[2] | 1 | #!/usr/bin/perl -w |
---|
| 2 | |
---|
| 3 | # a program to compute the frequencies of each motif at a window size, determined by the user, in both |
---|
| 4 | # upstream and downstream sequences flanking indels in all chromosomes. |
---|
| 5 | # the first input is a TABULAR format file containing the motif names and sequences, such that the file |
---|
| 6 | # consists of two columns: the left column represents the motif names and the right column represents |
---|
| 7 | # the motif sequence, one line per motif. |
---|
| 8 | # the second input is a TABULAR format file containing the windows into which both upstream and downstream |
---|
| 9 | # sequences flanking indels have been divided. |
---|
| 10 | # the fourth input is an integer number representing the number of windows to be considered in both |
---|
| 11 | # upstream and downstream flanking sequences. |
---|
| 12 | # the output is a TABULAR format file consisting of three columns: the left column represents the motif |
---|
| 13 | # name, the middle column represents the motif frequency in the window of the upstream sequence flanking |
---|
| 14 | # an indel, and the the right column represents the motif frequency in the window of the downstream |
---|
| 15 | # sequence flanking an indel, one line per indel. |
---|
| 16 | # The total number of lines in the output file = number of motifs x number of indels. |
---|
| 17 | |
---|
| 18 | use strict; |
---|
| 19 | use warnings; |
---|
| 20 | |
---|
| 21 | #variable to handle the window information |
---|
| 22 | my $window = ""; |
---|
| 23 | my $windowNumber = 0; |
---|
| 24 | my $totalWindowsNumber = 0; |
---|
| 25 | my $upstreamAndDownstreamFlankingSequencesWindows = ""; |
---|
| 26 | |
---|
| 27 | #variable to handle the motif information |
---|
| 28 | my $motif = ""; |
---|
| 29 | my $motifName = ""; |
---|
| 30 | my $motifSequence = ""; |
---|
| 31 | my $motifNumber = 0; |
---|
| 32 | my $totalMotifsNumber = 0; |
---|
| 33 | my $upstreamMotifFrequencyCounter = 0; |
---|
| 34 | my $downstreamMotifFrequencyCounter = 0; |
---|
| 35 | |
---|
| 36 | #arrays to sotre window and motif data |
---|
| 37 | my @windowsArray = (); |
---|
| 38 | my @motifNamesArray = (); |
---|
| 39 | my @motifSequencesArray = (); |
---|
| 40 | |
---|
| 41 | #variable to handle the indel information |
---|
| 42 | my $indelIndex = 0; |
---|
| 43 | |
---|
| 44 | #variable to store line counter value |
---|
| 45 | my $lineCounter = 0; |
---|
| 46 | |
---|
| 47 | # check to make sure having correct files |
---|
| 48 | my $usage = "usage: compute_motif_frequencies_for_all_motifs.pl [TABULAR.in] [TABULAR.in] [windowSize] [TABULAR.out] \n"; |
---|
| 49 | die $usage unless @ARGV == 4; |
---|
| 50 | |
---|
| 51 | #get the input arguments |
---|
| 52 | my $motifsInputFile = $ARGV[0]; |
---|
| 53 | my $indelFlankingSequencesWindowsInputFile = $ARGV[1]; |
---|
| 54 | my $numberOfConsideredWindows = $ARGV[2]; |
---|
| 55 | my $motifFrequenciesOutputFile = $ARGV[3]; |
---|
| 56 | |
---|
| 57 | #open the input files |
---|
| 58 | open (INPUT1, "<", $motifsInputFile) || die("Could not open file $motifsInputFile \n"); |
---|
| 59 | open (INPUT2, "<", $indelFlankingSequencesWindowsInputFile) || die("Could not open file indelFlankingSequencesWindowsInputFile \n"); |
---|
| 60 | open (OUTPUT, ">", $motifFrequenciesOutputFile) || die("Could not open file $motifFrequenciesOutputFile \n"); |
---|
| 61 | |
---|
| 62 | #store the motifs input file in the array @motifsData |
---|
| 63 | my @motifsData = <INPUT1>; |
---|
| 64 | |
---|
| 65 | #iterated through the motifs (lines) of the motifs input file |
---|
| 66 | foreach $motif (@motifsData){ |
---|
| 67 | chomp ($motif); |
---|
| 68 | #print ($motif . "\n"); |
---|
| 69 | |
---|
| 70 | #split the motif data into its name and its sequence |
---|
| 71 | my @motifNameAndSequenceArray = split(/\t/, $motif); |
---|
| 72 | |
---|
| 73 | #store the name of the motif into the array @motifNamesArray |
