-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathpre-processing.Rmd
More file actions
314 lines (221 loc) · 6.16 KB
/
pre-processing.Rmd
File metadata and controls
314 lines (221 loc) · 6.16 KB
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
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
---
title: "Pre-processing"
author: "Mark Dunning"
date: "20/11/2015"
output: html_document
---
# Pre-processing concerns for Microarray data
## Image-Processing
+ Calculate foreground & background
+ Background correct
+ The resulting values are found in the text files
+ The steps can be repeated in beadarray
+ *Usually we start with these values*
# Quality Assessment
## Spatial artefacts
Sometimes obvious artefacts can be apparent on the TIFF image itself
```{r}
library(affy)
setwd(system.file("extdata",package="estrogen"))
badc <- ReadAffy("bad.cel")
image(badc)
```
[http://plmimagegallery.bmbolstad.com/](A gallery of interesting / unusual artefacts)
***BASH*** method for Illumina
## Typical plots
- Boxplots of array distributions
```{r echo=FALSE,warning=FALSE,fig.height=4,fig.width=8,cache=TRUE,message=FALSE}
library(estrogen)
suppressPackageStartupMessages(library(affy))
targ <- paste0(system.file("extdata",package = "estrogen"),"/estrogen.txt")
pd <- read.AnnotatedDataFrame(targ,header=TRUE,sep="",row.names=1)
rawdata <- ReadAffy(filenames=paste(system.file("extdata",package = "estrogen"),rownames(pData(pd)),sep="/"),phenoData=pd)
boxplot(rawdata[,1:4])
```
## Typical plots
- MA plots
```{r echo=FALSE,fig.height=4,fig.width=8}
suppressPackageStartupMessages(library(affyPLM))
par(mfrow=c(1,3))
MAplot(rawdata[,1:3])
```
##Typical plots
- Density plots
```{r echo=FALSE,warning=FALSE,cache=TRUE}
eset <- expresso(rawdata, bgcorrect.method="rma",
normalize.method="constant",pmcorrect.method="pmonly",
summary.method="avgdiff")
plotDensity(log2(exprs(eset)))
```
## Normalisation
- We want to be observing *biological* and not *technical* variation
- We wouldn't expect such wholesale changes on a per-sample basis
- Easy option would to scale values for each array to median level
```{r echo=FALSE,warning=FALSE,fig.height=4,fig.width=8}
boxplot(rawdata)
x <- rawdata
tmp <- unlist(indexProbes(x, which="both"))
tmp <- tmp[seq(1, length(tmp), len = 5000)]
df <- data.frame(log2(intensity(x)[tmp, ]))
med <- median(as.matrix(df))
abline(h=med,lty=2,col="red")
```
## Simple scaling
- Genes on array 2 are on average `r median(df[,2]) - med` higher than the global median, so subtract `r median(df[,2]) - med` from each gene
- Genes on array 8 are on average `r abs(median(df[,8]) - med)` lower than the global median, so add `r abs(median(df[,8]) - med)` to each gene
- etc
## Non-linear effects
- We often compare to an *average* array which is the result of averaging each gene
- Different effects can be seen when comparing to this theoretical array
## Quantile normalisation
Consider the following matrix of values to be normalised
```{r echo=FALSE}
df <- data.frame(Array1 = c(1,3,9,2,4), Array2 = c(3,4,2,1,9), Array3 = c(9,1,5,7,6))
rownames(df) <- LETTERS[1:nrow(df)]
df
```
Genes A, B, C, D and E measured on three arrays
## Quantile normalisation
Determine ranks of each column
```{r echo=FALSE}
df
```
```{r echo=FALSE}
rks <- apply(df, 2, function(x) paste("Rank",rank(x,ties.method="min"),sep=""))
rks
```
##Quantile normalisation
Sort each column Largest...Smallest
Original data
```{r echo=FALSE}
df
```
***
Sorted data
```{r echo=FALSE}
apply(df, 2,sort)
```
Then calculate target distribution by averaging the sorted rows
```{r echo=FALSE}
target <- round(rowMeans(apply(df, 2,sort)),3)
names(target) <- paste("Rank", 1:length(target),sep="")
target
```
##Quantile normalisation
Go back to the rank matrix
```{r echo=FALSE}
rks
```
Substitue with values from the target distribution
```{r echo=FALSE}
target
```
```{r echo=FALSE}
rks[,1] <- gsub("Rank1",target["Rank1"],rks[,1])
rks
```
##Quantile normalisation
Go back to the rank matrix
```{r echo=FALSE}
rks
```
Substitue with values from the target distribution
```{r echo=FALSE}
target
```
```{r echo=FALSE}
rks[,1] <- gsub("Rank2",target["Rank2"],rks[,1])
rks
```
##Quantile normalisation
Go back to the rank matrix
```{r echo=FALSE}
rks
```
Substitue with values from the target distribution
```{r echo=FALSE}
target
```
```{r echo=FALSE}
rks[,1] <- gsub("Rank3",target["Rank3"],rks[,1])
rks
```
##Quantile normalisation
Go back to the rank matrix
```{r echo=FALSE}
rks
```
Substitue with values from the target distribution
```{r echo=FALSE}
target
```
```{r echo=FALSE}
rks[,1] <- gsub("Rank4",target["Rank4"],rks[,1])
rks
```
##Quantile normalisation
Go back to the rank matrix
```{r echo=FALSE}
rks
```
Substitue with values from the target distribution
```{r echo=FALSE}
target
```
```{r echo=FALSE}
rks[,1] <- gsub("Rank5",target["Rank5"],rks[,1])
rks
```
##Quantile normalisation
We then repeat to get the normalized matrix
```{r echo=FALSE}
for(i in 1:3){
rks[,i] <- gsub("Rank1",target["Rank1"],rks[,i])
rks[,i] <- gsub("Rank2",target["Rank2"],rks[,i])
rks[,i] <- gsub("Rank3",target["Rank3"],rks[,i])
rks[,i] <- gsub("Rank4",target["Rank4"],rks[,i])
rks[,i] <- gsub("Rank5",target["Rank5"],rks[,i])
}
rks <- as.data.frame(rks)
rownames(rks) <- rownames(df)
```
Original data
```{r echo=FALSE}
df
```
Normalised data
```{r echo=FALSE}
rks
```
##Final Code
```{r}
df <- data.frame(Array1 = c(1,3,9,2,4),
Array2 = c(3,4,2,1,9), Array3 = c(9,1,5,7,6))
rownames(df) <- LETTERS[1:nrow(df)]
rks <- apply(df, 2, function(x) paste("Rank",
rank(x,ties.method="min"),sep=""))
target <- round(rowMeans(apply(df, 2,sort)),3)
names(target) <- paste("Rank", 1:length(target),sep="")
for(i in 1:ncol(df)){
for(nm in names(target)){
rks[,i] <- gsub(nm,target[nm],rks[,i])
}
}
norm <- as.data.frame(rks)
```
##Effect of quantile normalisation
Caveats
- Distributions of samples are expected to be the same
- Majority of genes do not change between groups
```{r echo=FALSE,warning=FALSE,fig.height=4,fig.width=8,message=FALSE}
x <- rawdata
tmp <- unlist(indexProbes(x, which="both"))
tmp <- tmp[seq(1, length(tmp), len = 5000)]
df <- data.frame(log2(intensity(x)[tmp, ]))
med <- median(as.matrix(df))
par(mfrow=c(1,2))
boxplot(df,main="Before")
suppressPackageStartupMessages(library(limma))
boxplot(normalizeQuantiles(df),main="After")
```