HepG2_GAT-analysis.R

# This script will read blocks of GAT shuffling results (in whitelisted genome (WL), open chromatin (DHS), OQS and 'open' OQS) combine results with ENCODE meta data and print complete and trimmed spreadsheets of the combined results

# ======= Libraries and working directory
 library(data.table)
 library(dplyr)
 library(ggplot2)
 library(ggsignif)
 library(ggpubr)
 library(scales)

setwd('HepG2_GAT/')


# ==== Script parameters
plotdata <- 1
savetables <- 1


# ===  list of segments that were shuffled in different backgrounds
GAT_segments <- c('rep1-3') 

# ====== Parameters for Filtering the data ====
qValue_cutoff <-  1 ## (this currently not used)
PEAK_NUMBER_Cutoff <- 100 # --- filtering by peak number should be covered by too high q-values

Explicit_exclusion <- c() # Datasets have been flagged in original paper for K562 (Cell 2019 (10.1016/j.cell.2019.06.001). Not a problem in HepG2


# ====== Load meta data
ENCODE_bed_meta <- read.table(file = "HepG2/MetaData/ENCODE_HepG2_Mar2020_BED_Meta.tsv", sep = '\t', header = TRUE)
ENCODE_bed_meta$Experiment.target <- gsub("-human", "", ENCODE_bed_meta$Experiment.target)
ENCODE_bed_meta <- ENCODE_bed_meta[rowSums(is.na(ENCODE_bed_meta))  != ncol(ENCODE_bed_meta), ] # Remove empty rows


#read in peak numbers
Peak_Number <- read.table(file="HepG2/GAT_analysis/HepG2_NumberOfPeaks.txt", sep= '\t', header = FALSE)
colnames(Peak_Number)[1] <- "ENCODE_ID"
colnames(Peak_Number)[2] <- "Peaks"
Peak_Number$ENCODE_ID <- sapply(as.vector(Peak_Number$ENCODE_ID), function(x) {gsub("\\..*","",x)}) # strip extension
ENCODE_bed_meta$Peaks <- Peak_Number$Peaks[match(ENCODE_bed_meta$File.accession, Peak_Number$ENCODE_ID)]


# =====================================================================================================================================================
# == Cycle Through blocks of GAT results as defined in GAT_segments
# =====================================================================================================================================================
for (i in 1:length(GAT_segments))
{
    Gat_Seg <- (GAT_segments)[i]


# =====================================================================================================================================================
# Load result tables for four GAT runs using different workplaces (Whitelisted genome, DNAse-hypersensitivity sites, Observed Quadruplexes, OQs at DHS)
# =====================================================================================================================================================

# ====== shuffle in whitelisted genome  
Gat_WL <- read.table(file =paste("HepG2/GAT_output/GAT_HepG2_", Gat_Seg, ".WL.tsv",  sep="") , sep = '\t', header = TRUE) 
Gat_WL <- Gat_WL[, 2:11]
colnames(Gat_WL)[1] <- "ENCODE_ID"
Gat_WL$ENCODE_ID <- gsub(".cut.bed.gz", "", Gat_WL$ENCODE_ID)
###Sort by rank and include and extra column 'rank' to see how high the marker scored
Gat_WL <- Gat_WL[order(Gat_WL$fold, decreasing = TRUE), ]
Gat_WL$rank <- seq.int(nrow(Gat_WL))


# ===== shuffle in  DHS (open chromatin)
Gat_DHS <- read.table(file = paste("HepG2/GAT_output/GAT_HepG2_", Gat_Seg, ".DHS.tsv",  sep=""), sep = '\t', header = TRUE)
Gat_DHS <- Gat_DHS[, 2:11]
colnames(Gat_DHS)[1] <- "ENCODE_ID"
Gat_DHS$ENCODE_ID <- gsub(".cut.bed.gz", "", Gat_DHS$ENCODE_ID)
###Sort by rank and include and extra column 'rank' to see how high the marker scored
Gat_DHS <- Gat_DHS[order(Gat_DHS$fold, decreasing = TRUE), ]
Gat_DHS$rank <- seq.int(nrow(Gat_DHS))


