#####################################################################
#
#    Instructions on using this program
# 1. change the name of the Supplementary Functions file to Functions.R,
#    change the name of the Supplementary Data file to Data.csv,
#    and change extension of this file from .txt to .R
# 2. put this file, Functions.R and Data.csv in the same directory
# 3. change DIR variable to be the name of this directory
# 4. create two subdirectories called Figures and Tables
# 5. Install R 
#   (see www.r-project.org/ to get R and instructions on using R)
# 6. Run this program in R 
#
###################################################################
#
#    Description of Data File
#
#    The file is a comma separated file, with the first line giving 
#    the variable name. 
#
#    The variables are:
#
#      setting= setting number, see text and Table 1
#      study=study name
#      ug = dose for rPA vaccines
#      proportionofhumandose= dose for AVA vaccines
#      diluent= either "saline" or "adjuvanted" 
#      survive= 0 for died, 1 for survived
#      vaccine= C for AVA vaccine, A,B,D, or E for different rPA vaccines 
#      species= numeric symbol, 1=rhesus,2=rabbit,3=cynos,4=human
#      Species= either Rhesus, Rabbit, Cynos, or Human
#      challenge=number of weeks after initial vaccination of the challenge
#      ed50=TNA taken at week given by ed50_week variable
#      ed50_week = number of weeks after last vaccination (or after second vaccination for humans) 
#                  in which TNA was measured
#      elisa=ELISA
#      elisa_week = number of weeks after last vaccination (or after second vaccination for humans) 
#                  in which ELISA was measured
#      
#
#
###################################################################


#######################################
#  set DIR, nPerm, nBoot and determine 
#   where to put figures and tables
#######################################
DIR<-"H:/My Documents/extramural/anthrax Meta/R/forPaper/"
#DIR<-"C:/H/My Documents/extramural/anthrax Meta/R/forPaper/"

## use colors that are robust to colorblindness
# orange for rabbits, sky-blue for cynos, blue-green for rhesus
colorRabbits<-"#E69F00"
colorCynos<-"#56B4E9"
colorRhesus<-"#009E73"

# get functions from other file
source(paste(DIR,"Functions.R",sep=""))

## for final analysis use nPerm=2000 and nBoot=2000
## for quicker run can change that to smaller values
nPerm<-2000
nBoot<-2000
#nPerm<-3
#nBoot<-3
putplot<-function(name,width=3.5){
    ## for Windows based results, i.e., for importing into Word files use...
    #dev.print(win.metafile,filename=paste(DIR,"Figures/",name,".wmf",sep=""))
    ## or for more portability use
    if (!is.null(width)){
        dev.print(pdf,file=paste(DIR,"Figures/",name,".pdf",sep=""),width=width)
    } else {
        dev.print(pdf,file=paste(DIR,"Figures/",name,".pdf",sep=""))
    }
}
putcsv<-function(x,name,rowNames=FALSE){
    write.csv(x,file=paste(DIR,"Tables/",name,sep=""),row.names=rowNames)
}


###################################################################
#
#   Import data 
#
###################################################################
d<-read.csv(paste(DIR,"Data.csv",sep=""),header=TRUE)
### add log10 values
d<-data.frame(d,log10elisa=log10(d$elisa),log10ed50=log10(d$ed50))
# main data set, use only non-missing ed50=TNA
dm<-d[!is.na(d$ed50),]


###################################################################
#
#   Table: Settings
#
###################################################################


tableSettings<-function(d){
    n<-nrow(d)
    ### animals studies are 1-11
    ### for Human studies, only use setting=14, 2 doses IM
    S<-c(1:11,14)
    nanimal<-ndose<-rep(NA,length(S))
    for (i in 1:length(S)){
        di<-d[d$setting==S[i],]
        nanimal[i]<-nrow(di)
        if (di[1,"vaccine"]=="C"){
            ud<-unique(di[,"proportionofhumandose"])
            ndose[i]<-length(ud)
        } else {
            ud<-unique(di[,"ug"])
            ndose[i]<-length(ud)
        }
    }
    out<-data.frame(setting=S,ndose,nanimal)
    out
}
## count only non-missing TNA
tab1<-tableSettings(dm)
putcsv(tab1,"tableSettings.csv")





