---
title: "Pais_et_al_complete_statistics"
author: "Luis Teixeira and Ines Pais"
date: "25/06/2018"
output:
  html_document: default
  pdf_document: default
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r libraries for the whole script}
# R version 3.5.0

#load libraries
#tidyverse 1.2.1
library(tidyverse)

#reshape 0.8.7
library(reshape)

#reshape2 1.4.3
library(reshape2)

#lme4 1.1-17
library(lme4)

#lmerTest 3.0-1
library(lmerTest)

#multcomp 1.4-8
library(multcomp)

#lsmeans 2.27-62
library(lsmeans)

#multcompView 0.1-7
library(multcompView)
```

```{r options}

#change this value to "yes" to save some output files throughout the script
write_files <- "no"
```


##Figure 1

### Iso vials not flipped
```{r iso 10 days non flipping}

iso_vials_nf <- read.csv("S1_Data.csv",sep=",",head=T)
iso_vials_nf$replicate <- as.factor(iso_vials_nf$replicate)
iso_vials_nf$day <- as.factor(iso_vials_nf$day)

#create column with the cfu from the medium that present the highest number
iso_vials_nf$maxi <- apply(iso_vials_nf[,4:8], 1,max)

#create log from the maximum cfu
iso_vials_nf$log = log10(iso_vials_nf$maxi+1)

head(iso_vials_nf)

#plot data
ggplot(data=iso_vials_nf, aes(day,log)) +
  geom_jitter(height=0,width=.2) +
  expand_limits(y=c(0,7)) +
  facet_grid(replicate ~ .,margins = FALSE)


#statistical analyses
#lmer with replicate as random factor

modeliso_vials_nf <- lmer(log ~ day + (1 | replicate), data=iso_vials_nf)
summary(modeliso_vials_nf)


#estimates from model
#titres
lsmeansLT(modeliso_vials_nf)

values_iso_vials_nf <- data.frame(lsmeansLT(modeliso_vials_nf))

levels_iso_vials_nf <- values_iso_vials_nf[,c(3,7,8) ]

apply(levels_iso_vials_nf,1, function(i) '^'(10,i))

#fold change
'^'(10,levels_iso_vials_nf[2,1] - levels_iso_vials_nf[1,1])

```

### Iso vials flipped
```{r iso 10 days flipping}

iso_vials_f <- read.csv("S2_Data.csv",sep=",",head=T)
iso_vials_f$replicate <- as.factor(iso_vials_f$replicate)
iso_vials_f$day<-factor(iso_vials_f$day)

#create column with the cfu from the medium that present the highest number
iso_vials_f$maxi <- apply(iso_vials_f[,4:8], 1,max)

#create log from the maximum cfu
iso_vials_f$log = log10(iso_vials_f$maxi+1)

head(iso_vials_f)

#plot data
ggplot(data=iso_vials_f, aes(day,log)) +
  geom_jitter(height=0,width=.2) +
  expand_limits(y=c(0,7)) +
  facet_grid(replicate ~ .,margins = FALSE)


#statistical analyses
#lmer with replicate as random factor
model_iso_vials_f <- lmer(log ~ day + (1 | replicate), data=iso_vials_f)
summary(model_iso_vials_f)


#estimates from model
#titres
lsmeansLT(model_iso_vials_f)

values_iso_vials_f <- data.frame(lsmeansLT(model_iso_vials_f))
levels_iso_vials_f <- values_iso_vials_f[,c(3,7,8) ]
apply(levels_iso_vials_f,1, function(i) '^'(10,i))

#fold change
'^'(10,levels_iso_vials_f[1,1] - levels_iso_vials_f[2,1])
```

### Iso vials flipped qPCR
```{r iso 10 days flipping qPCR}
iso_vials_f_qpcr <- read.csv("S3_Data.csv",sep=",",head=T)
iso_vials_f_qpcr$replicate <- as.factor(iso_vials_f_qpcr$replicate)
iso_vials_f_qpcr$day<-factor(iso_vials_f_qpcr$day)

#create log
iso_vials_f_qpcr$log = log10(iso_vials_f_qpcr$relativelevels)

head(iso_vials_f_qpcr)

#plot
ggplot(data=iso_vials_f_qpcr, aes(day,log)) +
  geom_jitter(height=0,width=.2) +
  facet_grid(replicate ~ .,margins = FALSE)

#statistical analyses
#lmer with replicate as random factor
model_iso_vials_f_qpcr <- lmer(log ~ day + (1 | replicate), data=iso_vials_f_qpcr)
summary(model_iso_vials_f_qpcr)

#fold change
lsmeansLT(model_iso_vials_f_qpcr)

values_iso_vials_f_qpcr <- data.frame(lsmeansLT(model_iso_vials_f_qpcr))
levels_iso_vials_f_qpcr <- values_iso_vials_f_qpcr[,c(3,7,8) ]
apply(levels_iso_vials_f_qpcr,1, function(i) '^'(10,i))

#fold change
'^'(10,levels_iso_vials_f_qpcr[1,1] - levels_iso_vials_f_qpcr[2,1])
```

### growth on food
```{r growth on food}

growth_vials <- read.csv("S4_Data.csv",sep=",",head=T)
growth_vials$day<-factor(growth_vials$day)

#create column with the cfu from the medium that present the highest number
growth_vials$maxi <- apply(growth_vials[,4:8], 1,max)

#create log from the maximum cfu
growth_vials$log = log10(growth_vials$maxi + 1)

head(growth_vials)

#plot
ggplot(data=growth_vials, aes(day,log)) +
  geom_jitter(height=0,width=.2)

#statistical analyses
#lm
model_growth_vials <- lm(log ~day, data=growth_vials )
summary(model_growth_vials)

#estimates from model
lsmeans::lsmeans(model_growth_vials, specs = "day")

values_growth_vials <- summary(lsmeans::lsmeans(model_growth_vials, specs = "day"))
levels_growth_vials <- summary(lsmeans::lsmeans(model_growth_vials, specs = "day"))[2]

#fold change
'^'(10,levels_growth_vials[2,1] - levels_growth_vials[1,1])

```

### wild flies vials flipped
```{r wild flies flipped all data together}

wild_vials_f <- read.csv("S5_Data.csv",sep=",",head=T)
wild_vials_f$replicate <- as.factor(wild_vials_f$replicate)

#create column with the cfu from the medium that present the highest number
wild_vials_f$maxi <- apply(wild_vials_f[,4:8], 1,max)
 
#create log from the maximum cfu
wild_vials_f$log = log10(wild_vials_f$maxi+1)

head(wild_vials_f)

#plot data
ggplot(data=wild_vials_f, aes(as.factor(day),log)) +
  geom_jitter(height=0,width=.2) +
  expand_limits(y=c(0,7)) +
  facet_grid(replicate ~ .,margins = FALSE)

#statistical analysis
model_wild_vials_f <- lmer(log ~ day + (1 | replicate), data=wild_vials_f)
summary(model_wild_vials_f)

#estimates
values_wild_vials_f <- summary(lsmeans::lsmeans(model_wild_vials_f, specs = "day", at = list(day = c(0,10,20))))
levels_wild_vials_f <- data.frame(values_wild_vials_f[,c("lsmean")])

#fold change day 0 / day 10
'^'(10, levels_wild_vials_f[1,1] - levels_wild_vials_f[2,1])

apply(levels_wild_vials_f,1, function(i) '^'(10,i))
```

##Figures 2 and 3

```{r import table and process}
#import table
bacteria<-read.csv("S6_Data.csv",sep=",",head=T)

#create column with OTU only
bacteria$OTU <- sapply(strsplit(as.character(bacteria$greengenes_tax_string),'otu_'), "[", 2)
bacteria$OTU <- paste("OTU",bacteria$OTU,sep="")

#define OTUs that are contaminants
contaminats <-c("OTU2881","OTU2886","OTU3005")
#remove contaminants
bacteria <- filter(bacteria, !(OTU %in% contaminats))

#groups of bacteria with same colony morphology
Acetobacteraceae<-c("OTU2782", "OTU2781", "OTU2784","OTU2755","OTU2757","OTU2758","OTU2759","OTU2760","OTU2761","OTU2762","OTU2753","OTU2763")
Enterobacteriaceae <- c("OTU3643", "OTU3529", "OTU3638", "OTU3592", "OTU3635", "OTU3558", "OTU3570", "OTU3559","OTU3548")
Lactobacillus <- c("OTU1870","OTU1905","OTU1865")

#creat column with OTU or aggregate of OTUs when morphologies of colonies indistinguishable
bacteria$OTUtable <- bacteria$OTU
bacteria$OTUtable[bacteria$OTUtable %in% Lactobacillus] <- "Lacto"
bacteria$OTUtable[bacteria$OTUtable %in% Acetobacteraceae] <- "Aceto"
bacteria$OTUtable[bacteria$OTUtable %in% Enterobacteriaceae] <- "Entero"
```

### Analysis of diversity of bacteria
```{r script to get accumulation curve of OTUs}