# ========  shuffle in observed Quadruplexes
Gat_OQs <- read.table(file = paste("HepG2/GAT_output/GAT_HepG2_", Gat_Seg, ".OQS.tsv",  sep=""), sep = '\t', header = TRUE)
Gat_OQs <- Gat_OQs[, 2:11]
colnames(Gat_OQs)[1] <- "ENCODE_ID"
Gat_OQs$ENCODE_ID <- gsub(".cut.bed.gz", "", Gat_OQs$ENCODE_ID)
###Sort by rank and include and extra column 'rank' to see how high the marker scored
Gat_OQs <- Gat_OQs[order(Gat_OQs$fold, decreasing = TRUE), ]
Gat_OQs$rank <- seq.int(nrow(Gat_OQs))


# ==== shuffle in observed Quadruplexes at DHS
Gat_openOQs <- read.table(file = paste("HepG2/GAT_output/GAT_HepG2_", Gat_Seg, ".opOQS.tsv",  sep=""), sep = '\t', header = TRUE)
Gat_openOQs <- Gat_openOQs[, 2:11]
colnames(Gat_openOQs)[1] <- "ENCODE_ID"
Gat_openOQs$ENCODE_ID <- gsub(".cut.bed.gz", "", Gat_openOQs$ENCODE_ID)
###Sort by rank and include and extra column 'rank' to see how high the marker scored
Gat_openOQs <- Gat_openOQs[order(Gat_openOQs$fold, decreasing = TRUE), ]
Gat_openOQs$rank <- seq.int(nrow(Gat_openOQs))

# ============================================================================================================================================================
# === Generate a Merged data sheet comprising ENCODE meta data and GAT analysis for the different runs ====
# ============================================================================================================================================================

Merged_all <- ENCODE_bed_meta

#whitelisted genome
Merged_all$WL_observed <- Gat_WL$observed[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_expected <- Gat_WL$expected[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_CI95low <- Gat_WL$CI95low[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_CI95high <- Gat_WL$CI95high[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_stddev <- Gat_WL$stddev[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_fold <- Gat_WL$fold[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_l2fold <- Gat_WL$l2fold[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_pvalue <- Gat_WL$pvalue[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_qvalue <- Gat_WL$qvalue[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]
Merged_all$WL_rank <- Gat_WL$rank[match(Merged_all$File.accession, Gat_WL$ENCODE_ID)]

#DHS (open chromatin)
Merged_all$DHS_observed <- Gat_DHS$observed[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_expected <- Gat_DHS$expected[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_CI95low <- Gat_DHS$CI95low[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_CI95high <- Gat_DHS$CI95high[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_stddev <- Gat_DHS$stddev[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_fold <- Gat_DHS$fold[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_l2fold <- Gat_DHS$l2fold[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_pvalue <- Gat_DHS$pvalue[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_qvalue <- Gat_DHS$qvalue[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]
Merged_all$DHS_rank <- Gat_DHS$rank[match(Merged_all$File.accession, Gat_DHS$ENCODE_ID)]

#observed Quadruplexes
Merged_all$OQs_observed <- Gat_OQs$observed[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_expected <- Gat_OQs$expected[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_CI95low <- Gat_OQs$CI95low[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_CI95high <- Gat_OQs$CI95high[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_stddev <- Gat_OQs$stddev[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_fold <- Gat_OQs$fold[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_l2fold <- Gat_OQs$l2fold[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_pvalue <- Gat_OQs$pvalue[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_qvalue <- Gat_OQs$qvalue[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]
Merged_all$OQs_rank <- Gat_OQs$rank[match(Merged_all$File.accession, Gat_OQs$ENCODE_ID)]

#observed Quadruplexes at DHS
Merged_all$opOQs_observed <- Gat_openOQs$observed[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_expected <- Gat_openOQs$expected[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_CI95low <- Gat_openOQs$CI95low[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_CI95high <- Gat_openOQs$CI95high[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_stddev <- Gat_openOQs$stddev[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_fold <- Gat_openOQs$fold[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_l2fold <- Gat_openOQs$l2fold[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_pvalue <- Gat_openOQs$pvalue[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_qvalue <- Gat_openOQs$qvalue[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]
Merged_all$opOQs_rank <- Gat_openOQs$rank[match(Merged_all$File.accession, Gat_openOQs$ENCODE_ID)]

