calcModMoransI: calcModMoransI
In lmweber/spatzli: Methods for analyzing spatially resolved transcriptomics data

Description Usage Arguments Details Value Examples

View source: R/calcModMoransI.R

Calculate modified Moran's I statistic to rank spatially variable genes (SVGs).

calcModMoransI(
  spe,
  l_prop = 0.1,
  weights_min = 0.01,
  x_coord = "pxl_row_in_fullres",
  y_coord = "pxl_col_in_fullres",
  verbose = FALSE
)

`spe`	Input object (SpatialExperiment). Assumed to contain an assay named "logcounts" containing log-transformed normalized counts in sparse matrix format, and "spatialCoords" slot containing spatial coordinates.
`l_prop`	Value to set characteristic length parameter in squared exponential kernel used to calculate weights matrix. The characteristic length parameter is set to "l_prop" times the maximum range of the x or y coordinates. Default = 0.1.
`weights_min`	Minimum weights threshold. Weights (in the spatial covariance matrix) that are below "weights_min" times the maximum weights value are assumed to be zero.
`x_coord`	Name of column in spatialCoords slot containing x-coordinates. Default = "pxl_row_in_fullres".
`y_coord`	Name of column in spatialCoords slot containing x-coordinates. Default = "pxl_col_in_fullres".
`verbose`	Whether to print messages. Default = FALSE.

Fast implementation of modified Moran's I statistic for ranking spatially variable genes (SVGs).

We modify the definition of Moran's I statistic to make two sparsity-preserving assumptions, due to the sparse nature of spatial transcriptomics data and to speed up runtime.

- We assume that most genes are not detected in most spots (spatial coordinates), and perform calculations using only the values from the non-zero spots. For example, mean expression of each gene (which is used inside the Moran's I formula) is calculated as the mean of the non-zero spots.

- We assume that weights (in the spatial covariance matrix) below some threshold (e.g. below 1

Returns a list containing output values (one value per gene).