#get number of entries
n <- length(distinct(bacteria,Source,Day,Fly)[,1])
accumulation <- data.frame(source_day = rep(NA,n),fly = rep(NA,n),count = rep(NA,n))


bacteria$source_day<-with(bacteria,paste(Source," ",Day,"days",sep = ""))
conditions <- distinct(bacteria,source_day)
counter <- 1

for (i in c(1:length(conditions[,1]))){
   flies <-  distinct(filter(bacteria,source_day == conditions[i,]),Fly)
   for (n in c(1:length(flies[,1]))){
     flies.data <- filter(bacteria,source_day == conditions[i,], Fly %in% flies[1:n,])
     count <- distinct(flies.data, OTU)
     accumulation$source_day[counter] <- conditions[i,]
     accumulation$fly[counter] <- flies[n,]
     accumulation$count[counter] <- length(count[,1])
     counter <- counter + 1
     }
  }

accumulation

ggplot(accumulation, aes(x=fly, y = count, color = source_day)) +
  geom_line(size=1) +
  geom_text(data = filter(accumulation, fly == "10", source_day == "Wild 0days" | source_day == "Lab 10days"), aes(label = source_day), vjust = -.6, hjust = .8) +
  geom_text(data = filter(accumulation, fly == "10", source_day == "Wild 10days" | source_day == "Lab 0days"), aes(label = source_day), vjust = 1.7, hjust = .7) +
  scale_colour_discrete(guide = 'none') +
  theme_bw() +
  theme(axis.title.y=element_text(margin=margin(0,8,0,0))) +
  labs( y = "Total OTUs", x = "Individuals") +
  scale_x_continuous(breaks=seq(0,10,1)) +
  ylim(0,36)
  


fly_ID <- distinct(bacteria,fly_ID)
Diversity <- data.frame(fly_ID,OTU_number = rep(NA,length(fly_ID)))

for (j in c(1:length(Diversity$fly_ID))){
  flies.data <- filter(bacteria,fly_ID == Diversity$fly_ID[j])
  Diversity$OTU_number[j] <- length(distinct(flies.data, OTU)[,1])
  }

#manual addition of Lab_Day10_FlyB, which had 0 bacteria
Diversity <- rbind(Diversity, data.frame(fly_ID = "Lab_Day10_FlyB", OTU_number = 0))
Diversity
```

```{r Max of each OTU per fly}
#Make table with max of each OTU per fly

#Creat table with number of OTUtable per morphotypes
Combinations <- distinct(bacteria,OTUtable, Unique_morpho, fly_ID)
OTUs_per_morpho <- summarise(group_by(Combinations,Unique_morpho, fly_ID), count = n())

#select unique morphos that have 2 or more entries in this table
OTUs_per_morpho <- filter(OTUs_per_morpho,count > 1)


#identify unique morphotypes with 2 different OTUS assigned
#identify for which of these there is more information on the OTUs in the fly analysis
#remove these unique morphotypes from analysis

list_p <- c()
table_p <- data.frame(Unique_morpho = c(), OTUtable = c())

for (i in c(1:nrow(OTUs_per_morpho))){
  pOTUs <- filter(Combinations, Unique_morpho == OTUs_per_morpho$Unique_morpho[i])$OTUtable
  fly <- filter(bacteria, fly_ID == OTUs_per_morpho$fly_ID[i], OTUtable %in% pOTUs)
  morpho_per_OTU <- summarise(group_by(fly, OTUtable), count = n())
    for (j in c(1,2)){
      if (morpho_per_OTU[j,2] > 1) table_p <- rbind(table_p, data.frame(Unique_morpho = c(as.character(OTUs_per_morpho$Unique_morpho[i])), OTUtable = c(morpho_per_OTU[j,1])))
    }
}

table_p$morpho_OTU <- paste(table_p$Unique_morpho, table_p$OTUtable)
bacteria_for_max <- filter(bacteria, !(paste(Unique_morpho, OTUtable) %in% table_p$morpho_OTU))

#Make column of OTUtable and fly ID together
bacteria_for_max$flyOTU <- with(bacteria_for_max,paste(OTUtable,fly_ID,sep="_"))

#get maximum for each OTUtable in each fly
tableOTU_fly <- summarise(group_by(bacteria_for_max,flyOTU), MaxOTU = max(Cfu_gut, na.rm = TRUE))
tableOTU_fly <- tableOTU_fly %>% separate(flyOTU, c("OTU","source","day","fly"), "_")


#write table
if(write_files == "yes") {write.table(tableOTU_fly, file = "tableOTU_fly.csv", row.names=FALSE, sep = ",")}
```

```{r define functions to extract different frequencies and to plot}
#define functions to analyse frequencies of OTUs in different groups of bacteria

#tables

Frequencies <- function(bacgroup) {
  #get subset of bacteria
  bacteria_group <- filter(bacteria, OTU %in% bacgroup)
  bacteria_group$source_day<-with(bacteria_group,paste(Source,Day,sep = "_"))
  #create table with all possible combinations of OTU, source, and day
  OTUs <- distinct(bacteria_group,OTU)
  source_days <- distinct(bacteria_group,source_day)
  freq_group <- expand.grid(OTU = OTUs[,1], source_day = source_days[,1], stringsAsFactors = FALSE)
  freq_group$divide<-freq_group$source_day
  freq_group <- freq_group %>% separate(divide, c("source", "day"), "_")
  #get number of colonies sequenced of each OTU per source_day
  freq_OTU <- summarise(group_by(bacteria_group,source_day,OTU), count = n())
  freq_group$count <- freq_OTU$count[match(paste(freq_group$source_day,freq_group$OTU),paste(freq_OTU$source_day,freq_OTU$OTU))]
  freq_group$count[is.na(freq_group$count)] <- 0
  #get number of total colonies per source_day
  group_per_source_day <- summarise(group_by(bacteria_group,source_day), count = n())
  freq_group$total_per_source_day <- group_per_source_day$count[match(freq_group$source_day,group_per_source_day$source_day)]
  #get frequencies of OTU per source_day
  freq_group$freq <- with(freq_group,count / total_per_source_day *100)
  #calculate in how many flies at different days in different populations are each OTU
  fly_and_OTU <- distinct(group_by(bacteria_group,OTU),fly_ID)
  fly_and_OTU$divide <- fly_and_OTU$fly_ID
  fly_and_OTU <- fly_and_OTU %>% separate(divide, c("source", "day", "fly"), "_")
  fly_and_OTU$day <- substring(fly_and_OTU$day,4)
  fly_per_OTU <- summarise(group_by(fly_and_OTU,OTU,source,day), count = n())
  freq_group$nflies <- fly_per_OTU$count[match(paste(freq_group$source_day,freq_group$OTU,sep="_"),paste(fly_per_OTU$source,fly_per_OTU$day,fly_per_OTU$OTU,sep="_"))]
  freq_group$nflies[is.na(freq_group$nflies)] <- 0
  #reorder rows
  freq_group <- arrange(freq_group, source, day, OTU)
  #write table
  if(write_files == "yes") {write.table(freq_group, file = paste(filename,"_frequencies.csv",sep=""), row.names=FALSE, sep = ",")}
  return(freq_group)
  }

total_OTU <- function(group){
  allflies <- distinct(tableOTU_fly,source,day,fly)
  #add entry for a fly with no bacteria
  allflies <- rbind(allflies, c("Lab", "Day10", "FlyB"))
  allflies$fly_ID <- with(allflies,paste(source,day,fly,sep = "_"))
  bacteria_group <- filter(tableOTU_fly, OTU == group)
  bacteria_group$fly_ID <- with(bacteria_group,paste(source,day,fly,sep = "_"))
  allflies$cfu <- bacteria_group$MaxOTU[match(allflies$fly_ID, bacteria_group$fly_ID)]
  allflies$cfu[is.na(allflies$cfu)] <- 0
  if(write_files == "yes") {write.table(allflies, file = paste(filename,"_cfu_fly.csv",sep=""), row.names=FALSE, sep = ",")}
  return(allflies)
  }


Plotfreq <- function(freq_bact,pop) {
  #make table that includes missing values (required for proper display of bars' width)
  comb <- filter(freq_bact,source == pop)
  max_freq <- max(comb$freq)+5
  #plot
  freqplot <- ggplot(comb,aes(x=OTU,y=freq,fill = day)) +
    geom_col(position = "dodge") +
    geom_text(aes(label=nflies), position=position_dodge(width=0.9), vjust=-0.25, size = 6) +
    theme_bw() +
    theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust=1, size = 8),
          axis.title.y=element_text(margin=margin(0,8,0,0))
    ) +
    labs( y = "Frequency", x = "OTU") +
    scale_fill_manual(values=c("#CCCCCC", "#000000")) +
    scale_y_continuous(breaks=seq(0,max_freq,10), limits = c(0,max_freq))
  if(write_files == "yes") {ggsave(paste(filename,pop,"freq.tiff",sep ="_"), width = 20, height = 12.5, units = "cm", dpi = 300)}
  return(freqplot)
}


#statistics

Chisq_freq <- function(file) {
  statsfreq <- filter(file,source == "Wild")
  statsfreq <- dplyr::select(statsfreq, OTU, count,day)
  statsfreq <- spread(statsfreq, key = day, value = count)
  return(chisq.test(statsfreq[,2:3], simulate.p.value = TRUE, B=1000000))
  }
```

###Analysis of Acetobacteraceae
```{r analysis of Acetobacteraceae}

filename <- "Aceto"
freq_aceto <- Frequencies(Acetobacteraceae)
total_aceto <- total_OTU("Aceto")

