iscc.data.VisiumBatchHyperParams dataclass

VisiumBatchHyperParams(protocol: str = 'visium', spot_pitch: float = 2.0, spot_radius: float = 1.0, mu_counts: float = 20000.0, sigma_counts: float = 0.45, field_lengthscale: float = 18.0, field_sigma: float = 0.7, edge_sigma: float = 0.3, diffusion_sigma: float = 0.0, sigma_batch: float = 0.1, ambient_frac: float = 0.05, count_model: str = 'dm', kappa: float = 50.0, nb_dispersion: float = 0.3)

The Visium assay's technical parameters — a defaults container.

Holds the protocol-typical magnitudes of the Visium spatial count model: the spot geometry (pitch, radius), the per-spot library size, the spatially-correlated capture-efficiency field (lengthscale, strength, tissue-edge falloff), lateral mRNA diffusion, and the shared per-gene batch factor / ambient soup / pluggable count model. The Visium assay builds one and lets you override any field, and estimate_visium fits these same parameters from real data.

Attributes:

Name	Type	Description
protocol	str	Protocol preset the magnitudes come from (e.g. "visium").
spot_pitch	float	Spot centre-to-centre spacing (sets spot density / spots-per-tissue).
spot_radius	float	Spot capture radius (sets spot->cell aggregation, ~1-10 cells/spot).
mu_counts	float	Mean per-spot library size (total UMIs / spot).
sigma_counts	float	Per-spot library-size LogNormal sd.
field_lengthscale	float	Capture-field spatial autocorrelation scale (the SE-GP length-scale); larger -> smoother field -> higher Moran's I.
field_sigma	float	Capture-field strength (log-space sd of the smooth field).
edge_sigma	float	Tissue-boundary capture-efficiency falloff (0 = none; <1 keeps it positive).
diffusion_sigma	float	Lateral mRNA bleed: Gaussian kernel sd spreading expression to neighbours.
sigma_batch	float	Per-gene batch-factor LogNormal sd.
ambient_frac	float	Fraction of the spot library that is ambient ("soup") contamination.
count_model	str	Count emission {"dm" (default, compositional capture), "nb"}.
kappa	float	Dirichlet-Multinomial concentration (only for count_model="dm").
nb_dispersion	float	NB overdispersion phi (var = mu + phi*mu^2; only for count_model="nb").