Correlation, network and heatmap plots

Load R packages

library(tidyverse)
library(readxl)
library(stringr)
library(psych)
library(gdata)
library(qgraph)
library(kableExtra)
library(ComplexHeatmap)
library(circlize)

Load data from previous steps

load(file = paste("./saved_data/time_span_of_interest.RData", sep=""))
time_span_of_interest

## [1] "timeinterval_cde"

Step 1

if (time_span_of_interest=="timepoint_b"){
  fn ="001_read_and_format_data_and_stat_testing_tpb.RData"
}else if(time_span_of_interest=="timeinterval_cde"){
  fn ="001_read_and_format_data_and_stat_testing_ticde.RData"
}

loaded_objects_from_step_1 = load(file = paste("./saved_data/",fn, sep=""))
loaded_objects_from_step_1

##  [1] "cyts0.0"                    "gfs0.0"                    
##  [3] "clin_outs"                  "group_counts"              
##  [5] "group_counts_0"             "group_counts_1"            
##  [7] "corr_lst.0.0"               "corr_lst_clin_group_0.rval"
##  [9] "corr_lst_clin_group_0.pval" "corr_lst_clin_group_0.n"   
## [11] "corr_lst_clin_group_1.rval" "corr_lst_clin_group_1.pval"
## [13] "corr_lst_clin_group_1.n"    "paired_corr.lst.zval"      
## [15] "paired_corr.lst.pval"

Step 2

if (time_span_of_interest=="timepoint_b"){
  fn ="002_format_results_of_statistical_testing_tpb.RData"
}else if(time_span_of_interest=="timeinterval_cde"){
  fn ="002_format_results_of_statistical_testing_ticde.RData"
}
loaded_objects_from_step_2 = load(file = paste("./saved_data/",fn, sep=""))
loaded_objects_from_step_2

##  [1] "corr_lst.0.0"               "gfs0.0"                    
##  [3] "sign_corr_mats.lst"         "sign_corr_lst_clin_group_0"
##  [5] "sign_corr_lst_clin_group_1" "sign_cyts_and_gfs.lst"     
##  [7] "sign_cyts_and_gfs.lst_0"    "sign_cyts_and_gfs.lst_1"   
##  [9] "cor_diff.lst"               "cor_diff_groups.lst"       
## [11] "ranked_peps_tb.0.0"         "ranked_z_tb.0.0"

Heatmap

small_mat = as.matrix(ranked_z_tb.0.0)
#ranked_z_tb.0.0 %>% print(.,n = 50)
small_mat_text = as.matrix(ranked_peps_tb.0.0)
max <- max(small_mat)
min <- min(small_mat)
col_fun = colorRamp2(c(min,0.01, max), c("lightgrey", "green", "red"))
Heatmap(small_mat, name = "sum of z-score", col = col_fun,
show_column_dend = FALSE, show_row_dend = FALSE,
#width = unit(15, "cm"), height = unit(15, "cm"),
    cell_fun = function(j, i, x, y, width, height, fill) {
    if(small_mat[i, j] > 0)
        grid.text(sprintf("%s", small_mat_text[i, j]), x, y, gp = gpar(fontsize = 7))#%.1f
})

The color in each cell of this heatmap indicates the level at which a given cytokine or growth factor changes its interaction with other cytokines and growth factors. The color (and intesity) signifies the sum of the significant and standardized z-scores (mesures of change of correlation) for individual cytokines and growth factors. Each cytokine/growthfactor name found cell is trailed by an underscore and a number. The number indicates how many other cytokines it interacts with so that weather the sum is due to a few big interactions or many small. Cytokines with no significant change in correlation between children with and without clinical outcomeis excluded (grey)

Correlation plots with difference in correlation

For each plot triplet per row there is one plot for group 0 , one for group 1 and one for the difference between 0 and 1

for (i in 1:length(corr_lst_clin_group_0.rval)){
   corrplot::corrplot(corr_lst_clin_group_0.rval[[i]],
                      tl.pos="td",
                      tl.col="darkgrey",
                      mar=c(0,0,1,0),
                      number.cex=3,
                      tl.cex=1,  
                      type="upper",
                      pch.cex=0.8, 
                      diag=F,
                      cl.pos="n",
                      addgrid.col=NA,
                      title=paste(group_counts_0$clin_grouping[i],",0, numb samples =  ",corr_lst_clin_group_0.n[[i]],sep=""))
  
    corrplot::corrplot(corr_lst_clin_group_1.rval[[i]],
                      tl.pos="td",
                      tl.col="darkgrey",
                      mar=c(0,0,1,0),
                      number.cex=3,
                      tl.cex=1,  
                      type="upper",
                      pch.cex=0.8, 
                      diag=F,
                      cl.pos="n",
                      addgrid.col=NA,
                      title=paste(group_counts_1$clin_grouping[i],",1, numb samples =  ",corr_lst_clin_group_1.n[[i]],sep=""))
    
    #plot one plot fo rchange in correllation
    A <- paired_corr.lst.zval[[i]] 
    #Normalized Data
    normalized = (A-min(A,na.rm=T))/(max(A,na.rm=T)-min(A,na.rm=T))
    B <- paired_corr.lst.pval[[i]]
    sig_element = ifelse(B < 0.05, normalized, NA)
    sig_element[is.na(sig_element)] <- 0 
    
    corrplot::corrplot(sig_element,
                       tl.pos="td",
                       tl.col="darkgrey", 
                       mar=c(0,0,1,0),
                       number.cex=3,tl.cex=0.8,
                       type="upper",
                       pch.cex=0.8, diag=F,cl.pos="n")
        
      
      # Why am I not using cor_diff.lst here?
      # 
      # ,cor_diff.lst[[i]]
      #                 tl.pos="td",
      #                 tl.col="darkgrey",
      #                 mar=c(0,0,1,0),
      #                 number.cex=3,
      #                 tl.cex=1,  
      #                 type="upper",
      #                 pch.cex=0.8, 
      #                 diag=F,
      #                 cl.pos="n",
      #                 addgrid.col=NA,
      #                 title=paste(group_counts_1$clin_grouping[i],",1vs2",sep="")
      # 

                 
}

$Some caption$

Some caption

Network plots with difference in correlation