LSHTM_analysis/scripts/plotting/LINEAGE2.R

198 lines
No EOL
6.6 KiB
R

library(tidyverse)
#install.packages("ggforce")
library("ggforce")
#install.packages("gginference")
library(gginference)
library(ggpubr)
##################################################
#%% read data
# TODO: read data using gene and drug combination
# gene must be lowercase
# tolower(gene)
#################################################
df = read.csv("/home/tanu/git/Data/pyrazinamide/output/pnca_merged_df2.csv")
#df2 = read.csv("/home/tanu/git/Data/pyrazinamide/output/pnca_merged_df3.csv")
foo = as.data.frame(colnames(df))
cols_to_subset = c('mutationinformation'
, 'snp_frequency'
, 'pos_count'
, 'position'
, 'lineage'
, 'lineage_multimode'
, 'dst'
, 'dst_multimode'
#, 'dst_multimode_all'
, 'dst_mode')
my_df = df[ ,cols_to_subset]
#df2 = df2[ ,cols_to_subset]
r24p_embb = df_embb[df_embb$mutationinformation == "R24P",]
tm = c("A102P", "M1T")
test = my_df[my_df$mutationinformation%in%tm,]
#test$dst2[is.na(test$dst)] <-test$dst_mode
test$dst2 = ifelse(is.na(test$dst), test$dst_mode, test$dst)
sum(table(test$dst2)) == nrow(test)
# Now we need to make a column that fill na in dst with value of dst_mode
my_df$dst2 = ifelse(is.na(my_df$dst), my_df$dst_mode, my_df$dst)
#%% create sensitivity column ~ dst_mode
my_df$sensitivity = ifelse(my_df$dst2 == 1, "R", "S")
table(my_df$dst2)
if ( table(my_df$sensitivity)[2] == table(my_df$dst2)[1] && table(my_df$sensitivity)[1] == table(my_df$dst2)[2] ){
cat("\nProceeding with lineage resistance plots")
}else{
stop("FAIL: could not verify dst2 and sensitivity numbers")
}
#%%
# select only L1-L4 and LBOV
sel_lineages1 = c("LBOV", "")
my_df2 = my_df[!my_df$lineage%in%sel_lineages1,]
table(my_df2$lineage)
sel_lineages2 = c("L1", "L2", "L3", "L4")
my_df2 = my_df2[my_df2$lineage%in%sel_lineages2,]
table(my_df2$lineage)
sum(table(my_df2$lineage)) == nrow(my_df2)
table(my_df2$lineage)
# %%
# str(my_df2)
# my_df2$lineage = as.factor(my_df2$lineage)
# my_df2$sensitivity = as.factor(my_df2$sensitivity)
#%% get only muts which belong to > 1 lineage and have different sensitivity classifications
muts = unique(my_df2$mutationinformation)
#-----------------------------------------------
# step1 : get muts with more than one lineage
#-----------------------------------------------
lin_muts = NULL
for (i in muts) {
print (i)
s_mut = my_df2[my_df2$mutationinformation == i,]
s_tab = table(s_mut$lineage, s_mut$sensitivity)
#print(s_tab)
if (dim(s_tab)[1] > 1 && dim(s_tab)[2] > 1){
lin_muts = c(lin_muts, i)
}
}
cat("\nGot:", length(lin_muts), "mutations belonging to >1 lineage with differing drug sensitivities")
#-----------------------------------------------
# step 2: subset these muts for plotting
#-----------------------------------------------
plot_df = my_df2[my_df2$mutationinformation%in%lin_muts,]
cat("\nnrow of plot_df:", nrow(plot_df))
#-----------------------------------------------
# step 3: Add p-value
#-----------------------------------------------
plot_df$pval = NULL
for (i in lin_muts) {
print (i)
s_mut = plot_df[plot_df$mutationinformation == i,]
print(s_mut)
s_tab = table(s_mut$lineage, s_mut$sensitivity)
print(s_tab)
ft_pvalue_i = round(fisher.test(s_tab)$p.value, 3)
