Calculate gene spatial auto-correlations (for meta cells) — rs_mc_hotspot

This function implements the HotSpot auto-correlation functionality and will return to what extent a given gene shows auto-correlation in the kNN-graph over the meta cells. For details see DeTomaso, et al. This version works on MetaCell counts which are stored in memory directly. There is no streaming variant: streaming bounds disk re-reads, which is not a problem an in-memory matrix has.

Usage

rs_mc_hotspot_autocor(
  sparse_data,
  embd,
  knn_data,
  hotspot_params,
  cells_to_keep,
  genes_to_use,
  verbose,
  seed
)

Arguments

sparse_data: A named list that needs to have data, indptr, indices, nrow, ncol and cs_type. Shape is (metacells, genes) and the data are the raw counts.
embd: Numerical matrix. The embedding matrix from which to generate the kNN graph. Needs one row per entry of cells_to_keep.
knn_data: Optional list. This contains pre-computed kNN data (indices, dist, k) and the dist_metric it was built with. The user has to ensure consistency! If provided, this will be used rather than a graph built from the parameter list.
hotspot_params: List. The HotSpot parameter list. The kNN parameters are only read when no knn_data is provided.
cells_to_keep: Integer vector. 0-index vector indicating which meta cells to include in the analysis. Ensure that this is of same order/length as the embedding matrix.
genes_to_use: Integer vector. 0-index vector indicating which genes to include.
verbose: Integer. 0L - quiet; 1L - normal verbosity; 2L - detailed verbosity.
seed: Integer. Random seed for reproducibility.

Value

A list with the following elements.

gene_idx - 0-based integer indicating the gene index.
gaerys_c - Geary's C calculation for the autocorrelation coefficient.
z_score - Z-score of the auto-correlation.
pval - P-value derived from the Z-score.
fdr - False discovery rate based on the p-value.

References

DeTomaso, et al., Cell Systems, 2021