Mercurial > repos > devteam > dwt_var_perfeature

comparison execute_dwt_var_perFeature.R @ 3:6c29c7e347e8 draft default tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
"planemo upload for repository https://github.com/galaxyproject/tools-devteam/tree/master/tools/dwt_var_perfeature commit f929353ffb0623f2218d7dec459c7da62f3b0d24"
author devteam
date Mon, 06 Jul 2020 20:34:27 -0400
parents
children
comparison
equal deleted inserted replaced
2:7a15159140d1 3:6c29c7e347e8
1 #####################################################################
2 ## plot multiscale wavelet variance
3 ## create null bands by permuting the original data series
4 ## generate plots and table of wavelet variance including p-values
5 #######################################################################
6 options(echo = FALSE)
7 library("wavethresh");
8 library("waveslim");
9 library("bitops");
10
11 ## to determine if data is properly formatted 2^N observations
12 is_power2 <- function(x) {
13 x && !(bitops::bitAnd(x, x - 1));
14 }
15
16 ## dwt : discrete wavelet transform using Haar wavelet filter, simplest wavelet function but later can modify to let user-define the wavelet filter function
17 dwt_var_permut_get_max <- function(data, names, alpha, filter = 1, family = "DaubExPhase", bc = "symmetric", method = "kendall", wf = "haar", boundary = "reflection") {
18 title <- NULL;
19 final_pvalue <- NULL;
20 j <- NULL;
21 scale <- NULL;
22 out <- NULL;
23
24 print(class(data));
25 print(names);
26 print(alpha);
27
28 par(mar = c(5, 4, 4, 3), oma = c(4, 4, 3, 2), xaxt = "s", cex = 1, las = 1);
29
30 title <- c("Wavelet", "Variance", "Pvalue", "Test");
31 print(title);
32
33 for (i in seq_len(length(names))) {
34 temp <- NULL;
35 results <- NULL;
36 wave1_dwt <- NULL;
37
38 ## if data fails formatting check, do something
39 print(is.numeric(as.matrix(data)[, i]));
40 if (!is.numeric(as.matrix(data)[, i])) {
41 stop("data must be a numeric vector");
42 }
43 print(length(as.matrix(data)[, i]));
44 print(is_power2(length(as.matrix(data)[, i])));
45 if (!is_power2(length(as.matrix(data)[, i]))) {
46 stop("data length must be a power of two");
47 }
48 j <- wavethresh::wd(as.matrix(data)[, i], filter.number = filter, family = family, bc = bc)$nlevels;
49 print(j);
50 temp <- vector(length = j);
51 wave1_dwt <- waveslim::dwt(as.matrix(data)[, i], wf = wf, j, boundary = boundary);
52
53 temp <- waveslim::wave.variance(wave1_dwt)[- (j + 1), 1];
54 print(temp);
55
56 ##permutations code :
57 feature1 <- NULL;
58 null <- NULL;
59 var_lower <- NULL;
60 limit_lower <- NULL;
61 var_upper <- NULL;
62 limit_upper <- NULL;
63 med <- NULL;
64
65 limit_lower <- alpha / 2 * 1000;
66 print(limit_lower);
67 limit_upper <- (1 - alpha / 2) * 1000;
68 print(limit_upper);
69
70 feature1 <- as.matrix(data)[, i];
71 for (k in 1:1000) {
72 nk_1 <- NULL;
73 null_levels <- NULL;
74 var <- NULL;
75 null_wave1 <- NULL;
76
77 nk_1 <- sample(feature1, length(feature1), replace = FALSE);
78 null_levels <- wavethresh::wd(nk_1, filter.number = filter, family = family, bc = bc)$nlevels;
79 var <- vector(length = length(null_levels));
80 null_wave1 <- waveslim::dwt(nk_1, wf = wf, j, boundary = boundary);
81 var <- waveslim::wave.variance(null_wave1)[- (null_levels + 1), 1];
82 null <- rbind(null, var);
83 }
84 null <- apply(null, 2, sort, na.last = TRUE);
85 var_lower <- null[limit_lower, ];
86 var_upper <- null[limit_upper, ];
87 med <- (apply(null, 2, median, na.rm = TRUE));
88
89 ## plot
90 results <- cbind(temp, var_lower, var_upper);
91 print(results);
92 matplot(results, type = "b", pch = "*", lty = 1, col = c(1, 2, 2), xaxt = "n", xlab = "Wavelet Scale", ylab = "Wavelet variance");
93 mtext(names[i], side = 3, line = 0.5, cex = 1);
94 axis(1, at = 1:j, labels = c(2 ^ (0:(j - 1))), las = 3, cex.axis = 1);
95
96 ## get pvalues by comparison to null distribution
97 for (m in seq_len(length(temp))) {
98 print(paste("scale", m, sep = " "));
99 print(paste("var", temp[m], sep = " "));
100 print(paste("med", med[m], sep = " "));
101 pv <- NULL;
102 tail <- NULL;
103 scale <- NULL;
104 scale <- 2 ^ (m - 1);
105 if (temp[m] >= med[m]) {
106 ## R tail test
107 print("R");
108 tail <- "R";
109 pv <- (length(which(null[, m] >= temp[m]))) / (length(na.exclude(null[, m])));
110 } else {
111 if (temp[m] < med[m]) {
112 ## L tail test
113 print("L");
114 tail <- "L";
115 pv <- (length(which(null[, m] <= temp[m]))) / (length(na.exclude(null[, m])));
116 }
117 }
118 print(pv);
119 out <- rbind(out, c(paste("Scale", scale, sep = "_"), format(temp[m], digits = 3), pv, tail));
120 }
121 final_pvalue <- rbind(final_pvalue, out);
122 }
123 colnames(final_pvalue) <- title;
124 return(final_pvalue);
125 }
126
127 ## execute
128 ## read in data
129 args <- commandArgs(trailingOnly = TRUE)
130
131 data_test <- NULL;
132 final <- NULL;
133 sub <- NULL;
134 sub_names <- NULL;
135 data_test <- read.delim(args[1], header = FALSE);
136 pdf(file = args[5], width = 11, height = 8)
137 for (f in strsplit(args[2], ",")) {
138 f <- as.integer(f)
139 if (f > ncol(data_test))
140 stop(paste("column", f, "doesn't exist"));
141 sub <- data_test[, f];
142 sub_names <- colnames(data_test)[f];
143 final <- rbind(final, dwt_var_permut_get_max(sub, sub_names, as.double(args[3])));
144 }
145
146 dev.off();
147 write.table(final, file = args[4], sep = "\t", quote = FALSE, row.names = FALSE);