Merged_all <- Merged_all[rowSums(is.na(Merged_all)) != ncol(Merged_all), ] # remove empty rows 

# export all data unfiltered
if (savetables)
{
  write.csv(Merged_all, file =paste("HepG2/GAT_analysis/GAT_HepG2_", Gat_Seg, "_Gat-Analysis_all.csv",  sep=""))
}


#====== FILTER  q-values 
# keep cases where shuffling was successfull in case of DHS as this will be the reference and at least 100 peaks to remove low quality ChIP-seq maps
GAT_FILTERED <- Merged_all[ Merged_all$DHS_qvalue < qValue_cutoff , ]
GAT_FILTERED <- Merged_all[Merged_all$Peaks > PEAK_NUMBER_Cutoff, ]

#Remove data sets that have been explicitly flagged in publications, but not yet by ENCODE
GAT_FILTERED <- GAT_FILTERED[!(GAT_FILTERED$File.accession  %in% Explicit_exclusion), ]


# Update relative ranking after removing datasets
GAT_FILTERED <- GAT_FILTERED[order(GAT_FILTERED$WL_fold, decreasing = TRUE), ]
GAT_FILTERED$WL_rank <- seq.int(nrow(GAT_FILTERED))
GAT_FILTERED <- GAT_FILTERED[order(GAT_FILTERED$DHS_fold, decreasing = TRUE), ]
GAT_FILTERED$DHS_rank <- seq.int(nrow(GAT_FILTERED))
GAT_FILTERED <- GAT_FILTERED[order(GAT_FILTERED$OQs_fold, decreasing = TRUE), ]
GAT_FILTERED$OQs_rank <- seq.int(nrow(GAT_FILTERED))
GAT_FILTERED <- GAT_FILTERED[order(GAT_FILTERED$opOQs_fold, decreasing = TRUE), ]
GAT_FILTERED$opOQs_rank <- seq.int(nrow(GAT_FILTERED))

if(savetables)
{
  write.csv(GAT_FILTERED, file =paste("HepG2/GAT_analysis/GAT_HepG2_", Gat_Seg, '_', PEAK_NUMBER_Cutoff, "Peaks_Gat-Analysis_filt.csv",  sep=""))
}


# ===========Trimmed and Condensed version of filterd===================================
# Generate a trimmed  version only containing most relevant parameters
GAT_FILTERED_trim <- GAT_FILTERED[,  c("File.accession", "Experiment.accession", "Experiment.target", "Peaks", "WL_fold", "WL_qvalue", "WL_rank", "DHS_fold", "DHS_qvalue", "DHS_rank", "OQs_fold", "OQs_qvalue", "OQs_rank", "opOQs_fold", "opOQs_qvalue", "opOQs_rank")] 
GAT_FILTERED_trim$average_rank <- rowMeans(GAT_FILTERED_trim[, c("WL_rank",  "DHS_rank", "OQs_rank", "opOQs_rank")])
GAT_FILTERED_trim <- GAT_FILTERED_trim[order(GAT_FILTERED_trim$average_rank, decreasing = FALSE), ]

if(savetables)
{
  write.csv(GAT_FILTERED_trim, file =paste("HepG2/GAT_analysis/GAT_HepG2_", Gat_Seg, '_', PEAK_NUMBER_Cutoff, "Peaks_Gat-Analysis_trim.csv",  sep=""))
}



# ====== Aggregate by Target
# Comparing enrichment over different cell lines as different experiments in K562 resulted in different enrichments. Aggregate by target with both by highest and  average fold enrichment per target.
# Tagged and fusion proteins will be treated as unmodified proteins

