|
0
|
1 ### This method generates a row and column ordering given an input matrix and ordering methods.
|
|
|
2 ###
|
|
|
3 ### matrixData - numeric matrix
|
|
|
4 ### rowOrderMethod - Hierarchical, Original, Random
|
|
|
5 ### rowDistanceMeasure - For clustering, distance measure. May be: euclidean, binary, manhattan, maximum, canberra, minkowski, or correlation.
|
|
|
6 ### rowAgglomerationMethod - For clustering, agglomeration method. May be: 'average' for Average Linkage, 'complete' for Complete Linkage,
|
|
|
7 ### 'single' for Single Linkage, 'ward', 'mcquitty', 'median', or 'centroid'.
|
|
|
8 ### colOrderMethod
|
|
|
9 ### colDistanceMeasure
|
|
|
10 ### colAgglomerationMethod
|
|
|
11 ### rowOrderFile - output file of order of rows
|
|
|
12 ### rowDendroFile - output file of row dendrogram
|
|
|
13 ### colOrderFile - output file of order of cols
|
|
|
14 ### colDendroFile - output file of col dendrogram
|
|
|
15
|
|
|
16 performDataOrdering<-function(dataFile, rowOrderMethod, rowDistanceMeasure, rowAgglomerationMethod, colOrderMethod, colDistanceMeasure, colAgglomerationMethod,rowOrderFile, colOrderFile, rowDendroFile, colDendroFile)
|
|
|
17 {
|
|
|
18 dataMatrix = read.table(dataFile, header=TRUE, sep = "\t", row.names = 1, as.is=TRUE)
|
|
|
19 rowOrder <- createOrdering(dataMatrix, rowOrderMethod, "row", rowDistanceMeasure, rowAgglomerationMethod)
|
|
|
20 if (rowOrderMethod == "Hierarchical") {
|
|
|
21 writeHCDataTSVs(rowOrder, rowDendroFile, rowOrderFile)
|
|
|
22 } else {
|
|
|
23 writeOrderTSV(rowOrder, rownames(dataMatrix), rowOrderFile)
|
|
|
24 }
|
|
|
25
|
|
|
26 colOrder <- createOrdering(dataMatrix, colOrderMethod, "col", colDistanceMeasure, colAgglomerationMethod)
|
|
|
27 if (colOrderMethod == "Hierarchical") {
|
|
|
28 writeHCDataTSVs(colOrder, colDendroFile, colOrderFile)
|
|
|
29 } else {
|
|
|
30 writeOrderTSV(colOrder, colnames(dataMatrix), colOrderFile)
|
|
|
31 }
|
|
|
32 }
|
|
|
33
|
|
|
34 #creates output files for hclust ordering
|
|
|
35 writeHCDataTSVs<-function(uDend, outputHCDataFileName, outputHCOrderFileName)
|
|
|
36 {
|
|
|
37 data<-cbind(uDend$merge, uDend$height, deparse.level=0)
|
|
|
38 colnames(data)<-c("A", "B", "Height")
|
|
|
39 write.table(data, file = outputHCDataFileName, append = FALSE, quote = FALSE, sep = "\t", row.names=FALSE)
|
|
|
40
|
|
|
41 data=matrix(,length(uDend$labels),2);
|
|
|
42 for (i in 1:length(uDend$labels)) {
|
|
|
43 data[i,1] = uDend$labels[i];
|
|
|
44 data[i,2] = which(uDend$order==i);
|
|
|
45 }
|
|
|
46 colnames(data)<-c("Id", "Order")
|
|
|
47 write.table(data, file = outputHCOrderFileName, append = FALSE, quote = FALSE, sep = "\t", row.names=FALSE)
|
|
|
48 }
|
|
|
49
|
|
|
50 #creates order file for non-clustering methods
|
|
|
51 writeOrderTSV<-function(newOrder, originalOrder, outputHCOrderFileName)
|
|
|
52 {
|
|
|
53 data=matrix(,length(originalOrder),2);
|
|
|