###################################################################
#
#   Figures: ELISA vs TNA
#
###################################################################



#### Figure S1H Human
dh<-dm[dm$Species=="Human",]

### for Human studies, in cross-species comparisons 
### only use setting=14, 2 doses IM
### but for ELISA vs TNA correlations we can use all human data
#dh<-d[d$setting==14,]
par(mfrow=c(1,1))
ed50<-dh$ed50
tigg<-dh$elisa
I<-ed50>min(ed50,na.rm=TRUE) & tigg>min(tigg,na.rm=TRUE)
I<- !is.na(I) & I 
set.seed(1)
jig1<-exp(rnorm(length(ed50),0,sd=.05))
jig2<-exp(rnorm(length(ed50),0,sd=.05))

plot(ed50*jig1,tigg*jig2,log="xy",xlab="Toxin Neutralizing Activity",ylab="ELISA",axes=FALSE)
axis(1,at=c(min(ed50,na.rm=TRUE),10,10^2,10^3,10^4,10^5),label=c("BLD","10","100","1000","10,000","100,000"))
axis(2,at=c(min(tigg,na.rm=TRUE),10,10^2,10^3,10^4,10^5),label=c("BLD","10","100","1000","10,000","100,000"),las=1)
box()



putplot("FigureS1H",width=7)
#putplot("FigureS1H35",width=3.5)



ct1<-cor.test(log(ed50),log(tigg))
ct1

### calculate the proportion below the limit of detection
propBLD<-function(x){ x<-x[!is.na(x)]; 100*length(x[x==min(x)])/length(x) }
#propBLD(ed50)
#propBLD(tigg)

#### check correlation without values BLD
#### two different LoD values, use >15 and >1.
#I<-ed50>min(ed50,na.rm=TRUE) & tigg>min(tigg,na.rm=TRUE)
#I<-ed50>15 & tigg> 1
#I<- !is.na(I) & I 
#ct2<-cor.test(log(ed50[I]),log(tigg[I]))
#ct2



### Figure S1R Rabbit
dr<-dm[dm$Species=="Rabbit",]
ed50<-dr$ed50
tigg<-dr$elisa
I<-ed50>min(ed50,na.rm=TRUE) & tigg>min(tigg,na.rm=TRUE)
I<- !is.na(I) & I 
jig1<-exp(rnorm(length(ed50),0,sd=.05))
jig2<-exp(rnorm(length(ed50),0,sd=.05))

plot(ed50*jig1,tigg*jig2,log="xy",xlab="Toxin Neutralizing Activity",ylab="ELISA",axes=FALSE,col=colorRabbits)
axis(1,at=c(min(ed50,na.rm=TRUE),10,10^2,10^3,10^4,10^5),label=c("BLD","10","100","1000","10,000","100,000"))
axis(2,at=c(min(tigg,na.rm=TRUE),10,10^2,10^3,10^4,10^5),
    label=c("BLD","10","100","1000","10,000","100,000"),las=1)
box()

#propBLD(ed50)
#propBLD(tigg)

ct1<-cor.test(log(ed50),log(tigg))
ct1
#I<-ed50>min(ed50,na.rm=TRUE) & tigg>min(tigg,na.rm=TRUE)
#I<- !is.na(I) & I 
#ct2<-cor.test(log(ed50[I]),log(tigg[I]))
#ct2


putplot("FigureS3R",width=7)