# remove information on tag
ENCODE_bed_meta$Experiment.target <- gsub("-human", "", ENCODE_bed_meta$Experiment.target)



GAT_FILTERED_trim_by_Target <- GAT_FILTERED_trim[, c("Experiment.target", "WL_fold", "DHS_fold", "OQs_fold", "opOQs_fold")]
# remove information on tag
GAT_FILTERED_trim_by_Target$Experiment.target <- gsub("eGFP-", "", GAT_FILTERED_trim_by_Target$Experiment.target)
GAT_FILTERED_trim_by_Target$Experiment.target <- gsub("3xFLAG-", "" , GAT_FILTERED_trim_by_Target$Experiment.target)
GAT_FILTERED_trim_by_Target$Experiment.target <- gsub("phospho", "_Ph_" , GAT_FILTERED_trim_by_Target$Experiment.target)

# aggregate and keep maximum enrichment
GAT_FILTERED_trim_by_Target_max <- aggregate(GAT_FILTERED_trim_by_Target[,2:5], by = list(Exper_Target_List=GAT_FILTERED_trim_by_Target$Experiment.target), FUN = max , simplify = TRUE)

GAT_FILTERED_trim_by_Target_max <- GAT_FILTERED_trim_by_Target_max[order(GAT_FILTERED_trim_by_Target_max$WL_fold, decreasing = TRUE), ]
GAT_FILTERED_trim_by_Target_max$WL_rank <- seq.int(nrow(GAT_FILTERED_trim_by_Target_max))
GAT_FILTERED_trim_by_Target_max <- GAT_FILTERED_trim_by_Target_max[order(GAT_FILTERED_trim_by_Target_max$DHS_fold, decreasing = TRUE), ]
GAT_FILTERED_trim_by_Target_max$DHS_rank <- seq.int(nrow(GAT_FILTERED_trim_by_Target_max))
GAT_FILTERED_trim_by_Target_max <- GAT_FILTERED_trim_by_Target_max[order(GAT_FILTERED_trim_by_Target_max$OQs_fold, decreasing = TRUE), ]
GAT_FILTERED_trim_by_Target_max$OQs_rank <- seq.int(nrow(GAT_FILTERED_trim_by_Target_max))
GAT_FILTERED_trim_by_Target_max <- GAT_FILTERED_trim_by_Target_max[order(GAT_FILTERED_trim_by_Target_max$opOQs_fold, decreasing = TRUE), ]
GAT_FILTERED_trim_by_Target_max$opOQs_rank <- seq.int(nrow(GAT_FILTERED_trim_by_Target_max))


# aggregate using the mean enrichment
GAT_FILTERED_trim_by_Target_mean <- aggregate(GAT_FILTERED_trim_by_Target[,2:5], by = list(Exper_Target_List=GAT_FILTERED_trim_by_Target$Experiment.target), FUN = mean , simplify = TRUE)

GAT_FILTERED_trim_by_Target_mean <- GAT_FILTERED_trim_by_Target_mean[order(GAT_FILTERED_trim_by_Target_mean$WL_fold, decreasing = TRUE), ]
GAT_FILTERED_trim_by_Target_mean$WL_rank <- seq.int(nrow(GAT_FILTERED_trim_by_Target_mean))
GAT_FILTERED_trim_by_Target_mean <- GAT_FILTERED_trim_by_Target_mean[order(GAT_FILTERED_trim_by_Target_mean$DHS_fold, decreasing = TRUE), ]
GAT_FILTERED_trim_by_Target_mean$DHS_rank <- seq.int(nrow(GAT_FILTERED_trim_by_Target_mean))
GAT_FILTERED_trim_by_Target_mean <- GAT_FILTERED_trim_by_Target_mean[order(GAT_FILTERED_trim_by_Target_mean$OQs_fold, decreasing = TRUE), ]
GAT_FILTERED_trim_by_Target_mean$OQs_rank <- seq.int(nrow(GAT_FILTERED_trim_by_Target_mean))
GAT_FILTERED_trim_by_Target_mean <- GAT_FILTERED_trim_by_Target_mean[order(GAT_FILTERED_trim_by_Target_mean$opOQs_fold, decreasing = TRUE), ]
GAT_FILTERED_trim_by_Target_mean$opOQs_rank <- seq.int(nrow(GAT_FILTERED_trim_by_Target_mean))

if (savetables)
{
  write.csv(GAT_FILTERED_trim_by_Target_max, file =paste("HepG2/GAT_analysis/GAT_HepG2_", Gat_Seg, "_Gat-Analysis_byTarget-max.csv",  sep=""))
  write.csv(GAT_FILTERED_trim_by_Target_mean, file =paste("HepG2/GAT_analysis/GAT_HepG2_", Gat_Seg, "_Gat-Analysis_byTarget-mean.csv",  sep=""))
}

