# Script to test whether birds from a WWTP have a more diverse resistome, and produce nice graphs.

rm(list=ls(all=TRUE))

library(coin)
library (rcompanion)
library(ggplot2)

variables <- read.csv("variables.csv")

variables$Impact <- factor(variables$Impact,c("WTP","Other"))
variables$Loc <- factor(variables$Loc,c("WTP","Yallock Creek","King Island","Innaminkca","Kopaitik","GGV"))
variables$Order <- factor(variables$Order,c("Anseriformes","Charadriiformes","Sphenisciformes"))
variables$Impact2 <- factor(variables$Impact2,c("High","Medium","Low"))


#### Plots by pollution category #### 
bl <- "#4271AE"
gr = "grey56"

Impact.summary <- aggregate(. ~ Impact, mean, data=variables)


plot_AMRg <- ggplot(variables, aes(x = Impact, y = AMR_genes, color = Impact)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Impact.summary,aes(ymin = AMR_genes, ymax = AMR_genes),
                size=.7, width = .6) + theme_bw() + scale_color_grey(start=0, end=0.6)
plot_AMRg               


plot_antib <-  ggplot(variables, aes(x = Impact, y = Antibiotics, color = Impact)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Impact.summary,aes(ymin = Antibiotics, ymax = Antibiotics),
                size=.7, width = .6) + theme_bw() + scale_color_grey(start=0, end=0.6)
plot_antib

plot_AMRabund <- ggplot(variables, aes(x = Impact, y = AMR_abundance, color = Impact)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Impact.summary,aes(ymin = AMR_abundance, ymax = AMR_abundance),
                size=.7, width = .6) + theme_bw() + scale_color_grey(start=0, end=0.6)
plot_AMRabund


####  Plots by loc, using impact categories (low, medium and high) to draw median lines 
Impact2.summary <- aggregate(. ~ Impact2, mean, data=variables)

#AMR genes x Loc
ggplot(variables, aes(x = Loc, y = AMR_genes, color = Impact2)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Impact2.summary, aes(ymin = AMR_genes, ymax = AMR_genes),
                size=.7, width = 1) + theme_bw() + scale_color_brewer(palette="Set1")
  
# AMR Antibiotic claasses
ggplot(variables, aes(x = Loc, y = Antibiotics, color = Impact2)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Impact2.summary, aes(ymin = Antibiotics, ymax = Antibiotics),
                size=.7, width = 1) + theme_bw() + scale_color_brewer(palette="Set1")

# AMR abundance
ggplot(variables, aes(x = Loc, y = AMR_abundance, color = Impact2)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Impact2.summary, aes(ymin = AMR_abundance, ymax = AMR_abundance),
                size=.7, width = 1) + theme_bw() + scale_color_brewer(palette="Set1")



####  Plots by Bird Order ####
Order.summary <- aggregate(. ~ Order, mean, data=variables)

#AMR div
ggplot(variables, aes(x = Order, y = AMR_genes, color = Order)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Order.summary, aes(ymin = AMR_genes, ymax = AMR_genes),
                size=.7, width = 1) + theme_bw() + scale_color_brewer(palette="Accent")

#AMR Antibiotic classes
ggplot(variables, aes(x = Order, y = Antibiotics, color = Order)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Order.summary, aes(ymin = Antibiotics, ymax = Antibiotics),
                size=.7, width = 1) + theme_bw() + scale_color_brewer(palette="Accent")

#AMR Abundance
ggplot(variables, aes(x = Order, y = AMR_abundance, color = Order)) + geom_jitter(width=0.2, size=10) +
  geom_crossbar(data=Order.summary, aes(ymin = AMR_abundance, ymax = AMR_abundance),
                size=.7, width = 1) + theme_bw() + scale_color_brewer(palette="Accent")




### Statistics ###

## Impact ##
#kruskal-wallis test
kw_AMR_genes <- kruskal.test(variables$AMR_genes ~ variables$Impact)
kw_AMR_genes
#data:  variables$AMR_genes by variables$Impact
#Kruskal-Wallis chi-squared = 7.1296, df = 1, p-value = 0.007582

kw_antib <- kruskal.test(variables$Antibiotics ~ variables$Impact)
kw_antib
#data:  variables$Antibiotics by variables$Impact
#Kruskal-Wallis chi-squared = 4.353, df = 1, p-value = 0.03694

kw_AMR_ab <- kruskal.test(variables$AMR_abundance ~ variables$Impact)
kw_AMR_ab
#data:  variables$AMR_abundance by variables$Impact
#Kruskal-Wallis chi-squared = 7, df = 1, p-value = 0.008151


########
# Test for differences between 10 birds from the WWTP and 10 birds from PB
# VRMarcelino - 13-Sep-2018

rm(list=ls(all=TRUE))

library(ggplot2)

variables <- read.csv("00_AMRcounts.csv")
rownames(variables) <- variables[,1]
variables <- variables[,2:9]

variables$Impact <- factor(variables$Impact,c("WWTP","Pristine"))


## Diversity of genes
#kruskal-wallis test
kw_AMR_genes <- kruskal.test(variables$AMR_diversity ~ variables$Impact)
kw_AMR_genes
# Kruskal-Wallis chi-squared = 9.2551, df = 1, p-value = 0.002348


