diff compStrains.pl @ 8:5bd43e313819 draft

Uploaded
author antmarge
date Wed, 29 Mar 2017 14:02:02 -0400
parents
children
line wrap: on
line diff
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/compStrains.pl	Wed Mar 29 14:02:02 2017 -0400
@@ -0,0 +1,251 @@
+#!/usr/bin/perl -w
+
+#Margaret Antonio 16.01.13
+
+#DESCRIPTION: Takes two aggregate.pl outputs and compares them using mean difference, pval for each
+#gene. Can compare, for example, 19F in glucose and TIGR4 in glucose.
+#DIFFERENT GENOMES (ie. diff. strains).
+#Requires CONVERSION FILE
+
+#USAGE: perl compStrains.pl -c <conversion.csv> <options>
+    #[<aggregateFile1.csv aggregateFile2.csv> OR -indir <indir/>]
+
+use Data::Dumper;
+use strict;
+use Getopt::Long;
+use warnings;
+use File::Path;
+use File::Basename;
+use Statistics::Distributions;
+
+#ASSIGN INPUTS TO VARIABLES USING FLAGS
+our ($indir,$h,$out,$sortkey,$round,$l,$cfile);
+GetOptions(
+'d:s' => \$indir,
+'h' => \$h,
+'o:s' =>\$out,
+'s:i' => \$sortkey,
+'r:i'=> \$round,
+'l1:s'=> \$l1,
+'l2:s'=> \$l2,
+'c:s'=> \$cfile,
+);
+
+sub print_usage() {
+    print "\n";
+    print "\n##################################################################\n";
+    print "compStrains.pl: compare genes from a tn-seq experiment\n";
+    print "\tfor two DIFFERENT strains/genomes using aggregate files\n";
+    
+    print "\nDESCRIPTION: Takes two aggregate.pl outputs and compares them by\n";
+    print "calculating the difference in mean fitness.\n";
+    
+    print "Example: two strains tested under same condition.\n";
+    print "Note: For same strains (genomes), use compGenes.pl\n";
+    
+    print "\nUSAGE:\n";
+    print "perl compStrains.pl -c conversion.csv -d inputs/\n";
+    
+    print "\nREQUIRED:\n";
+    print " -d\tDirectory containing all input files (files from\n";
+    print "   \taggregate fitness script)\n";
+    print "   \tOR\n";
+    print "   \tIn the command line (without a flag), input the name(s) of\n";
+    print "   \ttwo files containing aggregate gene fitness values. \n\n";
+    print " -c\tConversion file: two columns with homologs for both organisms\n";
+
+    print "\nOPTIONAL:\n";
+    print " -h\tPrints usage and exits program\n";
+    print " -o\tOutput file for comparison data. Default: label1label2.csv\n";
+    print " -s\tSort output by this index of the file (indices begin at 0).\n";
+    print "   \tDefault: by mean\n";
+    print " -r\tRound final output numbers to this number of decimals\n";
+    print " -l\tLabels for input files. Default: filenames\n";
+    print "   \tTwo strings, comma separated (i.e. -l expt1,expt2).\n";
+    print "   \tOrder should match file order.\n";
+	print " \n~~~~Always check that file paths are correctly specified~~~~\n";
+    print "\n##################################################################\n";
+}
+if ($h){
+    print_usage();
+    exit;
+}
+if (!$indir and (scalar @ARGV==0)){
+	print "\nERROR: Please correctly specify input files or directory\n";
+    print_usage();
+	print "\n";
+	exit;
+}
+if (!$cfile){
+	print "\nERROR: Please correctly specify the required conversion file\n";
+    print_usage();
+	print "\n";
+	exit;
+}
+
+#THE @files ARRAY WILL CONTAIN INPUT FILE NAMES, EXTRACTED FROM A DIRECTORY (-indir) OR ARGV
+my @files;
+if ($indir){
+    my $directory="$indir";
+    opendir(DIR, $directory) or (print "Couldn't open $directory: $!\n" and print_usage() and exit);
+    my @direct= readdir DIR;
+    my $tail=".csv";
+    foreach (@direct){
+        if (index($_, $tail) != -1){
+            $_=$indir.$_;
+            push (@files,$_);
+        }
+    }
+    closedir DIR;
+}
+else{
+    @files=@ARGV;
+}
+
+#GET LABELS: USE (-l) OR USE FILNEAMES AS LABELS FOR COLUMNS IN OUTPUT FILE
+
+
+my @labels = ($l1,$l2);
+#if ($l){
+#    @labels=split(',',$l);
+#}
+#else{
+#    foreach (@files){
+#        my $filename=basename($_);
+#        my @temp=split('\\.',$filename);
+#        my $colName=$temp[0];
+#        push (@labels,$colName);
+  #  }
+#}
+
+#CHECK IF REQ. VARIABLES WERE DEFINED USING FLAGS. IF NOT THEN USE DEFAULT VALUES
+
+if (!$out) {$out="comp.".$labels[0].$labels[1].".csv"}
+if (!$round){$round='%.4f'}
+
+#OPEN INPUTTED AGGREGATE GENE FILES AND STORE THEIR CONTENTS INTO TWO HASHES
+#FILE1 GOES INTO HASH %ONE AND FILE2 GOES INTO HASH %TWO.
