Scaning and matching letter sequences using matrix - algorithm

Question

I have two txt files, then I put them into hash'es, sequence => title.
In file DUOMENYS.txt title is known, in file "DUOTA.txt" title is unknown.
So for every sequence in file "DUOTA.txt" I need to find similar sequence in file DUOMENYS.txt and then print that known title.
I tried to make this with slimple matching, printing title with more than 90% sequence symbol matches, but I was told that it is wrong, and I have to do it other way, with this table: http://www.ncbi.nlm.nih.gov/Class/FieldGuide/BLOSUM62.txt
I have to compare letters from known sequence and unknown sequence and get number
(-4)-9, if sum of all that numbers => length of sequence * 3, print that title
Example, ABW => TILE1, DUOMENYS.txt, ABF => UNKNOWNTITLE, DUOTA.txt,
A B W
A B F
4 4 1 sum = 9
length 3 x 3 = 9
9 => 9, true, print.
So the problem is I don't know how to make it happen....
#!/usr/bin/perl
use strict;
use Data::Dumper;
open (OUTPUT, ">OUTPUT.txt") or die "$!"; #Turimos vairenio sekos
open (DUOMENYS, "DUOMENYS.txt") or die "$!";
open (OUTPUT1, ">OUTPUT1.txt") or die "$!"; #Tiriamos sekos
open (DUOTA, "DUOTA.txt") or die "$!";
open (OUTPUT2, ">OUTPUT2.txt") or die "$!"; #rezultatai
open (MATRIX, "MATRIX.txt") or die "$!";
#--------------------DUOMENYS-HASH-----------------------------
#my $contentFile = $ARGV[0];
my $contentFile = <DUOMENYS>;
my %testHash = ();
my $currentKey = "";
my $seka = "";
my %nhash = ();
open(my $contentFH,"<",$contentFile);
while(my $contentLine = <DUOMENYS>){
chomp($contentLine);
next if($contentLine eq ""); # Empty lines.
if($contentLine =~ /^\>(.*)/){
$testHash{$currentKey} = $seka;
$currentKey= $1;
$seka = "";
}else{
$seka .= $contentLine;
}
}
#-------------------DUOTA-HASH-------------------------------------
#my $contentFile1 = $ARGV[0];
my $contentFile1 = <DUOTA>;
my %testHash1 = ();
my $currentKey1 = "";
my $seka1 = "";
my %nhash1 = ();
open(my $contentFH1,"<",$contentFile1);
while(my $contentLine1 = <DUOTA>){
chomp($contentLine1);
next if($contentLine1 eq ""); # Empty lines.
if($contentLine1 =~ /^\>(.*)/){
$testHash1{$currentKey1} = $seka1;
$currentKey1= $1;
$seka1 = "";
}else{
$seka1 .= $contentLine1;
}
}
#--------------------OUTPUT-HASH------------------------------------
%nhash = reverse %testHash;
print OUTPUT Dumper(\%nhash);
%nhash1 = reverse %testHash1;
print OUTPUT1 Dumper(\%nhash1);
#---------------------MATCHING---------------------------------------
my $klaidu_skaicius = 0;
my #sekos = keys %nhash;
my #duotos_sekos = keys %nhash1;
my $i = 0;
my $j = 0;
for($i = 0; $i <= scalar#sekos; $i++){
for($j = 0; $j <= scalar#duotos_sekos; $j++){
$klaidu_skaicius = (#sekos[$i] ^ #duotos_sekos[$j]) =~ tr/\0//c;
if($klaidu_skaicius <= length(#sekos[$i])/10){
print OUTPUT2 substr( $nhash{#sekos[$i]}, 0, 9 ), "\n";
}
else{
print OUTPUT2 "";
}
}
}
From comments
pastebin.com/7QnBDTDY – povilito May 30 at 11:57
Its too big (15mb) for pastebin.com – povilito May 30 at 12:01
filedropper.com/duomenys – povilito May 30 at 12:04
I think comparing "letter" with " "(space) or "" should give us number -4 – povilito May 30 at 12:28
It's enougth to find one title for one unknown sequence. – povilito May 30 at 12:45
So if there is 50 unknown sequences, output file should give us 50 titles, some tittles can be the same :)

Answer 1

Here is the basic version of solution to your problem, as per I understood. I considered both the sequence equal, you can make the changes easily if you need.
#!/usr/bin/perl
use strict;
use warnings;
# convert your matrix file to 2D hash
my %hash;
open my $fh, '<', 'matrix' or die "unable to open file: $!\n";
chomp(my $heading = <$fh>);
# strip out space from begining
$heading =~ s/^\s+//g;
my #headings = split (/\s+/, $heading);
while(<$fh>) {
chomp;
my #line = split;
my $key1 = shift #line;
foreach my $i( 0 .. $#line) {
$hash{$key1}{$headings[$i]} = $line[$i];
}
}
close $fh;
# Took from your example for reference
my $duameny = "ABW";
my $duota = "ABF";
my #duamenys = split (//,$duameny);
my #duotas = split (//,$duota);
# calculate sum from hash
# considering both sequences are equal
my $sum = 0;
foreach my $i (0 .. $#duamenys) {
$sum += $hash{$duamenys[$i]}{$duotas[$i]};
}
# calculate length from sequence
my $length = (length $duameny) * 3;
print "SUM: $sum, Length: $length\n";
if($sum >= $length) {
# probably you know how to print the title
# print the title from duamenys.txt
}
Below is summary of my approach.
First converted the matrix into 2D hash for lookup.
Calculated sum from the hash for sequences.
Calculated length from the sequences
Compare them and print the title if sum >= length .
output:
SUM: 9, Length: 9