Plotfreq(freq_aceto,"Lab")
Plotfreq(freq_aceto,"Wild")

Chisq_freq(freq_aceto)

ggplot(total_aceto,aes(x = day, y = log10(cfu+1))) +
  geom_jitter(width=.1, size = 2, shape = 21, fill = "white", stroke = .5) +
  facet_grid(.~source) +
  theme_bw()

model1_aceto <- lm(log10(cfu+1)~source*day, data=total_aceto)
summary(model1_aceto)

model2_aceto <- lm(log10(cfu+1)~day, data=filter(total_aceto, source == "Wild"))
summary(model2_aceto)

#estimates
values_acetobacteraceae <- summary(lsmeans::lsmeans(model1_aceto, specs = "day", by = "source"))
levels_acetobacteraceae <- data.frame(values_acetobacteraceae[,c("lsmean")])

'^'(10,levels_acetobacteraceae)

```

###Analysis of Lactobacillus
```{r analysis of Lactobacillus}

filename <- "Lacto"
freq_lacto <- Frequencies(Lactobacillus)
total_lacto <- total_OTU("Lacto")

Plotfreq(freq_lacto,"Lab")
Plotfreq(freq_lacto,"Wild")

Chisq_freq(freq_lacto)

ggplot(total_lacto,aes(x = day, y = log10(cfu+1))) +
  geom_jitter(width=.1, size = 2, shape = 21, fill = "white", stroke = .5) +
  facet_grid(.~source) +
  theme_bw()

model1_lacto <- lm(log10(cfu+1)~source*day, data=total_lacto)
summary(model1_lacto)

model2_lacto <- lm(log10(cfu+1)~day, data=filter(total_lacto, source == "Wild"))
summary(model2_lacto)

```

###Analysis of Enterobacteriaceae
```{r analysis of Enterobacteriaceae}

filename <- "Entero"
freq_entero <- Frequencies(Enterobacteriaceae)
total_entero <- total_OTU("Entero")
total_entero <- filter(total_entero, source == "Wild")

Plotfreq(freq_entero,"Wild")

Chisq_freq(freq_entero)

ggplot(total_entero,aes(x = day, y = log10(cfu+1))) +
  geom_jitter(width=.1, size = 2, shape = 21, fill = "white", stroke = .5) +
  theme_bw()

model1_entero <- lm(log10(cfu+1)~day, data=total_entero)
summary(model1_entero)
```

##Analysis of Leuconostoc
```{r Leuconostoc analysis}

filename <- "Leuconostoc"
total_leuco <- total_OTU("OTU1934")
total_leuco <- filter(total_leuco, source == "Wild")

ggplot(total_leuco,aes(x = day, y = log10(cfu+1))) +
  geom_jitter(width=.2, size = 2, shape = 21, fill = "white", stroke = .5) +
  theme_bw()

model1_leuco <- lm(log10(cfu+1)~day, data=filter(total_leuco, source == "Wild"))
summary(model1_leuco)
```

```{r test independence of frequencies of leuconostoco and lactobacillus}

#create matrix with data
#get information per fly
Lactos <- filter(tableOTU_fly, (OTU == "Lacto" | OTU == "OTU1934"), source == "Wild")
Lactos$MaxOTU <- 1
LactosS <- spread(Lactos,key= OTU, value=MaxOTU, fill = 0, drop = FALSE)

#create columns indicating that fly has 0, lacto, leuco, or both
LactosS$Lacto_only <- ifelse(LactosS$Lacto == 1 & LactosS$OTU1934 == 0, 1, 0)
LactosS$Leuco_only <- ifelse(LactosS$Lacto == 0 & LactosS$OTU1934 == 1, 1, 0)
LactosS$both <- ifelse(LactosS$Lacto == 1 & LactosS$OTU1934 == 1, 1, 0)
LactosS$none <- ifelse(LactosS$Lacto == 0 & LactosS$OTU1934 == 0, 1, 0)

#create matrix for chisquare
Lacto_table <- data.frame(c(0,0),c(0,0))
colnames(Lacto_table) <- c("Lacto_0","Lacto_1")
rownames(Lacto_table) <- c("Leuco_0","Leuco_1")
Lacto_table[1,1] <- sum(LactosS$none)
Lacto_table[1,2] <- sum(LactosS$Lacto_only)
Lacto_table[2,1] <- sum(LactosS$Leuco_only)
Lacto_table[2,2] <- sum(LactosS$both)

Lacto_table

#chi-square test
chisq.test(Lacto_table, simulate.p.value = TRUE, B=1000000)

```

##Figure 4

###load data stability in vials and cages
```{r load data and plot stability}
#load data
stab <- read.csv("S7_Data.csv",sep=",",head=T)
stab$replicate<-as.factor(stab$replicate)
stab$logcfu <- log10(stab$cfu+1)
head(stab)

#filter stability in cages and stability in vials
stabvials <- filter(stab, cagevial == "vial")
stabcages <- filter(stab, cagevial == "cage")

#PLOTS
##Stability in vials
ggplot(data=stabvials,aes(day,logcfu))+
	geom_jitter(width=.5) +
	facet_grid(bacteria ~ replicate,margins = "False")

##Stability in cages
ggplot(data=stabcages,aes(day,logcfu))+
	geom_jitter(width=.5) +
	facet_grid(bacteria ~ replicate,margins = "False")

```

###Analysis of stability in vials
```{r analysis of stability in vials}
#all data together
stabvials$day<-as.factor(stabvials$day)
stabvials$replicate<-as.factor(stabvials$replicate)

model_vials1 <- lmer(logcfu ~ day * bacteria + (1|replicate), data = stabvials)
model_vials2 <- lmer(logcfu ~ day + bacteria + (1|replicate), data = stabvials)
anova(model_vials1,model_vials2)

summary(model_vials1)

#estimates from the model
lsmeansLT(model_vials1)
values_vials1 <- data.frame(lsmeansLT(model_vials1))

apply(values_vials1[c(3,7,8)], 1, function(i) '^'(10,i))

#compare slopes
coef(model_vials1)

contrast_matrix_vials_slope <- rbind(
  "Ac. cibinongensis" =       c(0, 1, 0, 0, 0, 0, 0, 0),
  "Ac. OTU2753" =            c(0, 1, 0, 0, 0, 1, 0, 0),
  "Ac. thailandicus" =           c(0, 1, 0, 0, 0, 0, 1, 0),
  "L. pseudomesenteroides" = c(0, 1, 0, 0, 0, 0, 0, 1)
      )

summary(glht(model_vials1, contrast_matrix_vials_slope), test=adjusted("holm"))



```

###Analysis of stability during time-course analysis in cages 
###nls fit timecourse analysis
```{r nls timecourse}
#get only replicate 1 data
sbcg1 <- filter(stabcages, replicate == "1")

#models
model_all_1 <- nls(logcfu ~ a[bacteria] * exp(-b[bacteria]*day) + c[bacteria], data = sbcg1, start=list(a=c(4,4,4,4), b = c(1,1,1,1), c=c(0,0,0,0)))
summary(model_all_1)

model_all_2 <- nls(logcfu ~ a[bacteria] * exp(-b*day) + c[bacteria], data = sbcg1, start=list(a=c(4,4,4,4), b = 1, c=c(0,0,0,0)))
anova(model_all_1,model_all_2)
AIC(model_all_1)
AIC(model_all_2)


model_all_3 <- nls(logcfu ~ a * exp(-b*day) + c[bacteria], data = sbcg1, start=list(a=4, b = 1, c=c(0,0,0,0)))
anova(model_all_3,model_all_2)
AIC(model_all_3)

model_all_4 <- nls(logcfu ~ a[bacteria] * exp(-b*day) + c, data = sbcg1, start=list(a=c(4,4,4,4), b = 1, c=0))
anova(model_all_4,model_all_2)
AIC(model_all_4)

#model_all_2 is simplest model that explains the data

summary(model_all_2)
levels(sbcg1$bacteria)


coef_all_2 <- coef(model_all_2)

#half-life in hours (from b estimate)
log(2)*1/coef_all_2[5]*24

#compare asymptotes
contrast_matrix_asympt <- rbind(
  "cAc. cibinongensis vs. cAc. OTU2753" =            c(0, 0, 0, 0, 0, 1, -1, 0, 0),
  "cAc. cibinongensis vs. cAc. wild1153" =           c(0, 0, 0, 0, 0, 1, 0, -1, 0),
  "cAc. cibinongensis vs. cL. pseudomesenteroides" = c(0, 0, 0, 0, 0, 1, 0, 0, -1),
  "cAc. OTU2753 vs. cAc. wild1153" =                c(0, 0, 0, 0, 0, 0, 1, -1, 0),
  "cAc. OTU2753 vs. cL. pseudomesenteroides" =      c(0, 0, 0, 0, 0, 0, 1, 0, -1),
  "cAc. wild1153 vs. cL. pseudomesenteroides" =     c(0, 0, 0, 0, 0, 0, 0, 1, -1))


summary(glht(model_all_2, contrast_matrix_asympt), test=adjusted("holm"))


