#source("~/git/LSHTM_analysis/config/pnca.R")
#source("~/git/LSHTM_analysis/config/alr.R")
#source("~/git/LSHTM_analysis/config/gid.R")
#source("~/git/LSHTM_analysis/config/embb.R")
#source("~/git/LSHTM_analysis/config/katg.R")
#source("~/git/LSHTM_analysis/config/rpob.R")

source("/home/tanu/git/LSHTM_analysis/my_header.R")
#########################################################
# TASK: Generate averaged stability values 
# across all stability tools
# for a given structure
#########################################################

#=======
# output
#=======
outdir_images = paste0("~/git/Writing/thesis/images/results/", tolower(gene))
outfile_mean_ens_st_aff = paste0(outdir_images, "/", tolower(gene)
                                , "_mean_ens_stability.csv")
print(paste0("Output file:", outfile_mean_ens_st_aff))

#%%===============================================================

#=============
# Input
#=============
df3_filename = paste0("/home/tanu/git/Data/", drug, "/output/", tolower(gene), "_merged_df3.csv")
df3 = read.csv(df3_filename)
length(df3$mutationinformation)

# mut_info checks
table(df3$mutation_info)
table(df3$mutation_info_orig)
table(df3$mutation_info_labels_orig)

# used in plots and analyses
table(df3$mutation_info_labels) # different, and matches dst_mode
table(df3$dst_mode)

# create column based on dst mode with different colname
table(is.na(df3$dst))
table(is.na(df3$dst_mode))

#===============
# Create column: sensitivity mapped to dst_mode
#===============
df3$sensitivity = ifelse(df3$dst_mode == 1, "R", "S")
table(df3$sensitivity)

length(unique((df3$mutationinformation)))
all_colnames = as.data.frame(colnames(df3))
common_cols  = c("mutationinformation"
                 , "position"
                 , "dst_mode"
                 #, "mutation_info_labels"
                 , "sensitivity"
                 , "ligand_distance")

all_colnames$`colnames(df3)`[grep("scaled", all_colnames$`colnames(df3)`)]
scaled_cols = c("duet_scaled"          , "duet_stability_change"
                , "deepddg_scaled"      , "deepddg"   
                , "ddg_dynamut2_scaled" , "ddg_dynamut2"
                , "foldx_scaled"        , "ddg_foldx"
                , "affinity_scaled"     , "ligand_affinity_change"
                , "mmcsm_lig_scaled"    , "mmcsm_lig"                
                , "mcsm_ppi2_scaled"   , "mcsm_ppi2_affinity"
                , "mcsm_na_scaled"     , "mcsm_na_affinity"
                #, "consurf_scaled"      , "consurf_score"
                #, "snap2_scaled"        , "snap2_score"
                #, "provean_scaled"      , "provean_score"
)
all_colnames$`colnames(df3)`[grep("outcome", all_colnames$`colnames(df3)`)]
outcome_cols_aff = c("duet_outcome"
                     , "deepddg_outcome"
                     , "ddg_dynamut2_outcome"
                     , "foldx_outcome"
                     #, "ddg_foldx", "foldx_scaled"
                     , "ligand_outcome"
                     , "mmcsm_lig_outcome"
                     , "mcsm_ppi2_outcome"
                     , "mcsm_na_outcome"
                     # consurf outcome doesn't exist
                     #,"provean_outcome"
                     #,"snap2_outcome"
)
cols_to_consider = colnames(df3)[colnames(df3)%in%c(common_cols, scaled_cols,outcome_cols)]
cols_to_extract  = cols_to_consider[cols_to_consider%in%c(common_cols, outcome_cols)]

##############################################################
#####################
# Ensemble stability
#####################
# extract outcome cols and map numeric values to the categories
# Destabilising == 0, and stabilising == 1, so rescaling can let -1 be destabilising
df3_plot = df3[, cols_to_extract]

# assign numeric values to outcome
df3_plot[, outcome_cols] <- sapply(df3_plot[, outcome_cols]
                             , function(x){ifelse(x == "Destabilising", 0, 1)})
table(df3$duet_outcome)
table(df3_plot$duet_outcome)
#=====================================
# Stability (4 cols): average the scores
# across predictors ==> average by
# position ==> scale b/w -1 and 1

# column to average: ens_stability
#=====================================
cols_to_average = which(colnames(df3_plot)%in%outcome_cols)

# ensemble average across predictors
df3_plot$ens_stability = rowMeans(df3_plot[,cols_to_average])

head(df3_plot$position); head(df3_plot$mutationinformation)
head(df3_plot$ens_stability)
table(df3_plot$ens_stability)

# ensemble average of predictors by position
mean_ens_stability_by_position <- df3_plot %>%
  dplyr::group_by(position) %>%
  dplyr::summarize(avg_ens_stability = mean(ens_stability))

# REscale b/w -1 and 1
#en_stab_min = min(mean_ens_stability_by_position['avg_ens_stability'])
#en_stab_max = max(mean_ens_stability_by_position['avg_ens_stability']) 

# scale the average stability value between -1 and 1
# mean_ens_by_position['averaged_stability3_scaled'] = lapply(mean_ens_by_position['averaged_stability3']
#                                                        , function(x) ifelse(x < 0, x/abs(en3_min), x/en3_max))

mean_ens_stability_by_position['avg_ens_stability_scaled'] = lapply(mean_ens_stability_by_position['avg_ens_stability']
                                                                  , function(x) {
                                                                    scales::rescale(x, to  = c(-1,1)
                                                                                    #, from = c(en_stab_min,en_stab_max))
                                                                                    , from = c(0,1))
                                                                  })
cat(paste0('Average stability scores:\n'
           , head(mean_ens_stability_by_position['avg_ens_stability'])
           , '\n---------------------------------------------------------------'
           , '\nAverage stability scaled scores:\n'
           , head(mean_ens_stability_by_position['avg_ens_stability_scaled'])))

# convert to a data frame
mean_ens_stability_by_position = as.data.frame(mean_ens_stability_by_position)

#FIXME: sanity checks
# TODO: predetermine the bounds
# l_bound_ens = min(mean_ens_stability_by_position['avg_ens_stability_scaled'])
# u_bound_ens = max(mean_ens_stability_by_position['avg_ens_stability_scaled'])
# 
# if ( (l_bound_ens == -1) && (u_bound_ens == 1) ){
#   cat(paste0("PASS: ensemble stability scores averaged by position and then scaled"
#         , "\nmin ensemble averaged stability: ", l_bound_ens
#         , "\nmax ensemble averaged stability: ", u_bound_ens))
# }else{
#   cat(paste0("FAIL: avergaed duet scores could not be scaled b/w -1 and 1"
#         , "\nmin ensemble averaged stability: ", l_bound_ens
#         , "\nmax ensemble averaged stability: ", u_bound_ens))
#   quit()
# } 
##################################################################
# output
#write.csv(combined_df, outfile_mean_ens_st_aff
write.csv(mean_ens_stability_by_position
          , outfile_mean_ens_st_aff
          , row.names = F)
cat("Finished writing file:\n"
    , outfile_mean_ens_st_aff
    , "\nNo. of rows:", nrow(mean_ens_stability_by_position)
    , "\nNo. of cols:", ncol(mean_ens_stability_by_position))

# end of script
#===============================================================