merge_similar_clusters.R

#!/usr/bin/env Rscript

# Merging FASTA-files, which motifs are similar.
#
# @parameter tsv_in <string> Path to TSV file generated by Tomtom.
#                            The input for Tomtom is a from all clusters merged meme-file.
# @parameter file_list <string> Numerically sorted whitespace separated list of absolute fasta-file paths
# @parameter min_weight <INT> Minimum weight of edge allowed in graph clusters.


args = commandArgs(trailingOnly = TRUE)

tsv_in <- args[1]
file_list <- args[2]
min_weight <- args[3]

files <- unlist(as.list(strsplit(file_list, ",")))

# split the string on the character "." in the first to columns and safe the last value each, to get the number of the cluster.
tsv <- data.table::fread(tsv_in, header = TRUE, sep = "\t",colClasses = 'character')
query_cluster <- unlist(lapply(strsplit(tsv[["Query_ID"]],"\\."), function(l){
  tail(l,n=1)
}))
target_cluster <- unlist(lapply(strsplit(tsv[["Target_ID"]],"\\."), function(l){
  tail(l,n=1)
}))

# create data.table with only the cluster-numbers
sim_not_unique <- data.table::data.table(query_cluster,target_cluster)
# convert from character to numeric values
sim_not_unique[, query_cluster := as.numeric(query_cluster)]
sim_not_unique[, target_cluster := as.numeric(target_cluster)]

# remove rows if column 1 is idential to column 2
edgelist <- sim_not_unique[query_cluster != target_cluster]

# create graph from data.frame
g <- igraph::graph_from_edgelist(as.matrix(edgelist))
# converting graph to adjacency matrix
adj_matrix <- igraph::get.adjacency(g, names = T)
# generating weighted graph from adjacency matrix
g_adj <- igraph::graph_from_adjacency_matrix(adj_matrix, weighted = T)

# get subgraphs from graph with edges of weight > min_weight
s1 <- igraph::subgraph.edges(g_adj, igraph::E(g_adj)[igraph::E(g_adj)$weight>min_weight], del=F)
png('motif_clusters.png')
plot(s1)
dev.off()
clust <- igraph::clusters(s1)
if (clust$no < 1){
  b <- lapply(files, function(f){
    system(paste("cat",f,">",basename(f)))
  })
}
# merge FASTA-files depending on the clustered graphs
a <- lapply(seq(from = 1, to = clust$no, by = 1), function(i){
  cl <- as.vector(which(clust$membership %in% c(i)))
  fasta_list <- paste(files[cl], collapse = " ")
  name <- paste0("Cluster_",i,".fasta")
  system(paste("cat",fasta_list,">",name))
})