#estimates of asymptotes

values_all_2 <- data.frame(summary(model_all_2)$parameters)

table_estimates <- values_all_2[6:9,1:2]
table_estimates$bacteria <- levels(sbcg1$bacteria)
table_estimates$low_CI_log <- table_estimates$Estimate - 1.96 * table_estimates$Std..Error
table_estimates$high_CI_log <- table_estimates$Estimate + 1.96 * table_estimates$Std..Error
table_estimates$estimate <- '^'(10,table_estimates$Estimate)
table_estimates$low_CI <- '^'(10,table_estimates$low_CI_log)
table_estimates$high_CI <- '^'(10,table_estimates$high_CI_log)

table_estimates
```

###Analysis of stability in cages replicate 2
```{r analysis of stability in cages replicate 2}

#get replicate 2 only
sbcg2 <- filter(stabcages, replicate == "2")

sbcg2$day<-as.factor(sbcg2$day)

model_cg2 <- lm(logcfu ~ day * bacteria, data = sbcg2)
model_cg2b <- lm(logcfu ~ day + bacteria, data = sbcg2)

anova(model_cg2, model_cg2b)

summary(model_cg2b)


#test day5 logcfu different from 0
contrast_matrix_cfu_day5 <- rbind(
  "Ac. cibinongensis" =       c(1, 1, 0, 0, 0),
  "Ac. OTU2753" =            c(1, 1, 1, 0, 0),
  "Ac. wild1153" =           c(1, 1, 0, 1, 0),
  "L. pseudomesenteroides" = c(1, 1, 0, 0, 1)
      )

summary(glht(model_cg2b, contrast_matrix_cfu_day5), test=adjusted("holm"))

```

###localization of Ac. thailandicus
```{r analysis of localization of Ac. thailandicus}
cp <- read.csv("S8_Data.csv",sep=",",head=T)

cp$compartment <- factor(cp$compartment, levels = c("Crop","Anterior midgut","Middle midgut","Posterior midgut","Hindgut"))
cp$replicate<-as.factor(cp$replicate)
cp$day<-as.factor(cp$day)
cp$logcfu <- log10(cp$cfu+1)

##plot
ggplot(data=cp,aes(x = compartment,y=logcfu, fill=fly, colour=fly))+
	geom_jitter(width=0.1) +
	facet_grid(day ~ replicate ,margins = "False")+
  theme_bw()


model1_local <- lmer(logcfu ~ compartment*day + (1 | replicate), data = cp)
model2_local <- lmer(logcfu ~ compartment+day + (1 | replicate), data = cp)
anova(model1_local,model2_local)

summary(model1_local)
anova(model1_local)

lsmeansLT(model1_local)

#compare Crop and AM at day 5
contrast_matrix_day5_cp <- rbind(
  "Crop_day5 - AM_day5 = 0" =  c(0, -1, 0, 0, 0, 0, -1, 0, 0, 0),
  "Crop_day5 - MM_day5 = 0" =  c(0, 0, -1, 0, 0, 0, 0, -1, 0, 0),
  "Crop_day5 - PM_day5 = 0" =  c(0, 0, 0, -1, 0, 0, 0, 0, -1, 0),
  "Crop_day5 - H_day5 = 0"  =  c(0, 0, 0, 0, -1, 0, 0, 0, 0, -1),
  "AM_day5 - MM_day5 = 0"   =  c(0, 1, -1, 0, 0, 0, 1, -1, 0, 0),
  "AM_day5 - PM_day5 = 0"   =  c(0, 1, 0, -1, 0, 0, 1, 0, -1, 0),
  "AM_day5 - H_day5 = 0"    =  c(0, 1, 0, 0, -1, 0, 1, 0, 0, -1),
  "MM_day5 - PM_day5 = 0"   =  c(0, 0, 1, -1, 0, 0, 0, 1, -1, 0),
  "MM_day5 - H_day5 = 0"    =  c(0, 0, 1, 0, -1, 0, 0, 1, 0, -1),
  "PM_day5 - H_day5 = 0"    =  c(0, 0, 0, 1, -1, 0, 0, 0, 1, -1)
    )

summary(glht(model1_local, contrast_matrix_day5_cp, test=adjusted("holm")))
```

###ANALYSIS OF PROLIFERATION
```{r load data and plot proliferation}
#load data
profull <- read.csv("S9_Data.csv",sep=",",head=T)
profull$bacteria <- relevel(profull$bacteria, "Ac. OTU2753")
profull$replicate<-as.factor(profull$replicate)
profull$logcfu <- log10(profull$cfu+1)
profull$hours <- relevel(profull$hours,"0h")
head(profull)

##plot
ggplot(data=profull,aes(x = as.factor(log10titer),y=logcfu, fill=hours, colour = hours))+
	geom_jitter(width=0.2) +
	facet_grid(bacteria ~ replicate ,margins = "False")

#summmary
stats_pro <- as.data.frame(summarise(group_by(profull,bacteria, hours, log10titer),
                        median = median(logcfu),
                         mean = mean(logcfu),
                        meanCFUs = '^'(10,mean(logcfu)),
                         sd = sd(logcfu)))

stats_pro

## Linear models of individual bacteria
#Ac. OTU2753
pro2753 <- filter(profull, bacteria == "Ac. OTU2753")

modelpro2753_1 <- lmer(logcfu ~ log10titer*hours + (1 | replicate), data = pro2753)
model2753_2 <- lmer(logcfu ~ log10titer+hours + (1 | replicate), data = pro2753)
anova(modelpro2753_1,model2753_2)

summary(modelpro2753_1)

contrast_matrix_slopes_2753 <- rbind(
  "2753_2" =  c(0, 0, 1, 2), 
  "2753_3" =  c(0, 0, 1, 3),
  "2753_4" =  c(0, 0, 1, 4),
  "2753_5" =  c(0, 0, 1, 5)
  )

summary(glht(modelpro2753_1, contrast_matrix_slopes_2753, test=adjusted("holm")))




#Ac. cibinongensis
pro_cibinongensis <- filter(profull, bacteria == "Ac. cibinongensis")

model_cibinongensis_1 <- lmer(logcfu ~ log10titer*hours + (1 | replicate), data = pro_cibinongensis)
model_cibinongensis_2 <- lmer(logcfu ~ log10titer+hours + (1 | replicate), data = pro_cibinongensis)
anova(model_cibinongensis_1,model_cibinongensis_2)

summary(model_cibinongensis_2)


#Ac. thailandicus
pro_thailandicus <- filter(profull, bacteria == "Ac. thailandicus", hours != "24h_chaser")

model_thailandicus_1 <- lmer(logcfu ~ log10titer*hours + (1 | replicate), data = pro_thailandicus)
model_thailandicus_2 <- lmer(logcfu ~ log10titer+hours + (1 | replicate), data = pro_thailandicus)
anova(model_thailandicus_1,model_thailandicus_2)

summary(model_thailandicus_2)


stats_thai_prol <- as.data.frame(summarise(group_by(pro_thailandicus, replicate, hours, log10titer),
                        median = median(logcfu),
                         mean = mean(logcfu),
                        meanCFUs = '^'(10,mean(logcfu)),
                         sd = sd(logcfu)))

stats_thai_2 <- as.data.frame(summarise(group_by(filter(pro_thailandicus, hours =="0h"), replicate, log10titer),
                        mean_0h = mean(logcfu)
                       ))

mean(filter(pro_thailandicus, hours == "0h")$logcfu)

stats_thai_prol


#comparison between cibinogensis and thailandicus
pro_comp <- filter(profull, bacteria == "Ac. thailandicus" | bacteria == "Ac. cibinongensis", hours != "24h_chaser")
  
model_comp_1 <- lmer(logcfu ~ log10titer*hours*bacteria + (1 | replicate), data = pro_comp)
model_comp_2 <- lmer(logcfu ~ hours*bacteria + log10titer*bacteria + log10titer*hours + (1 | replicate), data = pro_comp)

anova(model_comp_1, model_comp_2)

model_comp_3 <- lmer(logcfu ~ hours*bacteria + log10titer*bacteria + log10titer + hours + (1 | replicate), data = pro_comp)

anova(model_comp_3, model_comp_2)

model_comp_4 <- lmer(logcfu ~ bacteria + log10titer*bacteria + log10titer + hours + (1 | replicate), data = pro_comp)
anova(model_comp_3, model_comp_4)

summary(model_comp_3)

#chaser flies
filter(profull,hours == "24h_chaser")

#L. pseudomesenteroides
pro_leuco <- filter(profull, bacteria == "L. pseudomesenteroides")

model_leuco_1 <- lmer(logcfu ~ log10titer*hours + (1 | replicate), data = pro_leuco)
model_leuco_2 <- lmer(logcfu ~ log10titer+hours + (1 | replicate), data = pro_leuco)
anova(model_leuco_1,model_leuco_2)

summary(model_leuco_2)
```

```{r load and analysis Lactobacillus colonization}
lacto <- read.csv("S10_Data.csv",sep=",",head=T)
#lacto$dilution <- as.factor(lacto$dilution)
lacto$day <- as.factor(lacto$day)

stats_lacto <- as.data.frame(summarise(group_by(lacto,bacteria, day, as.factor(dilution)),
                        median = median(logcfu),
                         mean = mean(logcfu),
                        meanCFUs = '^'(10,mean(logcfu)),
                         sd = sd(logcfu)))

stats_lacto 


#L. brevis
Lbrevis <- subset(lacto, lacto$bacteria == "L. brevis")

ggplot(Lbrevis, aes( day, logcfu)) +
  geom_boxplot(outlier.shape = NA) +
  geom_jitter(width = 0.1) +
  facet_grid(replicate ~ dilution,margins = FALSE, scales = "fixed") +
  theme_bw()

