Skip to contents

Object creation From Raw Data

Object_Creation.Rmd
Starting with a spatial seurat object, we can make a niche-DE object with the function ‘CreateNicheDEObjectFrom Seurat’. This function takes in 6 arguments
Arguments
  • counts mat: A counts matrix. The dimension should be #cells/spots by #genes.
    • coordinate mat: Coordinate matrix. Dimension should be #cells/spots by 2. Make sure that rownames match that of the counts matrix. The coordinates are scaled such that the median nearest neighbor distance is 100.
    • library mat: The average expression profile matrix calculated from a reference dataset.
    • deconv mat: The deconvolution matrix for the spatial dataset. The dimension should be #spots/cells by #cell types
    • sigma: A list of kernel bandwidths to use for effective niche calculation
    • Int: Logical if counts data supplied is integer. When performing niche-DE, Negative binomial regression is performed if True. Linear regression with a gene specific variance is performed if False. Default is true.
#load counts matrix
data("vignette_counts")
#load coordinate matrix
data("vignette_coord")
#load expression profile matrix
data("vignette_library_matrix")
#load deconvolution matrix
data("vignette_deconv_mat")
#make Niche-DE object
NDE_obj = CreateNicheDEObject(vignette_counts,vignette_coord,
                              vignette_library_matrix,vignette_deconv_mat,
                              sigma = c(1,100,250))
What’s in the object?

We see that their are 14 slots, 10 of which are populated when making the nicheDE object. Here we will explain what each slot should contain, except for the ones prefixed by niche_DE.

  • counts: The RNA count data of the spatial transcriptomics dataset. The dimension will be #cells/spots by #genes.Genes are filtered out if they do not exist within the scrna-seq reference dataset.
  • coord: The spatial coordinates matrix of the spatial transcriptomics dataset. It is scaled such that the median nearest neighbor distance is 100.
  • sigma: The kernel bandwidth(s) chosen for calculating the effective niche. Recommended values will be discussed shortly.
  • num_cells: A #cells/spots by #cell types matrix indicating the estimated number of cells of each cell type in each spot.
  • effective_niche: A list whose length is equal to the length of sigma. Each element of the list is a matrix of dimension #cells/spots by #cell types that measures how many of each cell type is in a given cell/spot’s neighborhood. For more information, please read the manuscript.
  • ref_expr: The average expression profile matrix. The dimension is #cell types by #genes. Each row gives the average expression of each gene for a given cell type based on the reference dataset supplied.
  • cell_names: The name of each cell. This will be used if the use wants to filter cells via the function ‘Filter’
  • gene_names: The gene names.
  • batch_ID: The batch ID for each cell/spot. This will be used when merging objects.
  • spot_distance: 100. Coordinates are scaled so that the distance between a spot and its nearest neighbor is 100.
  • scale: The scaling to convert our coordinates back to their original scale
  • Int: Is the data count data(True) or continuous (False)?