations, extending the gate to the lower axis should be the default. In practice, rectangular gates that extend to the axes provide a transparent and reproducible default. They make assumptions explicit, limit subjectivity, and allow others to understand exactly how populations were defined. When a different gate shape is used, it should reflect a deliberate analytical choice grounded in biology, instrument behavior, or data processing—not convenience or visual appeal. Making those assumptions explicit is essential for interpretability and reproducibility. Yolanda D. Mahnke: writing – review and editing, conceptualization. Bartek Rajwa: writing – review and editing, conceptualization, writing – original draft. The authors declare no conflicts of interest. Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
Mahnke et al. (Fri,) studied this question.