CODEWITHSUNDEEP
rcluster-analysisdata-visualizationbioconductorphylogeny
Rgeek
Rgeek

Reputation: 449

Construct a circular phylogenetic tree

I have a table of genes and diseases to which they are related.I want to construct a phylogenetic tree and group the genes to their diseases.Below is a sample dataset,where gene1 column belongs to disease1 and gene2 belongs to disease2.Primarily gene1 and gene2 are related to each other,and are mapped to diseases they belong to.

gene1   gene2   disease1           disease2
AGTR1   ACHE    cancer              tumor
AGTR1   ACHE    parkinson's         asthma
ALOX5   ADRB1   myocardial infarct  heart failure
AR      ADORA1  breast cancer       anxiety disorder

I want to have a circular phylogenetic tree for my purposes,given in the link below: http://itol.embl.de/itol.cgi

Any suggestions to do this in R or any softwares?

Thanks enter image description here

Code I am running ,now:

d=read.csv("genes_disease.txt",sep="\t",header=TRUE)
phyl_gad <-as.phylo(hclust(dist(d)))
 plot(phyl_gad,type="fan",edge.col=c("red","green","blue","orange","yellow","pink","magenta","white"),show.tip.label=FALSE)

If I do show.tip.label=TRUE,there are too many labels that get plotted and makes the tips cluttered up.

My modified dataset is only two columns now,one for gene,one for disease.

Upvotes: 2

Views: 6146

Answers (2)

Ferroao
Ferroao

Reputation: 3033

I think what you wanted to do was not a phylogeny but a clustering by distance. Here is a repeatable example.

library(XML)
library(RCurl)#geturl
library(rlist)
library(plyr)
library(reshape2)
library(ggtree)

#get the genes/ diseases info from internet
#example from http://www.musclegenetable.fr/
urllist<-paste0("http://195.83.227.65/4DACTION/GS/",LETTERS[1:24] )
theurl <- lapply(urllist, function(x) RCurl::getURL(x,.opts = list(ssl.verifypeer = T) ) )# wait
theurl2<-lapply(theurl, function(x) gsub("<span class='Style18'>","__",x))
tables <- lapply(theurl2, function (x) XML::readHTMLTable(x) )
tables2 <- lapply(tables, function(x) rlist::list.clean(x, fun = is.null, recursive = FALSE) )
unlist1 = lapply(tables2, plyr::ldply)
newdf<-do.call(rbind, unlist1)
colnames(newdf)[4]<-"diseases"
colnames(newdf)[2]<-"Gene"
newdf$gene<-sub("([A-z0-9]+)(__)(.*)","\\1",newdf$Gene)
newdf$diseases<-sub("(\\* )","",newdf$diseases, perl=T)

#split info of several diseases per gene, and simplify text 
#to allow better clustering
newdf2<-as.data.frame(data.table::setDT(newdf)[, strsplit(as.character(diseases), "* ", fixed=TRUE), by = .(gene, diseases)
][,.(diseases = V1, gene)])
newdf2$disease<-sub("([A-z0-9,\\-\\(\\)\\/ ]+)( \\- )(.*)","\\1",newdf2$diseases)
newdf2$disease<-gsub("[0-9,]","",newdf2$disease)
newdf2$disease<-gsub("( [A-Z]{1,2})$","",newdf2$disease)
newdf2$disease<-gsub("(\\-)","",newdf2$disease)
newdf2$disease<-gsub("\\s*\\([^\\)]+\\)","",newdf2$disease)
newdf2$disease<-gsub("\\s*type.*","",newdf2$disease, ignore.case = T)
newdf2$disease<-gsub("(X{0,3})(IX|IV|V?I{0,3})","", newdf2$disease)
newdf2$disease<-gsub("( [A-z]{1,2})$","",newdf2$disease)
newdf2$disease<-sub("^([a-z])(.*)","\\U\\1\\E\\2",newdf2$disease, perl=T)
newdf2$disease<-trimws(newdf2$disease)
newdf2<-newdf2[,c(2,3)]

#make clustering and tree
newcasted <- reshape2::dcast(newdf2, gene ~ disease)
phyl_gad <-ape::as.phylo(hclust(dist(newcasted)))

#use names of genes and diseases in tree
DT <- data.table::as.data.table(newdf2)
newdf4<-as.data.frame(DT[, lapply(.SD, paste, collapse=","), by = gene, .SDcols = 2])
newdf4$genemerge<-paste(newdf4$gene, newdf4$disease)
phyl_gad$tip.label<-newdf4$genemerge

#plot tree
ggtree::ggtree(phyl_gad, layout = "circular")+ ggtree::geom_tiplab2(offset=0.1, align = F, size=4)

enter image description here

Upvotes: 1

Max Candocia
Max Candocia

Reputation: 4376

Ah, I've done this before. As Bryan said, you want to use the ape package. Let's say that you have an hclust object. For example,

library(ape)
fit<-hclust(d,method='ward')
plot(as.phylo(fit),type='fan',label.offset=0.1,no.margin=TRUE)

If you want to modify the colors of the ends of the trees, you can use cutree and the tip.color parameter. This will create a repeating set of colors for the different clusters (e.g., color=c('red','blue') will have alternating blue and red text for the end of the branches.

nclus=...#insert number of clusters you want to cut to
color=...#insert a vector of colors here
fit<-hclust(d,method='ward')
color_list=rep(color,nclus/length(color))
clus=cutree(fit,nclus)
plot(as.phylo(fit),type='fan',tip.color=color_list[clus],label.offset=0.1,no.margin=TRUE)

I'm not sure what type of clustering method you want to use (I was using Ward's method), but that's how you do it.

Upvotes: 4

Related Questions