print(ft_pvalue_i)
plot_df$pval[plot_df$mutationinformation == i] <- ft_pvalue_i
#print(s_tab)
}
head(plot_df$pval)
# format p value
# TODO: add case statement for correct pvalue formatting
plot_df$pvalF = ifelse(plot_df$pval < 0.05, paste0(plot_df$pval, "*"), plot_df$pval )
plot_df$pvalF
#================================================
# Plot attempt 1 [no stats]: WORKS beeautifully
#================================================
ggplot(plot_df, aes(x = lineage
, fill = factor(sensitivity))) +
geom_bar(stat = 'count')+
#coord_cartesian(ylim = c(0, ypos_label)) +
facet_wrap(~mutationinformation
, scales = 'free_y')
#########################################################
#================================================
# Plot attempt 2 [with stats]:data wrangling to
# get ypos_label to place stats with geom_label
#================================================
# # small data set
# tm3 = c("F94L", "A102P", "L4S", "L4W")
# tm2 = c("L4W")
#
# # Calculate stats: example
# test2 = plot_df[plot_df$mutationinformation%in%tm2,]
# table(test2$mutationinformation, test2$lineage, test2$sensitivity)
# tm_tab = table(test2$lineage, test2$sensitivity)
# tm_tab
# Get the ypos for plotting the label for p-value
lin_muts_tb = plot_df %>%
group_by(mutationinformation) %>%
count(lineage) %>%
mutate(ypos_label = max(n))
head(lin_muts_tb); class(lin_muts_tb)
lin_muts_df = as.data.frame(lin_muts_tb)
class(lin_muts_df)
intersect(names(plot_df), names(lin_muts_df))
select_cols = c("mutationinformation", "ypos_label")
lin_muts_df2 = lin_muts_df[, select_cols]
names(lin_muts_df2) ; head(lin_muts_df2)
# remove duplicates before merging
lin_muts_df2U = lin_muts_df2[!duplicated(lin_muts_df2),]
class(lin_muts_df2); class(plot_df); class(lin_muts_df2U)
lin_muts_dfM = merge(plot_df, lin_muts_df2U, by = "mutationinformation", all.y = T)
if (nrow(lin_muts_dfM) == nrow(plot_df) ){
cat("\nPASS: plot_df now has ypos for label"
, "\nGenerating plot_df2 with sensitivity as factor\n")
str(lin_muts_dfM)
lin_muts_dfM$sensitivity = as.factor(lin_muts_dfM$sensitivity)
plot_df2 = lin_muts_dfM
}else{
stop("\nSomething went wrong. ypos_label could not be generated")
}
#================================================
# Plot: with stats (plot_df2)
# TODO:
#1) Add gene name from variable as plot title. <Placeholder provided>
#2) Add: facet_wrap_paginate () to allow graphs to span over multiple pages
#3) Add *: Extend yaxis for each plot to allow geom_label to have space (or see
# if this self resolving with facet_wrap_paginate())
#================================================
p_title = "<Insert gene>"
ggplot(plot_df2, aes(x = lineage
, fill = sensitivity)) +
geom_bar(stat = 'count') +
facet_wrap(~mutationinformation
, scales = 'free_y') +
theme(legend.position = "top")+
labs(title = p_title) +
#geom_text(aes(label = p.value, x = 0.5, y = 5))
geom_label(aes(label = paste0("p=",pvalF), x = 2.5, ypos_label+1), fill="white")# +
#geom_text(aes(label = paste0("p=",pvalF), x = 2.5, ypos_label+1))# +
#geom_segment(aes(x = 1, y = ypos_label+0.5, xend = 4, yend = ypos_label+0.5))
#geom_hline(data = lin_muts_dfM, aes(yintercept=ypos_label+0.5))
#geom_bracket(data=lin_muts_dfM, aes(xmin = 1, xmax = 4, y.position = ypos_label+0.5, label=''))