#lmer
lmer_Lbrevis_1 <- lmer(logcfu ~ day * dilution + (1 | replicate) ,data=Lbrevis)
lmer_Lbrevis_2 <- lmer(logcfu ~ day + dilution + (1 | replicate) ,data=Lbrevis)

anova(lmer_Lbrevis_1, lmer_Lbrevis_2)

summary(lmer_Lbrevis_2)

#L. paraplantarum
Lpara <- subset(lacto, lacto$bacteria == "L. paraplantarum")

ggplot(Lpara, aes( day, logcfu)) +
  geom_boxplot(outlier.shape = NA) +
  geom_jitter(width = 0.1) +
  facet_grid(replicate ~ dilution,margins = FALSE, scales = "fixed") +
  theme_bw()

#lmer
lmer_Lpara_1 <- lmer(logcfu ~ day * dilution + (1 | replicate) ,data=Lpara)
lmer_Lpara_2 <- lmer(logcfu ~ day + dilution + (1 | replicate) ,data=Lpara)

anova(lmer_Lpara_1, lmer_Lpara_2)

summary(lmer_Lpara_2)

```

###Bacteria growth on food
```{r analysis of proliferation of Acetobacter isolates on flyfood}

foodg <- read.csv("S11_Data.csv",sep=",",head=T)

foodg$replicate<-as.factor(foodg$replicate)
foodg$day<-as.factor(foodg$day)
foodg$logcfu <- log10(foodg$cfu+1)

##plot
ggplot(data=foodg,aes(x = day,y=logcfu))+
	geom_jitter(width=0.1) +
	facet_grid(bacteria ~ replicate ,margins = "False")

##linear model
#Ac. OTU2753
foodg2753 <- filter(foodg, bacteria == "Ac. OTU2753")
model2753_1 <- lmer(logcfu ~ day + (1 | replicate), data = foodg2753)
summary(model2753_1)

tukey2753 <- glht(model2753_1, mcp(day="Tukey"))
summary(tukey2753)
cld(tukey2753)

#Ac. thailandicus
foodgwild <- filter(foodg, bacteria == "Ac. thailandicus")
modelwild_1 <- lmer(logcfu ~ day + (1 | replicate), data = foodgwild)
summary(modelwild_1)

tukeywild <- glht(modelwild_1, mcp(day="Tukey"))
summary(tukeywild)
cld(tukeywild)

#Ac. cibinongensis
foodgcibi <- filter(foodg, bacteria == "Ac. cibinongensis")
head(foodgcibi)
modelcibi_1 <- lmer(logcfu ~ day + (1 | replicate), data = foodgcibi)
summary(modelcibi_1)

tukeycibi <- glht(modelcibi_1, mcp(day="Tukey"))
summary(tukeycibi)
cld(tukeycibi)

```


##Figure 5

###Analysis of proliferation in D. melanogaster and D. simulans
```{r load data and plot proliferation in different species}
#test protocol in bottles
#load data
bottle_test <- read.csv("S12_Data.csv",sep=",",head=T)
bottle_test$logcfu <- log10(bottle_test$cfu+1)
bottle_test$log10titer <- as.factor(bottle_test$log10titer)

ggplot(data=bottle_test, aes(y = logcfu, x = hours)) +
  geom_jitter(width = .3) +
  facet_grid(log10titer ~ bacteria)

#tahilandicus
bottle_test_thai <- filter(bottle_test, hours != "24h_chaser", bacteria == "Ac. thailandicus")

lm_bottle_test_thai_1 <- lm(logcfu ~ hours * log10titer, data = bottle_test_thai)
lm_bottle_test_thai_2 <- lm(logcfu ~ hours + log10titer, data = bottle_test_thai)
anova(lm_bottle_test_thai_1, lm_bottle_test_thai_2)

summary(lm_bottle_test_thai_1)

contrast_matrix_slopes_bottle_test_thai <- rbind(
  "titer3" =  c(0, 1, 0, 0), 
  "titer4" =  c(0, 1, 0, 1)
)

summary(glht(lm_bottle_test_thai_1, contrast_matrix_slopes_bottle_test_thai, adjust="holm"))

#increases with dose 3


#OTU2753
bottle_test_2753 <- filter(bottle_test, hours != "24h_chaser", bacteria == "Ac. OTU2753")

lm_bottle_test_2753_1 <- lm(logcfu ~ hours * log10titer, data = bottle_test_2753)
lm_bottle_test_2753_2 <- lm(logcfu ~ hours + log10titer, data = bottle_test_2753)
anova(lm_bottle_test_2753_1, lm_bottle_test_2753_2)

summary(lm_bottle_test_2753_1)

contrast_matrix_slopes_bottle_test_2753 <- rbind(
  "titer4" =  c(0, 1, 0, 0), 
  "titer5" =  c(0, 1, 0, 1)
)

summary(glht(lm_bottle_test_2753_1, contrast_matrix_slopes_bottle_test_2753, adjust="holm"))

#decreases in both doses



#load data of Dmel and Dsim
proms <- read.csv("S13_Data.csv",sep=",",head=T)

proms$replicate<-as.factor(proms$replicate)
proms$logcfu <- log10(proms$cfu+1)
proms$log10titer <- as.factor(proms$log10titer)
proms$hours <- as.factor(proms$hours)
proms$hours <- relevel(proms$hours,"0")

proms$stock <- factor(proms$stock, levels =c("w1118","canton-S","dmel O13","dsim A07","dsim J04","dsim O13"))
head(proms)

stats_proms <- as.data.frame(summarise(group_by(proms, stock, hours, log10titer),
                        median = median(logcfu),
                         mean = mean(logcfu),
                        meanCFUs = '^'(10,mean(logcfu)),
                         sd = sd(logcfu)))

stats_proms 

#plot
ggplot(data=filter(proms,replicate==1), aes(y = logcfu, x = hours)) +
  geom_jitter() +
  facet_grid(log10titer ~ stock)

ggplot(data=filter(proms,replicate==2), aes(y = logcfu, x = hours)) +
  geom_jitter() +
  facet_grid(log10titer ~ stock)


#test host interaction
modelms1 <- lmer(logcfu ~ host*log10titer*hours + (1 | replicate) + (1 | stock), data = proms)
modelms2 <- lmer(logcfu ~ host*log10titer + log10titer*hours + host*hours + (1 | replicate) + (1 | stock), data = proms)
anova(modelms1, modelms2)

drop1(modelms2, test = "Chisq")

modelms3 <- lmer(logcfu ~ log10titer*hours + host*hours + (1 | replicate) + (1 | stock), data = proms)

drop1(modelms3, test = "Chisq")

summary(modelms3)
lsmeansLT(modelms3)

contrast_matrix_slopes_hosts <- rbind(
  "melano_3" =    c(0, 0, 1, 0, 0, 0), 
  "melano_4" =    c(0, 0, 1, 0, 1, 0), 
  "simulans_3" =  c(0, 0, 1, 0, 0, 1), 
  "simulans_4" =  c(0, 0, 1, 0, 1, 1)
  )

summary(glht(modelms3, contrast_matrix_slopes_hosts, adjust="holm"))


#simpler model with no titer*hours interactions
modelms4 <- lmer(logcfu ~ log10titer + host*hours + (1 | replicate) + (1 | stock), data = proms)
summary(modelms4)

contrast_matrix_slopes_hosts_simple <- rbind(
  "melano" =    c(0, 0, 0, 1, 0), 
  "simulans" =  c(0, 0, 0, 1, 1)
  )

summary(glht(modelms4, contrast_matrix_slopes_hosts_simple, adjust="holm"))


#Difference between lines of Dmel
dmel <- filter(proms, host == "dmel")

lmer_dmel1 <-lmer(logcfu ~ hours*log10titer*stock + (1 | replicate), data = dmel)
lmer_dmel2 <-lmer(logcfu ~ hours*log10titer + log10titer*stock +  hours*stock + (1 | replicate), data = dmel)
anova(lmer_dmel1, lmer_dmel2)

drop1(lmer_dmel2, test = "Chisq")

lmer_dmel3 <-lmer(logcfu ~ log10titer*stock +  hours*stock + (1 | replicate), data = dmel)
drop1(lmer_dmel3, test = "Chisq")

summary(lmer_dmel3)


contrast_matrix_slopes_dmel <- rbind(
  "w1118" =     c(0, 0, 0, 0, 1, 0, 0, 0, 0), 
  "canton-S" =  c(0, 0, 0, 0, 1, 0, 0, 1, 0), 
  "dmel O13" =  c(0, 0, 0, 0, 1, 0, 0, 0, 1)
)

summary(glht(lmer_dmel3, contrast_matrix_slopes_dmel, adjust="holm"))