+
+#FILE1 OPENING ---> %one WHERE KEY:VALUE IS GENE_ID:(GENE_ID,INSERTIONS,MEAN,ETC.)
+my @header;
+my %one;
+
+open (F1,'<',$files[0]);
+
+#STORE COLUMN NAMES (FIRST LINE OF FILE1) FOR HEADER AND APPEND LABELS
+my $head=<F1>; #the header in the file
+my @cols=split(',',$head);
+@cols=@cols[0,1,2,3,4,5,6]; #get rid of blank columns
+for (my $j=0;$j<scalar @cols;$j++){
+    $cols[$j]=$cols[$j].'-'.$labels[0];   #mark each column name with file it comes from
+}
+push (@header,@cols);
+
+while (my $line=<F1>){
+    chomp $line;
+    my @info=split(",",$line);
+    #Only keep the first 7 columns (Ones about blanks aren't needed for comparisons)
+    @info=@info[0,1,2,3,4,5,6];
+    #Sometimes genes that don't have a gene name can't be blank, so fill with NA
+    if (!$info[5]){
+        $info[5]="NA";
+    }
+    #If there are no insertions in the column "total", then make it =0 rather than blank
+    if (!$info[6]){
+        $info[6]=0;
+    }
+    $one{$info[0]}=\@info;
+}
+close F1;
+
+#FILE2 OPENING ---> %two WHERE KEY:VALUE IS GENE_ID:(GENE_ID,INSERTIONS,MEAN,ETC.)
+
+my %two;
+open (F2,'<',$files[1]);
+
+#STORE COLUMN NAMES (FIRST LINE OF FILE2) FOR HEADER AND APPEND LABELS
+$head=<F2>; #the header in the file
+@cols=split(',',$head);
+@cols=@cols[0,1,2,3,4,5,6]; #get rid of blank columns
+for (my $j=0;$j<scalar @cols;$j++){
+    $cols[$j]=$cols[$j].'-'.$labels[1];   #mark each column name with file it comes from
+}
+push (@header,@cols);
+
+while (my $line=<F2>){
+    chomp $line;
+    my @info=split(",",$line);
+    @info=@info[0,1,2,3,4,5,6];
+    if (!$info[5]){
+        $info[5]="NA";
+    }
+    if (!$info[6]){
+        $info[6]=0;
+    }
+    $two{$info[0]}=\@info;
+}
+close F2;
+
+
+#READ CONVERSION FILE INTO ARRAY.
+#Conversion file must have strain 1 for file 1 in column 1 (index 0) and
+    #strain 2 for file 2 in column 2 (index 1)
+    #conversion file must be tab delimited with no NA fields
+#If homologs (exist then take info from hashes (%one and %two) by referring to gene_id in KEY
+
+my @all; #store all homologs in this hash
+open (CONV,'<',$cfile);
+while (my $line=<CONV>){
+    chomp $line;
+    my @genes=split("\t",$line);   #Array @genes will contain two genes (SP_0000,SPT_0000)
+    if (scalar @genes==2 and $genes[0] ne "" and $genes[1] ne ""){
+        my @info;
+        my @oneArray=@{$one{$genes[0]}};
+        my @twoArray=@{$two{$genes[1]}};
+        push (@info,@oneArray,@twoArray);
+        my $diff=sprintf("$round",($info[1]-$info[8]));
+        my $total1=$info[6];
+        my $total2=$info[13];
+        my $sd1=$info[3];
+        my $se1=$info[4];
+        my $sd2=$info[10];
+        my $se2=$info[11];
+        my $df=$total1+$total2-2;
+        my $tdist;
+        my $pval;
+        #TDIST, PVAL calculations with fail if standard dev, error, or counts are not real numbers
+        #or if 0 ends up in denominator
+        if ($se1 eq "X" or $se2 eq "X" or $sd1 eq "X" or $sd2 eq "X" or $total1==0 or $total2==0 or $sd1==0 or $sd2==0){
+            ($tdist,$pval)=("NA","NA");
+        }
+        else{
+            $tdist=sqrt((($diff)/(sqrt((($sd1**2)/$total1)+(($sd2**2)/$total2))))**2);
+            $pval=Statistics::Distributions::tprob($df,$tdist);
+        }
+        push (@info,$diff,$df,$tdist,$pval);
+        push (@all,\@info);
+    }
+}
+close CONV;
+
+#SORT THE HOMOLOGS BY THE SORTKEY OR BY DEFAULT DIFFERENCE IN MEAN FITNESSES
+if (!$sortkey){
+    $sortkey=14; #for mean difference
+}
+my @sorted = sort { $b->[$sortkey] <=> $a->[$sortkey] } @all;
+
+#FINISH THE HEADER BY ADDING COLUMN NAMES FOR MEAN-DIFF, DOF, TDIST, AND PVALUE
+my $field="MeanDiff(".$labels[0].'.'.$labels[1].")";
+push (@header,$field,"DOF","TDIST","PVALUE");
+
+#PRINT MATCHED HOMOLOG INFORMATION INTO A SINGLE OUTPUT FILE
+open OUT, '>',"$out";
+print OUT (join(',',@header),"\n");
+foreach (@sorted){
+    my @woo=@{$_};
+    print OUT join(',',@woo),"\n";
+    }
+
+close OUT;
+
+