### Figure 1NHP

dNHP<-dm[dm$Species=="Cynos" | dm$Species=="Rhesus",]
ed50<-dNHP$ed50
tigg<-dNHP$elisa
I<-ed50>min(ed50,na.rm=TRUE) & tigg>min(tigg,na.rm=TRUE)
I<- !is.na(I) & I 
jig1<-exp(rnorm(length(ed50),0,sd=.05))
jig2<-exp(rnorm(length(ed50),0,sd=.05))

plot(ed50*jig1,tigg*jig2,log="xy",xlab="Toxin Neutralizing Activity",ylab="ELISA",axes=FALSE,type="n")
I<-dNHP$Species=="Cynos"
points(ed50[I]*jig1[I],tigg[I]*jig2[I],col=colorCynos)
I<-dNHP$Species=="Rhesus"
points(ed50[I]*jig1[I],tigg[I]*jig2[I],col=colorRhesus)

axis(1,at=c(min(ed50,na.rm=TRUE),10,10^2,10^3,10^4,10^5),label=c("BLD","10","100","1000","10,000","100,000"))
axis(2,at=c(min(tigg,na.rm=TRUE),10,10^2,10^3,10^4,10^5),
    label=c("BLD","10","100","1000","10,000","100,000"),las=1)
box()


#propBLD(ed50)
#propBLD(tigg)

ct1<-cor.test(log(ed50),log(tigg))
ct1
#I<-ed50>min(ed50,na.rm=TRUE) & tigg>min(tigg,na.rm=TRUE)
#I<- !is.na(I) & I 
#ct2<-cor.test(log(ed50[I]),log(tigg[I]))
#ct2

putplot("FigureS2NHP",width=7)


###################################################################
#
#   Figure: logistic with Bootstrap CI
#
###################################################################
OUT<-matrix(NA,11,8,dimnames=list(paste("setting",1:11),c("estimate","lower","upper","n","var log(x)","mean log(x)","n Boot","n Boot missing")))
par(mfrow=c(3,4),oma=c(4,4,0,0),mar=c(1,1,3.5,1))
for (i in 1:11){
    OUT[i,]<-glmplotBoot(s=i,B=nBoot)
}
# if for the bootstrap data set all animals survive or all die, then set 
# values to missing, count how many are missing in such a way by OUT[,"n Boot missing"]



putcsv(round(OUT),"PA50with95CI.csv",rowNames=TRUE)
mtext("Toxin Neutralizing Activity",side=1,outer=TRUE,padj=2,cex=1.3)
mtext("Probability of Survival",outer=TRUE,side=2,padj=-2,cex=1.3)

#putplot("Fig11Panel",width=5)
putplot("Figure1",width=5)


# meta analysis for forest plots
metaRabbits<-meta(OUT[1:5,6],OUT[1:5,5],OUT[1:5,4],tau2=NULL)
metaCynos<-meta(OUT[6:7,6],OUT[6:7,5],OUT[6:7,4],tau2=NULL)
metaRhesus<-meta(OUT[8:11,6],OUT[8:11,5],OUT[8:11,4],tau2=NULL)

# fixed effects model, tau2=0
#metaRabbits<-meta(OUT[1:5,6],OUT[1:5,5],OUT[1:5,4],tau2=0)
#metaCynos<-meta(OUT[6:7,6],OUT[6:7,5],OUT[6:7,4],tau2=0)
#metaRhesus<-meta(OUT[8:11,6],OUT[8:11,5],OUT[8:11,4],tau2=0)


forest<-matrix(NA,14,5,dimnames=list(
    c(paste("Setting",1:5),"Overall rabbits",
      paste("Setting",6:7),"Overall cynos",
      paste("Setting",8:11),"Overall rhesus"),
     c("estimate","lower","upper","n","tau2")))
forest[1:5,1:4]<-OUT[1:5,c("estimate","lower","upper","n")]
forest[6,]<-metaRabbits
forest[7:8,1:4]<-OUT[6:7,c("estimate","lower","upper","n")]
forest[9,]<-metaCynos
forest[10:13,1:4]<-OUT[8:11,c("estimate","lower","upper","n")]
forest[14,]<-metaRhesus

par(mfrow=c(1,1),mar=c(4,4,1,1),oma=c(0,0,0,0))
plotForest(forest,bit=.1,plotstyle="traditional",XLIM=10^c(0,5.5),XLAB="Toxin Neutralizing Activity")
#plotForest(forest,bit=.1,plotstyle="traditional",XLAB="Toxin Neutralizing Activity")