---|
| 74 | push @motifNamesArray, $motifNameAndSequenceArray[0]; |
---|
| 75 | |
---|
| 76 | #store the sequence of the motif into the array @motifSequencesArray |
---|
| 77 | push @motifSequencesArray, $motifNameAndSequenceArray[1]; |
---|
| 78 | } |
---|
| 79 | |
---|
| 80 | #compute the size of the motif names array |
---|
| 81 | $totalMotifsNumber = @motifNamesArray; |
---|
| 82 | |
---|
| 83 | |
---|
| 84 | #store the first output file containing the windows of both upstream and downstream flanking sequences in the array @windowsData |
---|
| 85 | my @windowsData = <INPUT2>; |
---|
| 86 | |
---|
| 87 | #check if the number of considered window entered by the user is 0 or negative, if so make it equal to 1 |
---|
| 88 | if ($numberOfConsideredWindows <= 0){ |
---|
| 89 | $numberOfConsideredWindows = 1; |
---|
| 90 | } |
---|
| 91 | |
---|
| 92 | #iterated through the motif sequences to check their occurrences in the considered windows |
---|
| 93 | #and store the count of their occurrences in the corresponding ouput file |
---|
| 94 | for ($motifNumber = 0; $motifNumber < $totalMotifsNumber; $motifNumber++){ |
---|
| 95 | |
---|
| 96 | #get the motif name |
---|
| 97 | $motifName = $motifNamesArray[$motifNumber]; |
---|
| 98 | |
---|
| 99 | #get the motif sequence |
---|
| 100 | $motifSequence = $motifSequencesArray[$motifNumber]; |
---|
| 101 | |
---|
| 102 | #iterated through the lines of the second input file. Each line represents |
---|
| 103 | #the windows of the upstream and downstream flanking sequences of an indel |
---|
| 104 | foreach $upstreamAndDownstreamFlankingSequencesWindows (@windowsData){ |
---|
| 105 | |
---|
| 106 | chomp ($upstreamAndDownstreamFlankingSequencesWindows); |
---|
| 107 | $lineCounter++; |
---|
| 108 | |
---|
| 109 | #split both upstream and downstream flanking sequences into their windows |
---|
| 110 | my @windowsArray = split(/\t/, $upstreamAndDownstreamFlankingSequencesWindows); |
---|
| 111 | |
---|
| 112 | if ($lineCounter == 1){ |
---|
| 113 | $totalWindowsNumber = @windowsArray; |
---|
| 114 | $indelIndex = ($totalWindowsNumber - 1)/2; |
---|
| 115 | } |
---|
| 116 | |
---|
| 117 | #reset the motif frequency counters |
---|
| 118 | $upstreamMotifFrequencyCounter = 0; |
---|
| 119 | $downstreamMotifFrequencyCounter = 0; |
---|
| 120 | |
---|
| 121 | #iterate through the considered windows of the upstream flanking sequence and increment the motif frequency counter |
---|
| 122 | for ($windowNumber = $indelIndex - 1; $windowNumber > $indelIndex - $numberOfConsideredWindows - 1; $windowNumber--){ |
---|
| 123 | |
---|
| 124 | #get the window |
---|
| 125 | $window = $windowsArray[$windowNumber]; |
---|
| 126 | |
---|
| 127 | #if the motif is found in the window, then increment its corresponding counter |
---|
| 128 | if ($window =~ m/$motifSequence/i){ |
---|
| 129 | $upstreamMotifFrequencyCounter++; |
---|
| 130 | } |
---|
| 131 | } |
---|
| 132 | |
---|
| 133 | #iterate through the considered windows of the upstream flanking sequence and increment the motif frequency counter |
---|
| 134 | for ($windowNumber = $indelIndex + 1; $windowNumber < $indelIndex + $numberOfConsideredWindows + 1; $windowNumber++){ |
---|
| 135 | |
---|
| 136 | #get the window |
---|
| 137 | $window = $windowsArray[$windowNumber]; |
---|
| 138 | |
---|
| 139 | #if the motif is found in the window, then increment its corresponding counter |
---|
| 140 | if ($window =~ m/$motifSequence/i){ |
---|
| 141 | $downstreamMotifFrequencyCounter++; |
---|
| 142 | } |
---|
| 143 | } |
---|
| 144 | |
---|
| 145 | #store the result into the output file of the motif |
---|
| 146 | print OUTPUT $motifName . "\t" . $upstreamMotifFrequencyCounter . "\t" . $downstreamMotifFrequencyCounter . "\n"; |
---|
| 147 | } |
---|
| 148 | } |
---|
| 149 | |
---|
| 150 | #close the input and output files |
---|
| 151 | close(OUTPUT); |
---|
| 152 | close(INPUT2); |
---|
| 153 | close(INPUT1); |
---|