#Dsim
dsim <- filter(proms, host == "dsim")

lmer_dsim1 <-lmer(logcfu ~ hours*log10titer*stock + (1 | replicate), data = dsim)
lmer_dsim2 <-lmer(logcfu ~ hours*log10titer + log10titer*stock +  hours*stock + (1 | replicate), data = dsim)
anova(lmer_dsim1, lmer_dsim2)

drop1(lmer_dsim2, test = "Chisq")

lmer_dsim3 <-lmer(logcfu ~ log10titer*stock +  hours*stock + (1 | replicate), data = dsim)
drop1(lmer_dsim3, test = "Chisq")

lmer_dsim4 <-lmer(logcfu ~ log10titer +  hours*stock + (1 | replicate), data = dsim)
lmer_dsim5 <-lmer(logcfu ~ log10titer +  hours + stock + (1 | replicate), data = dsim)
anova(lmer_dsim5,lmer_dsim4)

summary(lmer_dsim5)




```

##Figure 6

###Analysis of flies monoassociated with Acetobacter species 
```{r load data and plot - flies associated with different Acetobacter}
#load data
mono <- read.csv("S14_Data.csv",sep=",",head=T)
mono$vials <- as.factor(paste(mono$bacteria, mono$vials, sep = "_"))

#make file with entry per day
m.fit<-reshape2:::melt.data.frame(mono, id=c( "condition","bacteria","vials"), variable.name="devday", value.name = "count", na.rm=TRUE)
m.fit$devday <- as.numeric(gsub("X", "", m.fit$devday))
m.fit$count <- as.numeric(m.fit$count)
head(m.fit)

##ANALYSIS OF DEVELOPMENT 
#make each entry correspond to one pupae or one adult emerged - untable
dg0ad<-untable(m.fit[,c(1,2,3,4)],num=m.fit$count)
dg0ad[1:20,]

##Statistics time to pupariation
#lm
model_pupae_mono_1 <- lm(devday ~ bacteria , data=filter(dg0ad, condition == "pupae"))
summary(model_pupae_mono_1)
tukey_pupae_mono_1 <- glht(model_pupae_mono_1, mcp(bacteria="Tukey"))
summary(tukey_pupae_mono_1)
cld(tukey_pupae_mono_1)

##Statistics time to adulthood
#lm
model_adult_mono_l <- lm(devday ~ bacteria , data=filter(dg0ad, condition == "adult"))
summary(model_adult_mono_l)
tukeyadult <- glht(model_adult_mono_l, mcp(bacteria="Tukey"))
summary(tukeyadult)
cld(tukeyadult)

lsmeans::lsmeans(model_adult_mono_l, "bacteria")

##ANALYSIS OF TOTAL PROGENY
mono$total <- rowSums(mono[,c(4:16)],na.rm = TRUE)

#Statistical analysis
#total pupae
model_tp_l<- lm(total ~ bacteria, data = filter(mono, condition == "pupae") )
summary(model_tp_l)
tukey_tp_l <- glht(model_tp_l, mcp(bacteria="Tukey"))
summary(tukey_tp_l)
cld(tukey_tp_l)

#total adults
model_ta_l<- lm(total ~ bacteria, data = filter(mono, condition == "adult"))
summary(model_ta_l)
tukey_ta_l <- glht(model_ta_l, mcp(bacteria="Tukey"))
summary(tukey_ta_l)
cld(tukey_ta_l)
```

###Analysis of flies exposed to changing environment
```{r analysis of gut bacterial levels from G0}
#load and plot data
g0gut <- read.csv("S15_Data.csv",sep=",",head=T)

#transform nc (not counteble) to 30000cfu per gut
levels(g0gut$cfu) <- c(levels(g0gut$cfu), 30000) 
g0gut$cfu[g0gut$cfu=="nc"] <- 30000
g0gut$cfu<-as.numeric(as.character(g0gut$cfu))
g0gut$logcfu <- log10(g0gut$cfu+1)

#Plot
ggplot(data=g0gut,aes(day,logcfu, fill=gender, colour=gender))+
	geom_jitter(width=.5) +
	facet_grid(bacteria ~ replicate,margins = "False")

#Statistics
summary_g0_gut_bact <- as.data.frame(summarise(group_by(g0gut, bacteria, day, gender),
                        medianCFU = median(cfu),                     
                        median = median(logcfu),
                         mean = mean(logcfu)))

summary_g0_gut_bact

#wilcoxon test because some values are high but undetermined - use rank test

wilcox.test(cfu ~ day, data = filter(g0gut, gender == "female", bacteria == "Ac. thailandicus"))
wilcox.test(cfu ~ day, data = filter(g0gut, gender == "female",  bacteria == "Ac. OTU2753"))
wilcox.test(cfu ~ bacteria, data = filter(g0gut, gender == "female", day == 10, bacteria != "GF"))
wilcox.test(cfu ~ bacteria, data = filter(g0gut, gender == "male", day == 10, bacteria != "GF"))
```

```{r analysis of bacteria from G0 transmitted to food}

#load data
bottles <- read.csv("S16_Data.csv",sep=",",head=T)

#reshape table to get one line per cfu reading
md <- melt(bottles, id=(c("replicate","bacteria", "cage")))

names(md)[names(md)=="value"] <- "colonies"
names(md)[names(md)=="variable"] <- "day"
bottles$cage<-as.factor(bottles$cage)

#remove X from day
md$day <- as.numeric(gsub("X", "", md$day))
md <- filter(md, bacteria != "GF")

md$presence[md$colonies == 0] = 0
md$presence[md$colonies > 0] = 1
md$presence[md$colonies == "nc"] = 1
md$presence[md$colonies == "No data"] <- NA
md$presence <- as.numeric(md$presence)

#statistics
#generalized linear model with binomial distribution
model1food <- glmer(presence~bacteria * day + (1|replicate) , data=md, family = binomial)
model2food <- glmer(presence~bacteria + day + (1|replicate) , data=md, family = binomial)
anova(model1food,model2food)

model3food <- glmer(presence~bacteria + (1|replicate) , data=md, family = binomial)
anova(model3food,model2food)

summary(model3food)
```

```{r analysis of number of eggs layed by G0}

#load data
eggs <- read.csv("S17_Data.csv",sep=",",head=T)

#make file with entry per day
m.eggs<-reshape2:::melt.data.frame(eggs, id=c("bacteria","cage","replicate"),variable.name="day",value.name = "eggs", na.rm=TRUE)
m.eggs$day <- as.numeric(gsub("X", "", m.eggs$day))
m.eggs$eggs <- as.numeric(m.eggs$eggs)
head(m.eggs)

#get total eggs per fly
eggs$total<-rowSums(eggs[,c(4:13)],na.rm = TRUE)
head(eggs)
boxplot(eggs$total ~eggs$bacteria)

#Analyse total number of eggs
model1eg<-lmer(total~bacteria + (1|replicate), data=eggs)
summary(model1eg)
tukey1eg<-glht(model1eg, mcp(bacteria="Tukey"))
summary(tukey1eg)
cld(tukey1eg)
```

```{r analysis of development time and total adults of F1}
f1 <- read.csv("S18_Data.csv",sep=",",head=T)

##DEVELOPMENT
#make one entry per day
m.f1<-reshape2:::melt.data.frame(f1, id=c("replicate", "day","bacteria","cage"),variable.name="devday",value.name = "adults", na.rm=TRUE)
m.f1$devday <- as.numeric(as.character(gsub("X", "", m.f1$devday)))
m.f1$adults <- as.numeric(m.f1$adults)
m.f1$day <- as.factor(m.f1$day)
m.f1$replicate <- as.factor(m.f1$replicate)

#one entry per fly
umf1<-untable(m.f1[,c(1,2,3,4,5)],num=m.f1$adults)
umf1[1:20,]

#histograms
ggplot(umf1, aes(devday, fill = bacteria)) +
  geom_histogram(breaks=seq(7, 20, by = 1), binwidth = 1, color = "black") +
  scale_x_continuous(breaks=7:20)+
  facet_grid(bacteria ~ replicate,margins = FALSE, scales = "fixed")+
  scale_fill_manual(values=c("salmon","olivedrab2","deepskyblue"))+
  theme_bw()

#Statistical analysis on developmental time to adulthood
lmer_dev <- lmer(devday~bacteria + (1|replicate/cage), data=umf1)
summary(lmer_dev)
tukey_lmer_dev <-glht(lmer_dev, mcp(bacteria="Tukey"))
summary(tukey_lmer_dev)
lsmeansLT(lmer_dev)

##TOTAL NUMBER OF ADULTS
f1$total <- rowSums(f1[,c(5:15)],na.rm = TRUE)

#remove individual days
tad <- f1[,-(5:15), drop = FALSE]
head(tad)

#dcast the days
tadcast <- reshape2::dcast(tad, replicate + bacteria + cage  ~ day, value.var = "total", na.rm = TRUE)

#make total of progeny for each pair summing all days
tadcast$total <- rowSums(tadcast[,c(4:13)],na.rm = TRUE)

#Boxplot
ggplot(tadcast, aes(bacteria, total, fill = bacteria)) +
  geom_boxplot() +
  facet_grid(. ~ replicate,margins = FALSE, scales = "fixed")+
  scale_fill_manual(values=c("salmon","olivedrab2","deepskyblue"))+
  theme_bw()

#Statistical analysis on the total number of adults
model1ta<-lmer(total~bacteria + (1|replicate) , data=tadcast)
summary(model1ta)
tukey1ta<-glht(model1ta, mcp(bacteria="Tukey"))
summary(tukey1ta)
```

