SpaNorm

SpaNorm implements spatially-aware library size normalisation for spatial transcriptomics data by concurrently modelling spatial library size effects and underlying smooth biology to remove library size effects without removing biological signals.

Key Features:

• Spatially-aware normalisation: Decomposes spatially smooth variation into components related and unrelated to library size using location-specific scaling factors.
• Gene- and cell-/spot-specific normalisation: Performs normalisation in a gene- and cell-/spot-specific manner to yield library size adjusted data.
• Platform versatility: Supports both spot-based and subcellular spatial transcriptomics data including Xenium, STOmics, CosMx, and segmented cell-level data.
• Biology retention: Outperforms single-cell RNA-seq normalisation approaches at preserving spatial domains and spatially variable genes (SVGs).

Scientific Applications:

• Spatial domain identification: Retains biological signals associated with spatial domains that would otherwise be excessively removed by single-cell normalisation methods.
• Spatially variable gene detection: Preserves spatially variable genes (SVGs) across 27 tissue samples from 6 datasets spanning 4 spatial platforms.
• Subcellular spatial transcriptomics analysis: Provides pronounced benefits for subcellular resolution data from Xenium, STOmics, and CosMx platforms where region-specific library size effects are more prevalent.

Methodology:

SpaNorm models smooth functions for each effect and performs optimal decomposition of spatially smooth variation into components related and unrelated to library size, applying location-specific scaling factors for gene- and cell-/spot-specific normalisation.