-
Notifications
You must be signed in to change notification settings - Fork 1
/
functions.R
224 lines (219 loc) · 9.33 KB
/
functions.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
######## Functions prepared by Zaid Abdo - 12/2017 - Microbial Genomics/Metagenomics Data Analysis
### Species bar plot per sample using most abundant in sample
## This function requires an otu table (on any taxonomic level) in the form of a data frame
## It only plots the most abundant based on the cutoff given (the cutoff defaults to 0 and hence plots everything)
## x label defaults to "Sample" but can be changed, and y label defaults to "OTU" but can be changed
## The function plots proportions and it is best to use the non-normalized data for this. It doesn't split the plot
## per treatment level; but combines all samples together.
bar.taxa.sample.ftn = function(d.df,cutoff=0,xlab="Sample",llab="OTU"){
d.rs = apply(d.df,1,sum)
ln = length(d.rs)
p.df = d.df/d.rs
if(cutoff > 0){
cut.b = apply(p.df,2,function(x)sum(x>cutoff))
p.df = p.df[,cut.b > 0]
p.rs = apply(p.df,1,sum)
p.df = p.df/p.rs
}
nm = colnames(p.df)
r.nm = row.names(p.df)
sp = rep(nm,ln)
sample = w = c()
for(i in 1:ln){
sample = c(sample,rep(r.nm[i],length(nm)))
w = c(w,t(p.df[i,]))
}
hist.df = data.frame(sample,sp,w)
print(qplot(data=hist.df,x=sample,fill=sp,geom="bar",weight=w)
+geom_bar(color="white",linetype=1)
+labs(x=xlab,y="Proportion")+scale_fill_discrete(name=llab)
+theme(axis.text.x=element_text(angle=0))
+scale_color_manual(values = palette(rainbow(19)))
+guides(fill=guide_legend(ncol=1))
+theme_minimal())
}
### Species bar plot per sample using most abundant in sample split by treatment level
## This function requires an otu table (on any taxonomic level) in the form of a data frame
## It only plots the most abundant based on the cutoff given (the cutoff defaults to 0.1)
## title defaults to "Experiment", x label defaults to "Sample" but can be changed, and
## y label defaults to "OTU" but can be changed. The function plots proportions and it is
## best to use the non-normalized data for this.
## It does splits the plot per treatment level as provided by Trt, which is required. Trt is a data frame
## that includes either 2 columns (the first is a treatment level and the second is the sample names) or
## 3 columns (first two columns include treatment levels for two factors and last includes sample names)
bar.sample.trt.ftn = function(data,Trt, cutoff=0.01, title="Experiment",xlab="Sample",llab="OTU"){
lt = length(Trt[1,])
nm = colnames(data)
# reducing the plot df using the lowest treatment
trt.ls = list()
for(i in 1:lt) trt.ls[[i]] = levels(factor(Trt[,i]))
if(lt == 2){
trt = p.df = c()
for(i in 1:length(trt.ls[[1]])){
for(j in 1:length(trt.ls[[2]])){
trt1 = c(trt.ls[[1]][i],trt.ls[[2]][j])
p.df = rbind(p.df, apply(data[Trt[,1]==trt.ls[[1]][i]&Trt[,2]==trt.ls[[2]][j],],2,sum))
trt = rbind(trt,trt1)
}
}
colnames(trt) = c("trt1","trt2")
Trt = trt
rs.vt = apply(p.df,1,sum)
Trt = Trt[rs.vt > 0, ]
p.df = p.df[rs.vt > 0, ]
rs.vt = rs.vt[rs.vt>0]
colnames(p.df) = nm
ln = length(p.df[,1])
p.df = p.df/rs.vt
if(cutoff > 0){
cut.b = apply(p.df,2,function(x)sum(x>cutoff))
p.df = p.df[,cut.b > 0]
p.rs = apply(p.df,1,sum)
p.df = p.df/p.rs
}
nm.trt = colnames(Trt)
hist.df = sp = w = c()
for(j in 1:length(p.df[1,])){
sp = c(sp,rep(nm[j],ln))
w = c(w,p.df[,j])
hist.df = rbind(hist.df,Trt)
}
trt1 = factor(hist.df[,1])
trt2 = factor(hist.df[,2])
hist.df = data.frame(trt1,trt2,sp,w)
hist.df = hist.df[order(hist.df$trt1,hist.df$trt2),]
print(ggplot(data=hist.df,aes(x=trt2,fill = sp))
+geom_bar(aes(weight=w),color="white",linetype=1)
+facet_wrap(~trt1,scales = "free")
+labs(title=title,colour=llab,x=xlab,y="Proportion")
+scale_fill_discrete(name = llab)
+theme_minimal()
+guides(fill=guide_legend(ncol=1))
+theme(axis.text.x=element_text(angle=90)))
}else if(lt==3){
trt = p.df = c()
for(i in 1:length(trt.ls[[1]])){
for(j in 1:length(trt.ls[[2]])){
for(k in 1:length(trt.ls[[3]])){
trt1 = c(trt.ls[[1]][i],trt.ls[[2]][j],trt.ls[[3]][k])
p.df = rbind(p.df, apply(data[Trt[,1]==trt.ls[[1]][i]&Trt[,2]==trt.ls[[2]][j]&Trt[,3]==trt.ls[[3]][k],],2,sum))
trt = rbind(trt,trt1)
}
}
}
colnames(trt) = c("trt1","trt2","trt3")
Trt = trt
rs.vt = apply(p.df,1,sum)
Trt = Trt[rs.vt > 0, ]
p.df = p.df[rs.vt > 0, ]