```{r analysis of F1 fertility}

#load and plot data
f1fert <- read.csv("S19_Data.csv",sep=",",head=T)

#convert number of adults per day to numbers, replicate to factor
for(i in (8:22)){
  f1fert[,i]<-as.numeric(f1fert[,i])
}

f1fert$replicate<-as.factor(f1fert$replicate)

#make total adults
f1fert$total <- rowSums(f1fert[,c(8:22)],na.rm = TRUE)
head(f1fert)

#remove individual days cagepair and daypair
tadf1 <- f1fert[,-(c(4,5,8:22)), drop = FALSE]

#dcast the vial days
tadcastf1 <- reshape2::dcast(tadf1, replicate + pairingdate + bacteria + pairs ~ vialday, value.var = "total", na.rm = TRUE)

#make total of progeny for each pair summing all vial days
tadcastf1$total <- rowSums(tadcastf1[,c(5:9)],na.rm = TRUE)

#plot total per pair and during the 10 days (total of total)
ggplot(tadcastf1, aes(bacteria, total, fill = bacteria)) +
  geom_boxplot() +
  facet_grid(. ~ replicate,margins = FALSE, scales = "fixed")+
  scale_fill_manual(values=c("salmon","olivedrab2","deepskyblue"))+
  theme_bw()

#Statistical analysis on the total number of adults
model1_taf1<-lmer(total~bacteria + (1|replicate) , data=tadcastf1)
summary(model1_taf1)
tukey1_taf1<-glht(model1_taf1, mcp(bacteria="Tukey"))
summary(tukey1_taf1)
```


###Analysis of crosses between Ac. thailandicus monoassociated and GF flies
```{r crosses monoassociated and GF parents}
#load data
f1_crosses <- read.csv("S20_Data.csv",sep=",",head=T)

##DEVELOPMENT ANALYSIS
##Transform data to have one entry per developed fly
md_f1_crosses<-reshape2:::melt.data.frame(f1_crosses, id=c( "Replicate","pair","pupaeadult","condition", "vialday"),variable.name="devday",value.name = "count", na.rm=TRUE)

md_f1_crosses$Replicate<-as.factor(md_f1_crosses$Replicate)

# Expanded table with number of rows correspondent to number of flies
expanded_f1_crosses <- untable(md_f1_crosses[,c(1,2,3,4,5,6)], num=md_f1_crosses[,7])
expanded_f1_crosses$devday <- as.numeric(gsub("X", "", expanded_f1_crosses$devday))

expanded_f1_crosses$Replicate<-as.factor(expanded_f1_crosses$Replicate)

#SUBSET PUPAE AND ADULTS
pupae_f1_crosses <-filter(expanded_f1_crosses, pupaeadult == "pupae") 
adults_f1_crosses <-filter(expanded_f1_crosses, pupaeadult == "adult") 

##TOTAL PROGENY
f1_crosses$total <- rowSums(f1_crosses[,c(6:23)],na.rm = TRUE)

#remove individual days
tad_f1_crosses <- f1_crosses[,-(6:23), drop = FALSE]


#dcast the days
tadcast_f1_crosses <- reshape2::dcast(tad_f1_crosses, Replicate + pair + pupaeadult + condition  ~ vialday, value.var = "total", na.rm = TRUE)

#make total of progeny for each pair summing all days
tadcast_f1_crosses$total <- rowSums(tadcast_f1_crosses[,c(5:9)],na.rm = TRUE)

tadp_f1_crosses = filter(tadcast_f1_crosses, pupaeadult == "pupae")
tada_f1_crosses = filter(tadcast_f1_crosses, pupaeadult == "adult")


###STATISTICAL ANALYSIS

##TOTAL PROGENY
#Total pupae
model_tp_f1_crosses_l<- lmer(total~condition + (1|Replicate) , data = tadp_f1_crosses)

summary(model_tp_f1_crosses_l)
tukey_tp_f1_crosses_l <- glht(model_tp_f1_crosses_l, mcp(condition="Tukey"))
summary(tukey_tp_f1_crosses_l)
cld(tukey_tp_f1_crosses_l)

#Total adults 
model_ta_f1_crosses_l<- lmer(total~condition + (1|Replicate), data = tada_f1_crosses)
summary(model_ta_f1_crosses_l)
tukey_ta_f1_crosses_l <- glht(model_ta_f1_crosses_l, mcp(condition="Tukey"))
summary(tukey_ta_f1_crosses_l)
cld(tukey_ta_f1_crosses_l)

##DEVELOPMENTAL TIME
#lmer pupae
model_pupae_f1_crosses_l <- lmer(devday~condition + (1|Replicate) + (1|pair), data=pupae_f1_crosses)
summary(model_pupae_f1_crosses_l)
tukey_pupae_f1_crosses_l <- glht(model_pupae_f1_crosses_l, mcp(condition="Tukey"))
summary(tukey_pupae_f1_crosses_l)
cld(tukey_pupae_f1_crosses_l)

#lmer adult
model_adult_f1_crosses_l <- lmer(devday~condition + (1|Replicate) + (1|pair), data=adults_f1_crosses)
summary(model_adult_f1_crosses_l)
tukey_adult_f1_crosses_l <- glht(model_adult_f1_crosses_l, mcp(condition="Tukey"))
summary(tukey_adult_f1_crosses_l)
cld(tukey_adult_f1_crosses_l)

```


##Figure 7

###Analysis of fitness of flies associated with different natural bacterial isolates
```{r analysis of fitness of flies associated with different natural bacterial isolates}
###Analysis of development and number of adults from G0
#load data
g0_dev_mono_eggs <- read.csv("S21_Data.csv",sep=",",head=T)
g0_dev_mono_eggs$bacteria = relevel(g0_dev_mono_eggs$bacteria, ref="GF Mannitol")

##DEVELOPMENT ANALYSIS
#transform data to have one entry per developed fly
#melt table(days in rows)
m_g0_dev_mono_eggs <- melt(g0_dev_mono_eggs , id=(c("bacteria","replicate","code","vial")))
names(m_g0_dev_mono_eggs)[names(m_g0_dev_mono_eggs)=="value"] <- "adults"
names(m_g0_dev_mono_eggs)[names(m_g0_dev_mono_eggs)=="variable"] <- "day"

# Expanded table with number of rows correspondent to number of flies
expandedm_g0_dev_mono_eggs <- untable(m_g0_dev_mono_eggs[,c(1,2,3,4,5)], m_g0_dev_mono_eggs[,6])
expandedm_g0_dev_mono_eggs$day <- as.numeric(gsub("X", "", expandedm_g0_dev_mono_eggs$day))

expandedm_g0_dev_mono_eggs$replicate<-as.factor(expandedm_g0_dev_mono_eggs$replicate)
expandedm_g0_dev_mono_eggs$vial<-as.factor(expandedm_g0_dev_mono_eggs$vial)
expandedm_g0_dev_mono_eggs$unique_v<-as.factor(paste(expandedm_g0_dev_mono_eggs$bacteria, expandedm_g0_dev_mono_eggs$replicate, expandedm_g0_dev_mono_eggs$vial))

head(expandedm_g0_dev_mono_eggs)

#mixed linear model
model_g0_dev_mono_eggs <- lmer(day~bacteria + (1|unique_v) + (1|replicate), data=expandedm_g0_dev_mono_eggs)
summary(model_g0_dev_mono_eggs)

#model_g0_dev_mono_eggs2 <- lmer(day~bacteria + (1|unique_v) + (1|replicate), data=expandedm_g0_dev_mono_eggs)

tukey_g0_dev_mono_eggs <- lsmeans::lsmeans(model_g0_dev_mono_eggs, list(pairwise ~ bacteria), adjust = "tukey")
tukey_g0_dev_mono_eggs

cld(tukey_g0_dev_mono_eggs,Letters = letters[seq(1:10)])

lsmeans_g0_dev_mono_eggs <- as.data.frame(summary(tukey_g0_dev_mono_eggs)$`lsmeans of bacteria`)
lsmeans_g0_dev_mono_eggs

##TOTAL ADULTS EMERGED ANALYSIS
head(g0_dev_mono_eggs)
g0_dev_mono_eggs$total <- rowSums(g0_dev_mono_eggs[,c(5:15)],na.rm = TRUE)
g0_dev_mono_eggs$replicate <- as.factor(g0_dev_mono_eggs$replicate)

#lm
lmer_adult_total_mono_eggs <- lmer(total~ bacteria + (1|replicate) , data=g0_dev_mono_eggs)
tukey_adult_total_mono_eggs <- lsmeans::lsmeans(lmer_adult_total_mono_eggs, list(pairwise ~ bacteria), adjust = "tukey")
tukey_adult_total_mono_eggs

cld(tukey_adult_total_mono_eggs)

###FERTILITY G0 ADULTS
#load and plot data
g0_fertility_mono_eggs <- read.csv("S22_Data.csv",sep=",",head=T)
g0_fertility_mono_eggs$bacteria <- relevel(g0_fertility_mono_eggs$bacteria, ref="GF Mannitol")
g0_fertility_mono_eggs$replicate <- as.factor(g0_fertility_mono_eggs$replicate)

ggplot(g0_fertility_mono_eggs, aes(bacteria, adults, fill = bacteria)) +
  geom_boxplot() +
  facet_grid(. ~ replicate,margins = FALSE, scales = "fixed")+
  theme_bw()

