×
Explore how Vaccinia virus (VacV) infection affects protein patterns and hence shifts cells in the Swin-tuned PATTERNS embedding space. You can select the gene to inspect how it was affected, or use the table with a ranking of most affected.
In Color cells by: Batch mode — Red: Infected batch (VacV, replicates B4, B5). Blue: Healthy batch (LiveDye, replicates C2, C3, B4, B5).
In Color cells by: Barcode mode — each barcode gets a paired colour (darker for infected, lighter for healthy LiveDye).
Grey background - initial-batch PATTERNS cells (reference only).
The genes in the table are ranked using a gated displacement score. Namely,
Cells are assigned to a precomputed k=400 k-means cluster (fitted on the PATTERNS data). Each cluster carries a soft compartment mixture from the Combined annotations. Healthy localisations = top compartments of the full Healthy batch, while Infected localisations = top compartments of the shifted cells, i.e., where the moved population went.
Random cells from the visible area of the UMAP can be previewed separately for each batch. Note that up to 9 cells are drawn and they can be shuffled with a click on the button.
Vaccinia Virus Infection Impact
Batches
Ranking
dispOverSc = ||μ_c − μ_b|| / σ_c (displacement of the infected cloud relative to its own spread, on L2-normalised embeddings), multiplied by a participation gate that keeps only shifts carried by enough cells: gate = min(1, shifted_count / N_0) * min(1, shifted_frac / R_MIN), where a c cell's soft "shifted" membership is a sigmoid of its distance to μ_b vs the healthy Q-percentile radius. The count floor rejects few-cell drift; the fraction floor rejects shifts where the bulk of c stays on healthy and only a small slice moved.Locations (descriptive)
Cell preview
The viewer is a result of a collaboration between the JSB Lab of the University of Bonn and Appsilon.
VacV-infected and healthy cells on PATTERNS (Swin-tuned) UMAP
Filter cells by gene