# gene-coastats9.R
## this variant reads go-spp-mod.tab instead of gene-spp-mod.tab
## drop/change ggo usage; sites here are GOid not GeneID
# /bio/bio-grid/dspp/blstats3/predhsp

library("ade4") # correspondence analysis

sppabbrev <- as.matrix( c( "C", "W", "W", "M", 
  "m", "s", "c", "y", "e", "a", "r", "p", "w", "j", "v", "g"))
row.names(sppabbrev) <- c("cele", "dpulex", "dpul", "agam", #
  "dmel","dsim","dsec","dyak","dere","dana","dper","dpse","dwil","dmoj","dvir","dgri")#

protdbs <- c( "modMM", "modDM", "modCE", "modSC")
protmat <- as.matrix( c( "Mouse", "Fruitfly", "Worm", "Yeast")) 
row.names(protmat) <- protdbs
proteomes <- protmat[protdbs,]
#

#?? need separate COA flyorder and ceorder, or allorder ??
allorder <- c("cele", "dpul", "dmel","dsim","dsec","dyak","dere","dana","dper","dpse","dwil","dmoj","dvir","dgri")#
flyorder <- c("dmel","dsim","dsec","dyak","dere","dana","dper","dpse","dwil","dmoj","dvir","dgri")#
ceorder  <- c("cele", "dpul", "dmel","dpse","dvir");#
#
modorder <- c( "modDM", "modMM", "modCE")

#........ go categories
gocats <- read.delim("../gocatall.tab") # id,PARENT,cat,term#

names(gocats) <- c("GOid","Parents","Class","Term")
row.names(gocats) <- gocats[,1] # GO:id#
golabs <- list(P="GO_Process",F="GO_Function",C="GO_CellLocation")#
## fix this mess:#
#[1] "nucleobase, nucleoside, nucleotide and nucleic acid metabolism"#
## levels(gocats[,3])[77] <-  "nucleic acid metabolism"#

# primary species x gene counts correspondence analysis

## do each protdbs modDM, modMM, modCE
# for (mod in modorder) {
for (mod in  "modDM") {

cat("#------ Working on ",mod,"---------------------\n")

genespp <- read.delim(fn<-paste("go-spp-",mod,".tab",sep=""))#
cat("# read ",fn, " dim=",dim(genespp),"\n")

# genego <- read.delim(fn<-paste("gene-go-",mod,".tab",sep=""))#
# cat("# read ",fn, "dim=",dim(genego),"\n")

gspp <- genespp[,-1]; rownames(gspp) <- genespp[,1]#

# ggo <- genego[,-1]; rownames(ggo) <- genego[,1]#
# names(ggo) <- gsub("\\.",":",names(ggo))#
# 
# ## need gene-go only here? using gene-go1 further category matching
# bool.nogo <-  (apply(ggo,1,sum)==0)#
# #genes with GO.unknown: #
# bool.ungo <- ( ggo[,c("GO:0008372","GO:0005554","GO:0000004")]>0 )
# bool.ungo2 <- (apply(bool.ungo,1,sum)>0)
# bool.keep <- (!(bool.ungo2 | bool.nogo))
bool.keep <- TRUE

for (iorder in 1:2) {
### for (iorder in 1) {

if(iorder==1) spporder <- flyorder else spporder <- ceorder
spp1 <- spporder[1]

###gspp.go <- gspp[bool.keep,spporder]
###gspp.go[gspp.go > 4] <- 4   # set upper limit to counts, e.g. 4 == 4+ level
gspp.go <- gspp[ ,spporder]
nspp <- ncol(gspp.go)

outf <- paste("gene5go-",mod,"-",spp1,"-coa.out",sep="")
cat("# output:",outf,"---------------------","\n")

capture.output(
  cat("#",outf,"\n"),
  cat("# Filtered data, dim(gspp.go)= ",dim(gspp.go),"\n"),
  cat("# Proteome:", protmat[mod,],", Species:", paste(names(gspp.go),collapse=","),"\n"),
  file=outf,append = FALSE)

