PageRenderTime 37ms CodeModel.GetById 18ms app.highlight 13ms RepoModel.GetById 2ms app.codeStats 0ms

/tools/discreteWavelet/execute_dwt_IvC_all.pl

https://bitbucket.org/cistrome/cistrome-harvard/
Perl | 210 lines | 147 code | 34 blank | 29 comment | 19 complexity | d6f92d43152dc5ebe684296c4b3065cc MD5 | raw file
  1#!/usr/bin/perl -w
  2use warnings;
  3use IO::Handle;
  4
  5$usage = "execute_dwt_IvC_all.pl [TABULAR.in] [TABULAR.in] [TABULAR.out] [PDF.out]  \n";
  6die $usage unless @ARGV == 4;
  7
  8#get the input arguments
  9my $firstInputFile = $ARGV[0];
 10my $secondInputFile = $ARGV[1];
 11my $firstOutputFile = $ARGV[2];
 12my $secondOutputFile = $ARGV[3];
 13
 14open (INPUT1, "<", $firstInputFile) || die("Could not open file $firstInputFile \n");
 15open (INPUT2, "<", $secondInputFile) || die("Could not open file $secondInputFile \n");
 16open (OUTPUT1, ">", $firstOutputFile) || die("Could not open file $firstOutputFile \n");
 17open (OUTPUT2, ">", $secondOutputFile) || die("Could not open file $secondOutputFile \n");
 18open (ERROR,  ">", "error.txt")  or die ("Could not open file error.txt \n");
 19
 20#save all error messages into the error file $errorFile using the error file handle ERROR
 21STDERR -> fdopen( \*ERROR,  "w" ) or die ("Could not direct errors to the error file error.txt \n");
 22
 23
 24print "There are two input data files: \n";
 25print "The input data file is: $firstInputFile \n";
 26print "The control data file is: $secondInputFile \n";
 27
 28# IvC test
 29$test = "IvC";
 30
 31# construct an R script to implement the IvC test
 32print "\n";
 33
 34$r_script = "get_dwt_IvC_test.r"; 
 35print "$r_script \n";
 36
 37# R script
 38open(Rcmd, ">", "$r_script") or die "Cannot open $r_script \n\n";
 39print Rcmd "
 40        ###########################################################################################
 41        # code to do wavelet Indel vs. Control
 42        # signal is the difference I-C; function is second moment i.e. variance from zero not mean
 43        # to perform wavelet transf. of signal, scale-by-scale analysis of the function 
 44        # create null bands by permuting the original data series
 45        # generate plots and table matrix of correlation coefficients including p-values
 46        ############################################################################################
 47        library(\"Rwave\");
 48        library(\"wavethresh\");
 49        library(\"waveslim\");
 50        
 51        options(echo = FALSE)
 52        
 53        # normalize data
 54        norm <- function(data){
 55            v <- (data - mean(data))/sd(data);
 56            if(sum(is.na(v)) >= 1){
 57                v <- data;
 58            }
 59            return(v);
 60        }
 61        
 62        dwt_cor <- function(data.short, names.short, data.long, names.long, test, pdf, table, filter = 4, bc = \"symmetric\", wf = \"haar\", boundary = \"reflection\") {
 63            print(test);
 64            print(pdf);
 65            print(table);
 66            
 67            pdf(file = pdf);
 68            final_pvalue = NULL;
 69            title = NULL;
 70                
 71            short.levels <- wd(data.short[, 1], filter.number = filter, bc = bc)\$nlevels;
 72            title <- c(\"motif\");
 73            for (i in 1:short.levels){
 74            	title <- c(title, paste(i, \"moment2\", sep = \"_\"), paste(i, \"pval\", sep = \"_\"), paste(i, \"test\", sep = \"_\"));
 75            }
 76            print(title);
 77        
 78            # loop to compare a vs a
 79            for(i in 1:length(names.short)){
 80        		wave1.dwt = NULL;
 81        		m2.dwt = diff = var.dwt = NULL;
 82        		out = NULL;
 83                out <- vector(length = length(title));
 84        
 85        		print(names.short[i]);
 86        		print(names.long[i]);
 87                        
 88        		# need exit if not comparing motif(a) vs motif(a)
 89        		if (names.short[i] != names.long[i]){
 90                	stop(paste(\"motif\", names.short[i], \"is not the same as\", names.long[i], sep = \" \"));
 91        		}
 92        		else {
 93                	# signal is the difference I-C data sets
 94                    diff<-data.short[,i]-data.long[,i];
 95        
 96                    # normalize the signal
 97                    diff<-norm(diff);
 98        
 99                    # function is 2nd moment
100                    # 2nd moment m_j = 1/N[sum_N(W_j + V_J)^2] = 1/N sum_N(W_j)^2 + (X_bar)^2 
101            		wave1.dwt <- dwt(diff, wf = wf, short.levels, boundary = boundary);
102            		var.dwt <- wave.variance(wave1.dwt);
103                	m2.dwt <- vector(length = short.levels)
104                    for(level in 1:short.levels){