########

# Script to produce a PCoA of the pathways found in bird microbiome,
# and test differences with random forest.
# VRMarcelino - 06 - Feb - 2018

rm(list=ls(all=TRUE))

library(ggplot2)
library(ape) # for pcoa function
library(randomForest)
library (rcompanion)


# Read data
paths_raw <- read.table("paths_ednames.tsv")
paths_raw <- t(paths_raw)
rownames(paths_raw) <- paths_raw[,1]
colnames(paths_raw) <- paths_raw[1,]
paths <- paths_raw[-1,-1]

# Remove Unintegrated and unmapped:
paths <- subset(paths, select = -c(UNINTEGRATED, UNMAPPED))

variables <- read.csv("variables.csv")
rownames(variables) <- variables[,1]


# Normalise for extreme distributions using the shifted logarithm (log2(n + 1)).
class(paths) <- "numeric"
paths_log <- log2(paths + 1)

dm <- dist(paths_log)

path.pcoa <- pcoa(dm)

# Merge data
PCi <- merge(data.frame(path.pcoa$vectors), variables, by="row.names")

# Plot Bird order and Impact
Order = PCi$Order
Impact <- PCi$Impact

# Order and Impact
ggplot(PCi,aes(x=Axis.1,y=Axis.2,group = interaction(Order,Impact))) + geom_point(aes(colour = Order, shape = Impact),size=5,alpha=0.8) + theme_classic()


# Random Forests

# merge tables (make sure they are in order!)
# Do not use logs (Random Forests does not require normalisation)
paths_and_vars_ord <- cbind(Order=as.character(variables$Order), paths)
paths_and_vars_imp <- cbind(Impact=as.character(variables$Impact), paths)

## Random Forest for Bird Order:
set.seed(71)
order_rf <- randomForest(Order ~ ., data = paths_and_vars_ord, ntree = 1000, importance=TRUE)
print(order_rf)  

#calculate importance of pathways in discriminating between groups:
imp_Order <- importance(order_rf,type=1)
write.csv(imp_Order, "imp_Order.csv")

## Impact
imp_rf <- randomForest(Impact ~ ., data = paths_and_vars_imp, ntree = 1000, importance=TRUE)
print(imp_rf)  

imp_Impact <- importance(imp_rf,type=1)
write.csv(imp_Impact, "imp_Impact.csv")

### Test correlation between bacterial reads and R genes:
# for spearman / pearson test - replace ''pearson with spearman'

### Test correlation between library size and R genes
ggscatter(variables, x = "lib_size", y = "AMR_genes", 
          add = "reg.line", conf.int = TRUE, 
          cor.coef = TRUE, cor.method = "pearson",
          xlab = "lib_size", ylab = "AMR gene diversity")

ggscatter(variables, x = "lib_size", y = "AMR_abundance", 
          add = "reg.line", conf.int = TRUE, 
          cor.coef = TRUE, cor.method = "pearson",
          xlab = "lib_size", ylab = "AMR_abundance")

ggscatter(variables, x = "lib_size", y = "Antibiotics", 
          add = "reg.line", conf.int = TRUE, 
          cor.coef = TRUE, cor.method = "pearson",
          xlab = "lib_size", ylab = "Antibiotics")

### Test correlation between bacterial reads and R genes:
ggscatter(variables, x = "bac_reads", y = "AMR_genes", 
          add = "reg.line", conf.int = TRUE, 
          cor.coef = TRUE, cor.method = "pearson",
          xlab = "microbial_reads", ylab = "AMR gene diversity")

ggscatter(variables, x = "bac_reads", y = "AMR_abundance", 
          add = "reg.line", conf.int = TRUE, 
          cor.coef = TRUE, cor.method = "pearson",
          xlab = "microbial_reads", ylab = "AMR_abundance")

ggscatter(variables, x = "bac_reads", y = "Antibiotics", 
          add = "reg.line", conf.int = TRUE, 
          cor.coef = TRUE, cor.method = "pearson",
          xlab = "microbial_reads", ylab = "Antibiotics")


####################### confounding tests:
# Script to test for confounding effects of library size size (sequencing depth)

rm(list=ls(all=TRUE))

library(ggplot2)
library(lme4)
library(ggpubr)

variables <- read.csv("variables.csv")

variables$Impact <- factor(variables$Impact,c("WTP","Other"))
variables$Loc <- factor(variables$Loc,c("WTP","Yallock Creek","King Island","Innaminkca","Kopaitik","GGV"))
variables$Order <- factor(variables$Order,c("Anseriformes","Charadriiformes","Sphenisciformes"))
variables$Impact2 <- factor(variables$Impact2,c("High","Medium","Low"))

###### Nested model ######

#### AMR genes ####
model1_no_conf <- lm(log(AMR_genes) ~ Impact, data=variables)
summary(model1_no_conf)
model1 <- lm(log(AMR_genes) ~ Impact + lib_size, data=variables)
summary(model1)
plot(model1)

#### Abundance ####
model2_no_conf <- lm(log(AMR_abundance) ~ Impact, data=variables)
summary(model2_no_conf)
model2 <- lm(log(AMR_abundance) ~ Impact + lib_size, data=variables)
summary(model2)
plot(model2)


#### Antibiotic classes ####
model3_no_conf <- lm(log(Antibiotics) ~ Impact, data=variables)
summary(model3_no_conf)
model3 <- lm(log(Antibiotics) ~ Impact + lib_size, data=variables)
summary(model3)
plot(model3)