Scattergram

The Scattergram engine generates interactive bivariate plots:

1. Axis selection — Choose any two measurements as X and Y axes (marker intensities, area, derived values).
2. Population display — Each cell appears as a point at its (X, Y) measurement coordinates. Color coding by phenotype, region, or other classification highlights population structure.
3. Density overlay — For large cell counts (50,000+), density contours replace individual points to prevent overplotting and reveal population density structure.
4. Gate drawing — Interactive rectangle, polygon, or quadrant gates can be drawn directly on the scattergram. Cells within the gate are classified accordingly.
5. Linked visualization — Selecting cells on the scattergram highlights their positions in the tissue image, and vice versa. This links statistical patterns to spatial locations.

Bivariate analysis (Dilbilir): A scattergram visualizes the joint distribution of two variables. While each variable's marginal distribution (histogram) shows univariate structure, the joint distribution reveals relationships: correlation, clustering, and non-linear associations. Two markers might both appear normally distributed in their individual histograms, but their scattergram reveals two distinct clusters — a bimodal joint distribution that is invisible in either marginal alone.

Colocalization metrics (Pawley): In confocal microscopy, scattergrams of two fluorescence channels are used to assess colocalization. Pearson's correlation coefficient R measures intensity correlation (-1 to +1). Manders' coefficients M1 and M2 measure the proportion of each channel that overlaps with the other. But Pawley warns: colocalization quantifies codistribution, not molecular interaction — the microscope's resolution (~200 nm) means molecules in the same voxel could be hundreds of nanometers apart. The same caution applies to tissue cytometry scattergrams: co-expression of two markers in the same cell doesn't necessarily mean the proteins interact.

The compensation analogy: In flow cytometry, spectral overlap between fluorophores produces false correlations in raw scattergrams — cells that are positive for only one marker appear to express both. Compensation (spectral unmixing) removes this artifact. In tissue cytometry, spectral unmixing serves the same purpose: the scattergram of unmixed signals shows true co-expression, while the raw scattergram may show artifactual correlation from spectral bleed-through.

Identifying immune cell subtypes in a CD3-vs-CD8 scattergram:

• Upper right cluster (CD3+CD8+): Cytotoxic T cells — high in both markers
• Left cluster (CD3+CD8−): Helper T cells — high CD3, low CD8
• Lower left cloud (CD3−CD8−): Non-T cells — negative for both
• Lower right (CD3−CD8+): Rare/artifact — NK cells or spectral contamination

Drawing quadrant gates at the valleys between clusters classifies all cells simultaneously. The linked tissue view shows where each population resides spatially — helper T cells in the stroma, cytotoxic T cells at the tumor border.