105                    	m2.dwt[level] <- var.dwt[level, 1] + (mean(diff)^2);
106                    }
107                                
108            		# CI bands by permutation of time series
109            		feature1 = feature2 = NULL;
110            		feature1 = data.short[, i];
111            		feature2 = data.long[, i];
112            		null = results = med = NULL; 
113            		m2_25 = m2_975 = NULL;
114            
115            		for (k in 1:1000) {
116                		nk_1 = nk_2 = NULL;
117                		m2_null = var_null = NULL;
118                		null.levels = null_wave1 = null_diff = NULL;
119                		nk_1 <- sample(feature1, length(feature1), replace = FALSE);
120                		nk_2 <- sample(feature2, length(feature2), replace = FALSE);
121                		null.levels <- wd(nk_1, filter.number = filter, bc = bc)\$nlevels;
122                		null_diff <- nk_1-nk_2;
123                		null_diff <- norm(null_diff);
124                		null_wave1 <- dwt(null_diff, wf = wf, short.levels, boundary = boundary);
125                        var_null <- wave.variance(null_wave1);
126                		m2_null <- vector(length = null.levels);
127                		for(level in 1:null.levels){
128                        	m2_null[level] <- var_null[level, 1] + (mean(null_diff)^2);
129                		}
130                		null= rbind(null, m2_null);
131            		}
132                
133            		null <- apply(null, 2, sort, na.last = TRUE);
134            		m2_25 <- null[25,];
135            		m2_975 <- null[975,];
136            		med <- apply(null, 2, median, na.rm = TRUE);
137
138            		# plot
139            		results <- cbind(m2.dwt, m2_25, m2_975);
140            		matplot(results, type = \"b\", pch = \"*\", lty = 1, col = c(1, 2, 2), xlab = \"Wavelet Scale\", ylab = c(\"Wavelet 2nd Moment\", test), main = (names.short[i]), cex.main = 0.75);
141            		abline(h = 1);
142
143            		# get pvalues by comparison to null distribution
144            		out <- c(names.short[i]);
145            		for (m in 1:length(m2.dwt)){
146                    	print(paste(\"scale\", m, sep = \" \"));
147                        print(paste(\"m2\", m2.dwt[m], sep = \" \"));
148                        print(paste(\"median\", med[m], sep = \" \"));
149                        out <- c(out, format(m2.dwt[m], digits = 4));	
150                        pv = NULL;
151                        if(is.na(m2.dwt[m])){
152                        	pv <- \"NA\"; 
153                        } 
154                        else {
155                        	if (m2.dwt[m] >= med[m]){
156                            	# R tail test
157                                tail <- \"R\";
158                                pv <- (length(which(null[, m] >= m2.dwt[m])))/(length(na.exclude(null[, m])));
159                            }
160                            else{
161                                if (m2.dwt[m] < med[m]){
162                                	# L tail test
163                                    tail <- \"L\";
164                                    pv <- (length(which(null[, m] <= m2.dwt[m])))/(length(na.exclude(null[, m])));
165                                }
166                            }
167                        }
168                        out <- c(out, pv);
169                        print(pv);  
170                        out <- c(out, tail);
171                    }
172                    final_pvalue <-rbind(final_pvalue, out);
173                    print(out);
174                }
175            }
176            
177            colnames(final_pvalue) <- title;
178            write.table(final_pvalue, file = table, sep = \"\\t\", quote = FALSE, row.names = FALSE);
179            dev.off();
180        }\n";
181
182print Rcmd "
183        # execute
184        # read in data 
185        
186        inputData <- read.delim(\"$firstInputFile\");
187        inputDataNames <- colnames(inputData);
188        
189        controlData <- read.delim(\"$secondInputFile\");
190        controlDataNames <- colnames(controlData);
191        
192        # call the test function to implement IvC test
193        dwt_cor(inputData, inputDataNames, controlData, controlDataNames, test = \"$test\", pdf = \"$secondOutputFile\", table = \"$firstOutputFile\");
194        print (\"done with the correlation test\");
195\n";
196
197print Rcmd "#eof\n";
198
199close Rcmd;
200
201system("echo \"wavelet IvC test started on \`hostname\` at \`date\`\"\n");
202system("R --no-restore --no-save --no-readline < $r_script > $r_script.out\n");
203system("echo \"wavelet IvC test ended on \`hostname\` at \`date\`\"\n");
204
205#close the input and output and error files
206close(ERROR);
207close(OUTPUT2);
208close(OUTPUT1);
209close(INPUT2);
210close(INPUT1);