Identify multiomic features correlated with an distance to an image feature

Source: R/distance_based_analysis.R

distance_based_analysis: Identifies proteins/features that show a strong correlation between spatial distance (centroid based) from a specified ROI's samples and protein/feature abundance differences between samples.

distance_based_analysis(
  spe,
  assay_name,
  spotHeightCol = "spot_height",
  spotWidthCol = "spot_width",
  sampleCategoryCol,
  sampleCategoryValue,
  featuresNameCol,
  corr_type = "spearman",
  corr_thresh = 0.5,
  min_samples = 5,
  allowOverlaps = TRUE
)

Arguments

spe

SpatialExperiment object containing spatial omics data

assay_name

Name of the assay stored in spe that is to be used for distance based analysis. Example: "znormalized_log2"

spotHeightCol

Column containing height of spot

spotWidthCol

Column containing width of spot

sampleCategoryCol

Column name in metadata (colData(spe)) that should be used for selecting samples of certain type, to define the "origin" region for distance based analysis

sampleCategoryValue

Sample category to be used for defining the "origin" region for distance based analysis

featuresNameCol

Name of column containing features (example: proteins) in rowData(spe). It is assumed that the data provided in assay(spe) is in the same order as the order in which the features are listed under the featuresNameCol

corr_type

Choose from "pearson" (default), "spearman." Correlation method to be used for calculating correlation between distance between samples and protein abundance differences. Both types of correlation provide a measure of monotonic association between two variables. Pearson is better suited for linear relationships while Spearman is better for non-linear monotonic relationships.

corr_thresh

Minimum correlation value to be used for identifying proteins that have a correlation between protein abundance differences and distance between samples. Values greater than or equal to this theshold will be used.

min_samples

of a minimum number of sample points for calculating correlation. For proteins with less than this number of sample points, correlation value is reported as NA.

allowOverlaps

allow overlaps of regions

Value

a SpatialExperiment object with the distance based data in the colData

Details

Identifies proteins/features that show a strong correlation between spatial distance and protein/feature abundance differences between a specified ROI all other ROIs (Regions Of Interest) in the dataset. Spatial distance between ROIs is calculated based on their euclidean centroid distance. Features (e.g. proteins) of interest are then extracted by correlating (Pearson or Spearman correlation) feature intensities across samples/ROIs with the previously calculated pairwise distances. NOTE: No differential expression is calculated here. For spatial differential expression please use calc_spatial_diff_ex & enrich_ORA.

Examples

data(pancMeta)
data(protMeta)
data(smallPancData)
img0.spe <- convert_to_spe(smallPancData$Image_0,
  pancMeta,
  protMeta,
  sample_id = "Image0",
  spatial_coords_colnames = c("x_pixels", "y_pixels"),
  feature_meta_colname = "pancProts",
  image_files = system.file("extdata", "Image_0.png", package = "spammR"),
  image_ids = "Image0"
)
#> Note: Only mapping metadata for 2986 features out of 3000 data points

img0.spe <- distance_based_analysis(img0.spe,
  "proteomics",
  sampleCategoryCol = "IsletOrNot",
  sampleCategoryValue = "Islet"
)