#lmer
model_g0_fertility_mono_eggs <- lmer(adults~ bacteria + (1|replicate), g0_fertility_mono_eggs)

summary(model_g0_fertility_mono_eggs)

tukey_g0_fertility_mono_eggs <- lsmeans::lsmeans(model_g0_fertility_mono_eggs, list(pairwise ~ bacteria), adjust = "tukey")
tukey_g0_fertility_mono_eggs
cld(tukey_g0_fertility_mono_eggs, Letters = letters[seq(1:10)])

lsmeans_g0_fertility_mono_eggs <- as.data.frame(summary(tukey_g0_fertility_mono_eggs)$`lsmeans of bacteria`)


### analysis of correlation between development time and fertility

correlation_data <- merge(lsmeans_g0_fertility_mono_eggs[,1:2], lsmeans_g0_dev_mono_eggs[,1:2], by = "bacteria", all = TRUE)
names(correlation_data)[names(correlation_data) == 'lsmean.x'] <- 'fertility'
names(correlation_data)[names(correlation_data) == 'lsmean.y'] <- 'development_time'
head(correlation_data)


ggplot(correlation_data, aes(development_time, fertility)) +
  geom_point() +
  geom_text(data = correlation_data, aes(label = bacteria), vjust = -1.5,  size = 1.5) +
  theme_bw()

cor.test(correlation_data$fertility, correlation_data$development_time, method = c("pearson"))

#wihtout GF samples statistics are very similar
correlation_data_bact_only <- filter(correlation_data, bacteria != "GF", bacteria != "GF Mannitol", bacteria != "GF MRS")

cor.test(correlation_data_bact_only$fertility, correlation_data_bact_only$development_time, method = c("pearson"))

```

###Fitness analysis of flies grown on sterile fig homogenate
```{r analysis of development and number of adults from G0 in sterile figs homogenates}

#load and plot data
g0_fig_dev_sterile <- read.csv("S23_Data.csv",sep=",",head=T)
g0_fig_dev_sterile$bacteria = relevel(g0_fig_dev_sterile$bacteria, ref="GF")

##DEVELOPMENT
m_g0_fig_dev_sterile <- melt(g0_fig_dev_sterile, id=(c("bacteria","repvial","replicate")))
names(m_g0_fig_dev_sterile)[names(m_g0_fig_dev_sterile)=="value"] <- "adults"
names(m_g0_fig_dev_sterile)[names(m_g0_fig_dev_sterile)=="variable"] <- "day"
m_g0_fig_dev_sterile$replicate<-as.factor(m_g0_fig_dev_sterile$replicate)
m_g0_fig_dev_sterile$repvial<-as.factor(m_g0_fig_dev_sterile$repvial)

# Expanded table with number of rows correspondent to number of flies
e_g0_fig_dev_sterile <- untable(m_g0_fig_dev_sterile[,c(1,2,3,4)], num=m_g0_fig_dev_sterile[,5])

#Take out "X" from day column
e_g0_fig_dev_sterile$day <- as.numeric(gsub("X", "", e_g0_fig_dev_sterile$day))

##Plot
ggplot(e_g0_fig_dev_sterile, aes( bacteria, day)) +
  geom_boxplot() +
  facet_grid(. ~ replicate,margins = FALSE, scales = "fixed")+
  theme_bw()
  
##STATISTICS
#Developmental time
model_g0_fig_dev_sterile <- lmer(day ~ bacteria + (1|replicate), data = e_g0_fig_dev_sterile )
summary(model_g0_fig_dev_sterile)
lsmeansLT(model_g0_fig_dev_sterile)
tukey_model_g0_fig_dev_sterile<-glht(model_g0_fig_dev_sterile, mcp(bacteria="Tukey"))
cld(tukey_model_g0_fig_dev_sterile)
summary(tukey_model_g0_fig_dev_sterile)

#total number of adults
head(g0_fig_dev_sterile)
g0_fig_dev_sterile$total <- rowSums(g0_fig_dev_sterile[,c(4:30)],na.rm = TRUE)

#plot
ggplot(g0_fig_dev_sterile, aes( bacteria, total)) +
  geom_boxplot(outlier.shape = NA) +
  geom_jitter(width = 0.1) +
  facet_grid(. ~ replicate,margins = FALSE, scales = "fixed") +
  theme_bw()

#lmer
model_g0_fig_total_sterile<-lmer( total ~ bacteria + (1|replicate), data = g0_fig_dev_sterile)
summary(model_g0_fig_total_sterile)
lsmeansLT(model_g0_fig_total_sterile)
tukey_model_g0_fig_total_sterile<-glht(model_g0_fig_total_sterile, mcp(bacteria="Tukey"))
cld(tukey_model_g0_fig_total_sterile)
summary(tukey_model_g0_fig_total_sterile)

#G0 fertility from sterile figs homogenate
#load data
g0_fertility_fig_sterile <- read.csv("S24_Data.csv",sep=",",head=T)
g0_fertility_fig_sterile$bacteria = relevel(g0_fertility_fig_sterile$bacteria, ref="GF")
g0_fertility_fig_sterile$replicate<-as.factor(g0_fertility_fig_sterile$replicate)
g0_fertility_fig_sterile$vialrep<-as.factor(g0_fertility_fig_sterile$vialrep)
g0_fertility_fig_sterile

#plot
ggplot(g0_fertility_fig_sterile, aes( bacteria, adults)) +
  geom_boxplot(outlier.shape = NA) +
  geom_jitter(width = 0.1) +
  facet_grid(. ~ replicate,margins = FALSE, scales = "fixed")+
  theme_bw()

#Statistics
model_g0_fertility_fig_sterile <- lmer(adults ~ bacteria + (1|replicate) ,data=g0_fertility_fig_sterile)
summary(model_g0_fertility_fig_sterile)
tukey_model_g0_fertility_fig_sterile<-glht(model_g0_fertility_fig_sterile, mcp(bacteria="Tukey"))
cld(tukey_model_g0_fertility_fig_sterile)
summary(tukey_model_g0_fertility_fig_sterile)
```


###Fitness analysis of flies grown on collected figs
```{r analysis of G0 development and total number of adults}
#Load data and process
g0_dev_fig_tree <- read.csv("S25_Data.csv",sep=",",head=T)
g0_dev_fig_tree$bacteria = relevel(g0_dev_fig_tree$bacteria, ref="GF")
g0_dev_fig_tree$rep <- as.factor(g0_dev_fig_tree$rep)

g0_dev_fig_tree$Unique_fig <- paste(g0_dev_fig_tree$rep,g0_dev_fig_tree$fig,sep="_")
g0_dev_fig_tree$Unique_repvial <- paste(g0_dev_fig_tree$rep,g0_dev_fig_tree$repvial,g0_dev_fig_tree$bacteria,sep="_")

##Process data for DEVELOPMENT analysis
m_g0_dev_fig_tree <- reshape2::melt(g0_dev_fig_tree, id=(c("rep","repvial","fig","bacteria","Unique_fig","Unique_repvial")))
names(m_g0_dev_fig_tree)[names(m_g0_dev_fig_tree)=="value"] <- "adults"
names(m_g0_dev_fig_tree)[names(m_g0_dev_fig_tree)=="variable"] <- "day"
m_g0_dev_fig_tree$Unique_fig<-as.factor(m_g0_dev_fig_tree$Unique_fig)
m_g0_dev_fig_tree$Unique_repvial<-as.factor(m_g0_dev_fig_tree$Unique_repvial)

# Expanded table with number of rows correspondent to number of flies
e_g0_dev_fig_tree <- reshape::untable(m_g0_dev_fig_tree[,c(1:7)], num=m_g0_dev_fig_tree[,8])

#Take out "X" from day column
e_g0_dev_fig_tree$day <- as.numeric(gsub("X", "", e_g0_dev_fig_tree$day))

##Process data for TOTAL number analysis
g0_dev_fig_tree$total <- rowSums(g0_dev_fig_tree[,c(5:21)],na.rm = TRUE)
g0_dev_fig_tree$Unique_fig <- as.factor(g0_dev_fig_tree$Unique_fig)

#Statistics
#total number of adults

model_total_fig_tree_1 <- lmer(total ~ bacteria * rep + (1|Unique_fig), data= g0_dev_fig_tree)
model_total_fig_tree_2 <- lmer(total ~ bacteria + rep + (1|Unique_fig), data= g0_dev_fig_tree)

anova(model_total_fig_tree_1, model_total_fig_tree_2)

summary(model_total_fig_tree_2)

# development time

model_dev_fig_tree_1 <- lmer(day~ bacteria * rep + (1|Unique_repvial) + (1|Unique_fig), data= e_g0_dev_fig_tree)
model_dev_fig_tree_2 <- lmer(day~ bacteria + rep + (1|Unique_repvial) + (1|Unique_fig), data= e_g0_dev_fig_tree)
anova(model_dev_fig_tree_1, model_dev_fig_tree_2)

summary(model_dev_fig_tree_1)

lsmeansLT(model_dev_fig_tree_1)
pairs(lsmeans::lsmeans(model_dev_fig_tree_1, list(pairwise ~ bacteria | rep)), adjust="holm")

```