#-----
  
gspp.coa <- dudi.coa(gspp.go, scann = FALSE, nf=nspp-1)

## for GOgroups try PCA, as values are much different (bigger) than gene-spp dataset
## ** PCA no good, gives oddities ***
# gspp.coa <- dudi.pca(gspp.go, scann = FALSE, nf=nspp-1)
# capture.output(
#   cat("# USING *** Principle Components dudi.PCA **** (not dudi.COA)\n"),
#   file=outf,append = FALSE)

pclb <- sppabbrev[ row.names(gspp.coa$co),]

#.. print summary of gspp.coa: $eig plus spp factors
#.. add plot?  pdf ok

sppdf <- data.frame(matrix(nrow=0,ncol=nspp))
names(sppdf)<-names(gspp.coa$tab)
for (ip in 1:gspp.coa$nf) {
  v<-gspp.coa$co[,ip]
  sppdf <- rbind(sppdf, trunc(100*(v)/max(abs(v))))
}
sppdf[abs(sppdf)<10] <- "."
capture.output(
  cat("# COA eigenvalues:",gspp.coa$eig,"\n"),
  cat("# COA factors, Species weight, for ", protmat[mod,],"proteome\n"),
  cat("# ---------------------------------------------------\n"),
  print(sppdf,digits=1),
  cat("# ---------------------------------------------------\n"),
  file=outf,append = TRUE)

# write to file gspp.coa ****
gspp.coa.l <- gspp.coa
class(gspp.coa.l) <- "list"
capture.output( gspp.coa.l, file=paste(outf,".data",sep=""), append = FALSE)

pdf( file=fn<-paste("gene5go-",mod,"-",spp1,"-coa1.pdf",sep=""), width=8, height=10)
nc <- min(5,gspp.coa$nf)
plot(gspp.coa$co[,1:nc],pch=pclb,cex=1.5,
  main=paste( paste(protmat[mod,],"COA species x ngenes"),
  paste(paste(names(pclb),pclb,sep="="),collapse=", "),sep="\n"))
dev.off()
capture.output( cat("# plot COA.",1:nc," fn=",fn,"\n"), file=outf,append = TRUE)

if(nc < gspp.coa$nf) {
nc2 <- min(10,gspp.coa$nf)
nc3 <- c(1,nc:nc2)
pdf( file=fn<-paste("gene5go-",mod,"-",spp1,"-coa2.pdf",sep=""), width=8, height=10)
plot(gspp.coa$co[,nc3],pch=pclb,cex=1.5,
  main=paste( paste(protmat[mod,],"COA species x ngenes"),
  paste(paste(names(pclb),pclb,sep="="),collapse=", "),sep="\n"))
dev.off()
capture.output( cat("# plot COA.",nc3," fn=",fn,"\n"), file=outf,append = TRUE)
}

#-- want this here or longer list of all GO terms per COA factor ??

ig <- row.names(gspp.coa$li)
# ngo <- apply( ggo[ig,],2,sum)
ngo <- apply( gspp.go,1,max) # bad?
## names(ngo)<- ig #??
for (ip in 1:gspp.coa$nf) {
  wt <- gspp.coa$li[ ,ip]
  # wtgo <- apply( wt * ggo[ig,],2,sum) # mangles names GO: to GO.
  wtgo <- wt
  
  igo <- order(abs(wtgo),decreasing=TRUE)
  gon1 <- ig[igo] ##? names(ngo[igo])
  
  dfgo1 <- data.frame(
    GO=gon1,
    COA_wt=wtgo[igo], 
    Ngenes=ngo[igo], 
    Term=strtrim(as.character(gocats[gon1,"Term"]),30),
    Kind=as.character(gocats[gon1,"Class"]),  # "Class","Term"
    row.names=NULL
    )
  
  capture.output(
    cat("# COA factor.",ip,",", protmat[mod,],"proteome, species x ngene\n"),
    print(dfgo1[1:15,],digits=2),
    cat("#-------------------------------------------------------\n\n"),
    file=outf,append = TRUE)
} # end coa list

} # end spporder

} # end mod