54 for (i in 1:length(originalOrder)) {
|
|
|
55 data[i,1] = originalOrder[i];
|
|
|
56 data[i,2] = which(newOrder==originalOrder[i]);
|
|
|
57 }
|
|
|
58 colnames(data)<-c("Id", "Order")
|
|
|
59 write.table(data, file = outputHCOrderFileName, append = FALSE, quote = FALSE, sep = "\t", row.names=FALSE)
|
|
|
60 }
|
|
|
61
|
|
|
62
|
|
|
63
|
|
|
64 createOrdering<-function(matrixData, orderMethod, direction, distanceMeasure, agglomerationMethod)
|
|
|
65 {
|
|
|
66 ordering <- NULL
|
|
|
67
|
|
|
68 if (orderMethod == "Hierarchical")
|
|
|
69 {
|
|
|
70
|
|
|
71 # Compute dendrogram for "Distance Metric"
|
|
|
72 distVals <- NULL
|
|
|
73 if(direction=="row") {
|
|
|
74 if (distanceMeasure == "correlation") {
|
|
|
75 geneGeneCor <- cor(t(matrixData), use="pairwise")
|
|
|
76 distVals <- as.dist((1-geneGeneCor)/2)
|
|
|
77 } else {
|
|
|
78 distVals <- dist(matrixData, method=distanceMeasure)
|
|
|
79 }
|
|
|
80 } else { #column
|
|
|
81 if (distanceMeasure == "correlation") {
|
|
|
82 geneGeneCor <- cor(matrixData, use="pairwise")
|
|
|
83 distVals <- as.dist((1-geneGeneCor)/2)
|
|
|
84 } else {
|
|
|
85 distVals <- dist(t(matrixData), method=distanceMeasure)
|
|
|
86 }
|
|
|
87 }
|
|
|
88
|
|
|
89 if (agglomerationMethod == "ward") {
|
|
|
90 ordering <- hclust(distVals * distVals, method="ward.D2")
|
|
|
91 } else {
|
|
|
92 ordering <- hclust(distVals, method=agglomerationMethod)
|
|
|
93 }
|
|
|
94 }
|
|
|
95 else if (orderMethod == "Random")
|
|
|
96 {
|
|
|
97 if(direction=="row") {
|
|
|
98 headerList <- rownames(matrixData)
|
|
|
99 ordering <- sample(headerList, length(headerList))
|
|
|
100 } else {
|
|
|
101 headerList <- colnames(matrixData)
|
|
|
102 ordering <- sample(headerList, length(headerList))
|
|
|
103 }
|
|
|
104 }
|
|
|
105 else if (orderMethod == "Original")
|
|
|
106 {
|
|
|
107 if(direction=="row") {
|
|
|
108 ordering <- rownames(matrixData)
|
|
|
109 } else {
|
|
|
110 ordering <- colnames(matrixData)
|
|
|
111 }
|
|
|
112 } else {
|
|
|
113 stop("createOrdering -- failed to find ordering method")
|
|
|
114 }
|
|
|
115 return(ordering)
|
|
|
116 }
|
|
|
117 ### Initialize command line arguments and call performDataOrdering
|
|
|
118
|
|
|
119 options(warn=-1)
|
|
|
120
|
|
|
121 args = commandArgs(TRUE)
|
|
|
122
|
|
|
123 performDataOrdering(dataFile=args[1], rowOrderMethod=args[2], rowDistanceMeasure=args[3], rowAgglomerationMethod=args[4], colOrderMethod=args[5], colDistanceMeasure=args[6], colAgglomerationMethod=args[7],rowOrderFile=args[8], colOrderFile=args[9], rowDendroFile=args[10], colDendroFile=args[11])
|
|
|
124
|
|
|
125 #suppressWarnings(performDataOrdering(dataFile=args[1], rowOrderMethod=args[2], rowDistanceMeasure=args[3], rowAgglomerationMethod=args[4], colOrderMethod=args[5], colDistanceMeasure=args[6], colAgglomerationMethod=args[7],rowOrderFile=args[8], colOrderFile=args[9], rowDendroFile=args[10], colDendroFile=args[11]))
|