rs.vt = rs.vt[rs.vt>0]
colnames(p.df) = nm
ln = length(p.df[,1])
p.df = p.df/rs.vt
if(cutoff > 0){
cut.b = apply(p.df,2,function(x)sum(x>cutoff))
p.df = p.df[,cut.b > 0]
p.rs = apply(p.df,1,sum)
p.df = p.df/p.rs
}
nm.trt = colnames(Trt)
hist.df = sp = w = c()
for(j in 1:length(p.df[1,])){
sp = c(sp,rep(nm[j],ln))
w = c(w,p.df[,j])
hist.df = rbind(hist.df,Trt)
}
trt1 = factor(hist.df[,1])
trt2 = factor(hist.df[,2])
trt3 = factor(hist.df[,3])
hist.df = data.frame(trt1,trt2,trt3,sp,w)
#hist.df = hist.df[order(hist.df$trt1,hist.df$trt2,hist.df$trt3),]
print(ggplot(data=hist.df,aes(x=trt3,fill=sp))
+geom_bar(aes(weight=w),color="white",linetype=1)
+facet_wrap(trt1~trt2,scales = "free")
+labs(title=title,colour=llab,x=xlab,y="Proportion")
+scale_fill_discrete(name = llab)
+theme_minimal()
+guides(fill=guide_legend(ncol=1))
+theme(axis.text.x=element_text(angle=90)))
}else{
print("Too many treatment levels")
}
}
######### Clustering and heatmaps using pheatmap
### This function plots heatmaps and clusters the different samples identifying the different treatment levels
### per sample. It requires an otu table (on any taxonomic level) in a data frame format, and the experimental
### design (meta.df data frame) with sample names as last column and either 1 factor or two factors (meta.df
### with either 2 or three columns). Number of clusters is set to three by defauls but can be changed, in will
### cluster by column if col.clust = T using euclidian distance but col.clust can equal to a clustering using
### the function hclust (see Analysis-Normalization-Ordination-Clustering.R file). Cell height and width default
### to 10 but can be changed and font size defaults to 10 but can be changed.
pheatmap.3.ftn = function(d.df,trt.df, cutoff=3,col.clust = T, cellh=10,cellw=10,fonts=10){
if(length(trt.df)==2){
Trt = trt.df[,1,drop=FALSE]
print(pheatmap(as.matrix(d.df),
cluster_rows = F,
cluster_cols = col.clust,
cutree_rows = cutoff,
cellwidth = cellw,
cellheight = cellh,
annotation_col = Trt,
fontsize = fonts,
height = 7,
width = 6.5))
}else{
Trt = trt.df[,1:2]
print(pheatmap(as.matrix(d.df),
cluster_rows = F,
cluster_cols = col.clust,
cutree_rows = cutoff,
cellwidth = cellw,
cellheight = cellh,
annotation_col = Trt,
fontsize = fonts,
height = 7,
width = 6.5))
}
}
###### Non metric multidimensional scalling ordination plot
#### This function creates an ordination plot using an otu table provided by the user and a vector of treatment levels
#### that is formated as factor. The otu table can be provided on any taxonomic level. The title defaults to "Experiment"
#### but can be changed, the dimensions of the plot are set by x and y (default between -1 and 1 but should be changed)
#### there are 7 types that can be used for plotting (numbers 1-7). font size can be set by cex and line width by lwd.
#### distance used in ordination can be set by distance and defaults to "bray" (bray-curtis)
#### plot types: 1) elipsoides colored by treatment, 2) only sample names with color by treatment,
#### 3) spider (ray) plot colored by treatment 4) elipsoide and sample names, 5) spider and sample names,
#### 6) elipsoid and spider, and 7) elipsoide, spider and sample names.
ord.plot.ftn <- function(d.df,Trt,title="Experiment",x=c(-1,1),y=c(-1,1),type=1,cex=0.5,lwd=1,distance="bray"){
l = levels(factor(Trt))
c = seq(1,length(l))
c1 = c()
for(i in 1:length(l)){
c1 = c(c1,rep(i,length(Trt[Trt==l[i]])))
}
s <- apply(d.df,2,sum)
d.df <- d.df[,s>0]
ord <- metaMDS(d.df,distance=distance,try=100,trymax=10000)
plot(x=NULL,y=NULL,xlim=x,ylim=y,xlab="NMDS1", ylab="NMDS2")
title(main = title)
if(type==1){
ordiellipse(ord, Trt, col=c,lwd=lwd,label=TRUE,cex=cex)
}else if(type == 2){
text(ord,display="site",cex=cex,col=c1)
}else if(type == 3){
ordispider(ord, Trt, col=c, cex = cex, label = TRUE)
}else if(type == 4){
text(ord,display="site",cex=cex,col=c1)
ordiellipse(ord, Trt, col=c,lwd=lwd,label=TRUE)
}else if(type == 5){
text(ord,display="site",cex=cex,col=c1)
ordispider(ord, Trt, col=c, cex = cex, label = TRUE)
}else if(type==6){
ordiellipse(ord, Trt, col=c,lwd=lwd,label=TRUE)
ordispider(ord, Trt, col=c, cex = cex, label = FALSE)
}else{
text(ord,display="site",cex=cex,col=c1)
ordiellipse(ord, Trt, col=c,lwd=lwd,label=TRUE)
ordispider(ord, Trt, col=c, cex = cex, label = FALSE)
}
}