# ==========
# Plot Data  
# =========
if (plotdata)
{
#for plots remove all the additional tag information: eGFP, FLAG, etc.
GAT_FILTERED_trim$Experiment.target <- gsub("eGFP-", "", GAT_FILTERED_trim$Experiment.target)
GAT_FILTERED_trim$Experiment.target <- gsub("3xFLAG-", "" , GAT_FILTERED_trim$Experiment.target)
#also trim phosphoS...
GAT_FILTERED_trim$Experiment.target <- gsub("phospho", "_P_", GAT_FILTERED_trim$Experiment.target)
GAT_FILTERED_trim$Experiment.target <- gsub("POLR", "Pr", GAT_FILTERED_trim$Experiment.target)



# assign colors based on known features
Known_G4_proteins <- read.csv(file = "K562_Rerun_Jan2019/GAT_analysis/G4IPD_Oct2019.csv", header = F)
temp <- GAT_FILTERED_trim
temp$G4IPD <- ifelse(temp$Experiment.target %in% Known_G4_proteins$V1, "#aad2a5", "grey50")


# highlight candidates that are not considered "canonical TFs" (see http://humantfs.ccbr.utoronto.ca/) 
Known_TF <- read.csv(file = "GAT rerun 2019_new/DatabaseExtract_v_1.01.csv", header = T)
Known_TF <- Known_TF[, c("HGNC.symbol", "Is.TF.", "TF.assessment" )]
Known_TF <- Known_TF[Known_TF$TF.assessment == "ssDNA/RNA binding",]
temp$Is.TF <- ifelse(temp$Experiment.target %in% Known_TF$HGNC.symbol, "#E98B0B", "grey50")
#nrow(temp[temp$Is.TF == "#E98B0B",])


# =====
# =DHS
# =====

temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]

# all
gg <- ggplot(temp_plot, aes(x=as.factor(rev(DHS_rank)), y=DHS_fold)) + 
  geom_bar(stat="identity", fill=rev(temp_plot$G4IPD) ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  scale_x_discrete(labels="") +
  geom_hline(yintercept=5, linetype="dashed", color = "grey")+
  
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_all_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 22/2.54, limitsize = FALSE)


# top 20
temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]
temp_plot <- temp_plot[1:20,]

gg <- ggplot(temp_plot, aes(x=as.factor(rev(DHS_rank)), y=DHS_fold)) + 
  geom_bar(stat="identity",color="black", fill=rev(temp_plot$G4IPD), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  scale_x_discrete(labels=rev(temp_plot$Experiment.target)) +
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_top20_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 8/2.54, limitsize = FALSE)


# bottom 20
temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]
temp_plot <- tail(temp_plot, 20)

gg <- ggplot(temp_plot, aes(x=as.factor(rev(DHS_rank)), y=DHS_fold)) + 
  geom_bar(stat="identity",color="black", fill=rev(temp_plot$G4IPD), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  scale_x_discrete(labels=rev(temp_plot$Experiment.target)) +
  scale_y_log10(breaks = trans_breaks("log10", function(x) 10^x), labels = trans_format("log10", math_format(10^.x))) +
  #annotation_logticks(sides= "l") + # unfortunately, this seems incompatible with coord_flip() -> have to edit this manually in illustrator :/
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_bottom20_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 8/2.54, limitsize = FALSE)



#####################
### other orientation

# all
temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]

gg <- ggplot(temp_plot, aes(x=as.factor((DHS_rank)), y=DHS_fold)) + 
  geom_bar(stat="identity", fill=(temp_plot$G4IPD) ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank())+
  scale_x_discrete(labels="") +
  geom_hline(yintercept=5, linetype="dashed", color = "grey30", size = 0.3 )



ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_all_", Gat_Seg, "_Gat-Analysis_byTarget-mean_rotated.pdf",  sep=""), width = 20/2.54, height = 6/2.54, limitsize = FALSE)

# top 20
temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]
temp_plot <- temp_plot[1:20,]

gg <- ggplot(temp_plot, aes(x=as.factor((DHS_rank)), y=DHS_fold)) + 
  geom_bar(stat="identity",color="black", fill=(temp_plot$G4IPD), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90))+
  scale_x_discrete(labels=(temp_plot$Experiment.target)) 


ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_top20_", Gat_Seg, "_Gat-Analysis_byTarget-mean_rotated.pdf",  sep=""), width = 9/2.54, height = 6/2.54, limitsize = FALSE)




# bottom 20
temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]
temp_plot <- tail(temp_plot, 20)

gg <- ggplot(temp_plot, aes(x=as.factor((DHS_rank)), y=DHS_fold)) + 
  annotation_logticks(sides= "l")+
  geom_bar(stat="identity",color="black", fill=(temp_plot$G4IPD), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90))+
  scale_x_discrete(labels=(temp_plot$Experiment.target)) +
  scale_y_log10(breaks = trans_breaks("log10", function(x) 10^x), labels = trans_format("log10", math_format(10^.x))) 
   # unfortunately, this seems incompatible with coord_flip() -> have to edit this manually in illustrator :/

ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_bottom20_", Gat_Seg, "_Gat-Analysis_byTarget-mean_rotated.pdf",  sep=""), width = 9/2.54, height = 6/2.54, limitsize = FALSE)


#############################
##Highlight non-canonical TFs
# all
temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]

gg <- ggplot(temp_plot, aes(x=as.factor((DHS_rank)), y=DHS_fold)) + 
  geom_bar(stat="identity", fill=(temp_plot$Is.TF) ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank())+
  scale_x_discrete(labels="") +
  geom_hline(yintercept=5, linetype="dashed", color = "grey30", size = 0.3 )


ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_all_", Gat_Seg, "_Gat-Analysis_byTarget-mean_TF.pdf",  sep=""), width = 20/2.54, height = 6/2.54, limitsize = FALSE)

# top 20
temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]
temp_plot <- temp_plot[1:20,]

gg <- ggplot(temp_plot, aes(x=as.factor((DHS_rank)), y=DHS_fold)) + 
  geom_bar(stat="identity",color="black", fill=(temp_plot$Is.TF), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90))+
  scale_x_discrete(labels=(temp_plot$Experiment.target)) 


ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_top20_", Gat_Seg, "_Gat-Analysis_byTarget-mean_TF.pdf",  sep=""), width = 9/2.54, height = 6/2.54, limitsize = FALSE)




# bottom 20
temp_plot <- temp[order(temp$DHS_fold, decreasing = TRUE), ]
temp_plot <- tail(temp_plot, 20)

gg <- ggplot(temp_plot, aes(x=as.factor((DHS_rank)), y=DHS_fold)) + 
  annotation_logticks(sides= "l")+
  geom_bar(stat="identity",color="black", fill=(temp_plot$Is.TF), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in DHS") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90))+
  scale_x_discrete(labels=(temp_plot$Experiment.target)) +
  scale_y_log10(breaks = trans_breaks("log10", function(x) 10^x), labels = trans_format("log10", math_format(10^.x))) 

ggsave(file =paste("HepG2/GAT_analysis/figures/DHS_bottom20_", Gat_Seg, "_Gat-Analysis_byTarget-mean_TF.pdf",  sep=""), width = 9/2.54, height = 6/2.54, limitsize = FALSE)








# =====
# =opOQS
# =====

temp_plot <- temp[order(temp$opOQs_fold, decreasing = TRUE), ]

# all
gg <- ggplot(temp_plot, aes(x=as.factor(rev(opOQs_rank)), y=opOQs_fold)) + 
  geom_bar(stat="identity", fill=rev(temp_plot$G4IPD) ) +
  xlab("") +
  ylab("fold enrichment in opOQs") +
  theme_minimal() +
  scale_x_discrete(labels="") +
  geom_hline(yintercept=5, linetype="dashed", color = "grey")+
  
