###################
# Clear workspace #
###################

rm(list = ls())

##################
# Load libraries #
##################

library(tidyverse)
library(melonnpan)
library(data.table)

#################
# Load datasets #
#################

metab<-fread('Metabolite training.txt')
metag<-fread('GENEFAMILY_TRAINING.txt')

###################
# Assign rownames #
###################

metab<-column_to_rownames(metab, 'SAMPLE')
metag<-column_to_rownames(metag, 'SAMPLE')

##########################
# Non-specific filtering #
##########################

abd_threshold = 1e-08
prev_threshold = 0.05
metab<- metab[, colSums(metab > abd_threshold) > nrow(metab)* prev_threshold] # Did not remove anything

###################
# Train MelonnPan #
###################

DD<-melonnpan::melonnpan.train(metab = metab, 
                               metag = metag, 
                               output = getwd()) 

############################
# Subset valid predictions #
############################

predicted_summary<-fread('./MelonnPan_Training_Summary.txt')
predicted_summary<-predicted_summary[predicted_summary$Model.Size>1,]
predicted_summary$Q.value<-p.adjust(predicted_summary$P.value)
predicted_summary<-predicted_summary[order(predicted_summary$Q.value, decreasing = FALSE),]

################################################
# Subset weight matrix and predicted compounds #
################################################

predicted_compounds<-fread('./MelonnPan_Trained_Metabolites.txt')
predicted_compounds<-column_to_rownames(predicted_compounds, 'ID')
predicted_compounds<-predicted_compounds[, predicted_summary$ID]
predicted_compounds<-rownames_to_column(predicted_compounds, 'ID')

weight_matrix<-fread('./MelonnPan_Trained_Weights.txt')
weight_matrix<-column_to_rownames(weight_matrix, 'ID')
weight_matrix<-weight_matrix[, predicted_summary$ID]
weight_matrix<-rownames_to_column(weight_matrix, 'ID')

##########################
# Replace original files #
##########################

write.table(predicted_summary, 'MelonnPan_Training_Summary_2.txt', row.names = F, col.names = T,  quote = F, sep = "\t")
write.table(predicted_compounds, 'MelonnPan_Trained_Metabolites_2.txt', row.names = F, col.names = T,  quote = F, sep = "\t")
write.table(weight_matrix, 'MelonnPan_Trained_Weights_2.txt', row.names = F, col.names = T,  quote = F, sep = "\t")