In AskExplain/GeneCodeR: Transforms between image and gene expression modalities from Spatial Transcriptomics

GeneCodeR spatial rotation to test invariance

Reload important files recently saved:

main_path <- "~/Documents/main_files/AskExplain/Q4_2022/gcode/"

# Please replace this path
path_to_save <- paste(main_path,"./temp_save_dir/",sep="")

load(file = paste(sep="",path_to_save,"all_genecoder.RData"))

Set up the test configuration for GeneCodeR

# Set up genecoder transform information
genecoder.config <- GeneCodeR::extract_config_framework(F)
genecoder.config$transform$from <- 1
genecoder.config$transform$to <- 2
genecoder.config$extract_spots$window_size <- 30

Set up validation functions to evaluate statistically significant differences via a t-test, and, cosine similarity.

# Testing functions

# cosine metric for similarity between observations

test_sample_and_genes <- function(a,b,non_zero_markers,test_type="cosine"){

  if (test_type == "t.test"){

    return(
      list(

        sample_wise = do.call('c',parallel::mclapply(c(1:dim(a)[1]),function(X){

          t.test(as.numeric(a[X,non_zero_markers[X,]]),as.numeric(b[X,non_zero_markers[X,]]))$p.value

        },mc.cores = 8)),

        gene_wise = do.call('c',parallel::mclapply(c(1:dim(a)[2]),function(X){

          t.test(as.numeric(a[non_zero_markers[,X],X]),as.numeric(b[non_zero_markers[,X],X]))$p.value

        },mc.cores = 8))

      )
    )
  } 

  if (test_type == "cosine"){
    return(
      list(

        sample_wise = do.call('c',parallel::mclapply(c(1:dim(a)[1]),function(X){

          lsa::cosine(as.numeric(a[X,non_zero_markers[X,]]),as.numeric(b[X,non_zero_markers[X,]]))

        },mc.cores = 8)),

        gene_wise = do.call('c',parallel::mclapply(c(1:dim(a)[2]),function(X){

          lsa::cosine(as.numeric(a[non_zero_markers[,X],X]),as.numeric(b[non_zero_markers[,X],X]))

        },mc.cores = 8))

      )
    )
  } 
}

Rotation testing is used to evaluate how rotated images affect the differences between the rotated image transformed to gene expression and the observed gene expression. It is expected that regardless of rotation, the gene expression should not be different across rotated images transformed to gene expression, and thus have a high cosine similarity correlation (close to 1).

# Spatial rotation testing


rotate_spot2gex <- list()
for (rotate_val in c(0,90,180,270)){

  genecoder.config$extract_spots$rotation <- rotate_val

  rotate_test_spot_data <- GeneCodeR::prepare_spot(file_path_list = test_file_path_list,meta_info_list = meta_info_list,config = genecoder.config, gex_data = test_gex_data$gex)

  rotate_spot2gex[[as.character(rotate_val)]] <- GeneCodeR::genecoder(model=genecoder.model, x = rotate_test_spot_data$spot, config = genecoder.config, model_type = "gcode")

}

count <- 0
cosine.simil_scores <- c()
for (i in c(1:4)){
  for (j in c(1:4)){
    if (i>j){
      count <- count + 1
      cosine.simil_scores[[count]] <- test_sample_and_genes(a = rotate_spot2gex[[i]],b = rotate_spot2gex[[j]], non_zero_markers = non_zero_markers, test_type = "t.test") 
    }
  }
}
rotate_spot2gex <- cosine.simil_scores

print(c("rotate cosine correlation; sample-wise",paste(c("0vs90:    ","0vs180:    ","0vs270:    ", "90vs180","90vs270","180vs270"),round(do.call('c',lapply(rotate_spot2gex,function(X){mean(X$sample_wise)})),5))))

print(c("rotate cosine correlation; gene-wise",paste(c("0vs90:    ","0vs180:    ","0vs270:    ", "90vs180","90vs270","180vs270"),round(do.call('c',lapply(rotate_spot2gex,function(X){mean(X$gene_wise)})),5))))


rm(list=ls())
gc()

AskExplain/GeneCodeR documentation built on Jan. 3, 2023, 3:52 a.m.