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/opOQs_all_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 22/2.54, limitsize = FALSE)


# top 20
temp_plot <- temp[order(temp$opOQs_fold, decreasing = TRUE), ]
temp_plot <- temp_plot[1:20,]

gg <- ggplot(temp_plot, aes(x=as.factor(rev(opOQs_rank)), y=opOQs_fold)) + 
  geom_bar(stat="identity",color="black", fill=rev(temp_plot$G4IPD), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in opOQs") +
  theme_minimal() +
  scale_x_discrete(labels=rev(temp_plot$Experiment.target)) +
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/opOQs_top20_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 8/2.54, limitsize = FALSE)


# bottom 20
temp_plot <- temp[order(temp$opOQs_fold, decreasing = TRUE), ]
temp_plot <- tail(temp_plot, 20)

gg <- ggplot(temp_plot, aes(x=as.factor(rev(opOQs_rank)), y=opOQs_fold)) + 
  geom_bar(stat="identity",color="black", fill=rev(temp_plot$G4IPD), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in opOQs") +
  theme_minimal() +
  scale_x_discrete(labels=rev(temp_plot$Experiment.target)) +
  scale_y_log10(breaks = trans_breaks("log10", function(x) 10^x), labels = trans_format("log10", math_format(10^.x))) +
  #annotation_logticks(sides= "l") + # unfortunately, this seems incompatible with coord_flip() -> have to edit this manually in illustrator :/
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/opOQs_bottom20_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 8/2.54, limitsize = FALSE)


# =====
# = WL
# =====

temp_plot <- temp[order(temp$WL_fold, decreasing = TRUE), ]

# all
gg <- ggplot(temp_plot, aes(x=as.factor(rev(WL_rank)), y=WL_fold)) + 
  geom_bar(stat="identity", fill=rev(temp_plot$G4IPD) ) +
  xlab("") +
  ylab("fold enrichment in WL") +
  theme_minimal() +
  scale_x_discrete(labels="") +
  geom_hline(yintercept=100, linetype="dashed", color = "grey")+
  geom_hline(yintercept=200, linetype="dashed", color = "grey")+
  geom_hline(yintercept=300, linetype="dashed", color = "grey")+
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/WL_all_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 22/2.54, limitsize = FALSE)


# top 20
temp_plot <- temp[order(temp$WL_fold, decreasing = TRUE), ]
temp_plot <- temp_plot[1:20,]

gg <- ggplot(temp_plot, aes(x=as.factor(rev(WL_rank)), y=WL_fold)) + 
  geom_bar(stat="identity",color="black", fill=rev(temp_plot$G4IPD), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in WL") +
  theme_minimal() +
  scale_x_discrete(labels=rev(temp_plot$Experiment.target)) +
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/WL_top20_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 8/2.54, limitsize = FALSE)


# bottom 20
temp_plot <- temp[order(temp$WL_fold, decreasing = TRUE), ]
temp_plot <- tail(temp_plot, 20)

gg <- ggplot(temp_plot, aes(x=as.factor(rev(WL_rank)), y=WL_fold)) + 
  geom_bar(stat="identity",color="black", fill=rev(temp_plot$G4IPD), width=0.8 ) +
  xlab("") +
  ylab("fold enrichment in WL") +
  theme_minimal() +
  scale_x_discrete(labels=rev(temp_plot$Experiment.target)) +
  scale_y_log10(breaks = trans_breaks("log10", function(x) 10^x), labels = trans_format("log10", math_format(10^.x))) +
  #annotation_logticks(sides= "l") + # unfortunately, this seems incompatible with coord_flip() -> have to edit this manually in illustrator :/
  coord_flip()
ggsave(file =paste("HepG2/GAT_analysis/figures/WL_bottom20_", Gat_Seg, "_Gat-Analysis_byTarget-mean.pdf",  sep=""), width = 6/2.54, height = 8/2.54, limitsize = FALSE)

}




} # close loop for different Gat_Seg