#putplot("ForestPA50",width=3.5)
putplot("Figure2",width=3.5)


putcsv(forest,"MetaPA50.csv",rowNames=TRUE)





### Now do permutation tests to make sure the slopes are all significantly
### different from zero

set.seed(43943591)
pvalues<-rep(NA,11)
names(pvalues)<-paste("settting=",1:11)
for (i in 1:11){
    pvalues[i]<-permTestOneSetting(i,nperm=nPerm)
}
#### these are one-sided p-values, so multiply by 2 to make 2-sided
#### with nPerm=2000 all are the lowest possible, p=2*1/2001
2*pvalues



###################################################################
#
#   Figure: VE Combined
#
###################################################################
par(mfrow=c(1,1),mar=c(4,4,1,1)+.1)
## remember all functions are in Functions.R
figVECombined()
#putplot("Fig2b3noci",width=3.5)
putplot("Figure3",width=3.5)



#### Now do permutation test

set.seed(43943591)
pvalues<-rep(NA,3)
names(pvalues)<-c("rabbits","cynos","rhesus")

### p-values are one-sided, double to get 2 sided ones
outRabbits<-permTest(1:5,nperm=nPerm)
2*outRabbits$p.value
outCynos<-permTest(6:7,nperm=nPerm)
2*outCynos$p.value
outRhesus<-permTest(8:11,nperm=nPerm)
2*outRhesus$p.value

###################################################################
#
#   Table: PPE 
#
###################################################################
t0<-proc.time()
tabPPE<-calcCoD.PPE(1:11,x="log10ed50")
putcsv(tabPPE,"tablePPE.csv")
tabPPEboot<-ppeboot(nBoot)
putcsv(tabPPEboot,"tablePPEboot.csv")

###################################################################
#
#   Figure: Model 2 for Setting 3 
#
###################################################################
par(mfrow=c(1,1),mar=c(5,4,1,1)+.1)
#graph.studyModel(d=dm[dm$setting==3 & dm$ug>0,],
#     xvar="log10ed50",svar="ug",fit="none",XLAB="TNA at Week 4",LEGX=1.1,DOSE0=.005)
graph.studyModel(d=dm[dm$setting==3,],
     xvar="log10ed50",svar="ug",fit="none",XLAB="TNA at Week 4",LEGX=1.1,DOSE0=.005)

#putplot("FigModel2Setting3",width=3.5)
putplot("Figure4",width=3.5)


#### test for dose effect from Model 2
d3<-dm[dm$setting==3,]
logdoseS3<-log10(d3$ug)
logdoseS3[logdoseS3==-Inf]<-log10(0.005)
d3<-data.frame(d3,logdose=logdoseS3)
gModel2Setting3<-glm(survive~log10ed50+logdose,data=d3)
summary(gModel2Setting3)


###################################################################
#
# Cross Species Pairwise Tests on PA50
#
###################################################################
test36<-pwe(3,6,B=nBoot)
test38<-pwe(3,8,B=nBoot)
test68<-pwe(6,8,B=nBoot)

test47<-pwe(4,7,B=nBoot)
pairwiseTests<-rbind(test36,test38,test68,test47)
putcsv(pairwiseTests,"pairwiseTestsSpecies.csv")



###################################################################
#
#   Table: Cross-species
#
###################################################################
## setting 14 -  AVRP_Human_2IM has day 56 data 
## from people who have received 2 intra-muscular shots at 0 and 28 days.
strata<-list(c(3,6,8),c(4,7))
set.seed(123011)
tabObsPred<-obspredTable(d=dm,dH=dm[dm$setting==14,],strata,B=nBoot)
putcsv(tabObsPred,"crossSpeciesObsPred.csv")
######################################################
#### Figure Cross Species Prediction
######################################################
par(mar=c(4,4,1,1)+.1)
crossSpeciesPlot(3,6,colorRabbits,colorCynos,XLAB="Toxin Neutralizing Activity")
#putplot("crossSpecies36",width=3.5)
putplot("Figure5",width=3.5)