# $Id: SiteMatrix.pm 16123 2009-09-17 12:57:27Z cjfields $
#---------------------------------------------------------

=head1 NAME

Bio::Matrix::PSM::SiteMatrix - SiteMatrixI implementation, holds a
position scoring matrix (or position weight matrix) and log-odds

=head1 SYNOPSIS

  use Bio::Matrix::PSM::SiteMatrix;
  # Create from memory by supplying probability matrix hash
  # both as strings or arrays
  # where the frequencies  $a,$c,$g and $t are supplied either as
  # arrayref or string. Accordingly, lA, lC, lG and lT are the log
  # odds (only as arrays, no checks done right now)
  my ($a,$c,$g,$t,$score,$ic, $mid)=@_; 
  #or
  my ($a,$c,$g,$t,$score,$ic,$mid)=('05a011','110550','400001',
                                    '100104',0.001,19.2,'CRE1');
  #Where a stands for all (this frequency=1), see explanation bellow
  my %param=(-pA=>$a,-pC=>$c,-pG=>$g,-pT=>$t,
             -lA=>$la, -lC=>$lc,-lG=>$lg,-lT=>$l,
             -IC=>$ic,-e_val=>$score, -id=>$mid);
  my $site=Bio::Matrix::PSM::SiteMatrix->new(%param);
  #Or get it from a file:
  use Bio::Matrix::PSM::IO;
  my $psmIO= Bio::Matrix::PSM::IO->new(-file=>$file, -format=>'transfac');
  while (my $psm=$psmIO->next_psm) {
    #Now we have a Bio::Matrix::PSM::Psm object, 
    # see Bio::Matrix::PSM::PsmI for details
    #This is a Bio::Matrix::PSM::SiteMatrix object now
    my $matrix=$psm->matrix;  
  }

  # Get a simple consensus, where alphabet is {A,C,G,T,N}, 
  # choosing the character that both satisfies a supplied or default threshold
  # frequency and is the most frequenct character at each position, or N.
  # So for the position with A, C, G, T frequencies of 0.5, 0.25, 0.10, 0.15,
  # the simple consensus character will be 'A', whilst for 0.5, 0.5, 0, 0 it
  # would be 'N'.
  my $consensus=$site->consensus;

  # Get the IUPAC ambiguity code representation of the data in the matrix.
  # Because the frequencies may have been pseudo-count corrected, insignificant
  # frequences (below 0.05 by default) are ignored. So a position with
  # A, C, G, T frequencies of 0.5, 0.5, 0.01, 0.01 will get the IUPAC code 'M',
  # while 0.97, 0.01, 0.01, 0.01 will get the code 'A' and
  # 0.25, 0.25, 0.25, 0.25 would get 'N'.
  my $iupac=$site->IUPAC;

  # Getting/using regular expression (a representation of the IUPAC string)
  my $regexp=$site->regexp;
  my $count=grep($regexp,$seq);
  my $count=($seq=~ s/$regexp/$1/eg);
  print "Motif $mid is present $count times in this sequence\n";

=head1 DESCRIPTION

SiteMatrix is designed to provide some basic methods when working with position
scoring (weight) matrices, such as transcription factor binding sites for
example. A DNA PSM consists of four vectors with frequencies {A,C,G,T}. This is
the minimum information you should provide to construct a PSM object. The
vectors can be provided as strings with frequenciesx10 rounded to an int, going
from {0..a} and 'a' represents the maximum (10). This is like MEME's compressed
representation of a matrix and it is quite useful when working with relational
DB. If arrays are provided as an input (references to arrays actually) they can
be any number, real or integer (frequency or count).

When creating the object you can ask the constructor to make a simple pseudo
count correction by adding a number (typically 1) to all positions (with the
-correction option). After adding the number the frequencies will be
calculated. Only use correction when you supply counts, not frequencies.

Throws an exception if: You mix as an input array and string (for example A
matrix is given as array, C - as string). The position vector is (0,0,0,0). One
of the probability vectors is shorter than the rest.

Summary of the methods I use most frequently (details bellow):

  iupac - return IUPAC compliant consensus as a string
  score - Returns the score as a real number
  IC - information content. Returns a real number
  id - identifier. Returns a string
  accession - accession number. Returns a string
  next_pos - return the sequence probably for each letter, IUPAC
      symbol, IUPAC probability and simple sequence
  consenus letter for this position. Rewind at the end. Returns a hash.
  pos - current position get/set. Returns an integer.
  regexp - construct a regular expression based on IUPAC consensus.
      For example AGWV will be [Aa][Gg][AaTt][AaCcGg]
  width - site width
  get_string - gets the probability vector for a single base as a string.
  get_array - gets the probability vector for a single base as an array.
  get_logs_array - gets the log-odds vector for a single base as an array.

New methods, which might be of interest to anyone who wants to store
PSM in a relational database without creating an entry for each
position is the ability to compress the PSM vector into a string with
losing usually less than 1% of the data.  this can be done with:

  my $str=$matrix->get_compressed_freq('A');
or
  my $str=$matrix->get_compressed_logs('A');

Loading from a database should be done with new, but is not yest implemented.
However you can still uncompress such string with:

  my @arr=Bio::Matrix::PSM::_uncompress_string ($str,1,1); for PSM
or
  my @arr=Bio::Matrix::PSM::_uncompress_string ($str,1000,2); for log odds

=head1 FEEDBACK

=head2 Mailing Lists

User feedback is an integral part of the evolution of this and other
Bioperl modules. Send your comments and suggestions preferably to one
of the Bioperl mailing lists.  Your participation is much appreciated.

  bioperl-l@bioperl.org                  - General discussion
  http://bioperl.org/wiki/Mailing_lists  - About the mailing lists

=head2 Support 

Please direct usage questions or support issues to the mailing list:

I<bioperl-l@bioperl.org>

rather than to the module maintainer directly. Many experienced and 
reponsive experts will be able look at the problem and quickly 
address it. Please include a thorough description of the problem 
with code and data examples if at all possible.

=head2 Reporting Bugs

Report bugs to the Bioperl bug tracking system to help us keep track
the bugs and their resolution.  Bug reports can be submitted via the
web:

  http://bugzilla.open-bio.org/

=head1 AUTHOR - Stefan Kirov

Email skirov@utk.edu

=head1 CONTRIBUTORS

Sendu Bala, bix@sendu.me.uk

=head1 APPENDIX

The rest of the documentation details each of the object methods.
Internal methods are usually preceded with a _

=cut

# Let the code begin...
package Bio::Matrix::PSM::SiteMatrix;
use strict;

use base qw(Bio::Root::Root Bio::Matrix::PSM::SiteMatrixI);

=head2 new

 Title   : new
 Usage   : my $site=Bio::Matrix::PSM::SiteMatrix->new(-pA=>$a,-pC=>$c,
						     -pG=>$g,-pT=>$t,
						     -IC=>$ic,
						     -e_val=>$score, 
						     -id=>$mid);
 Function: Creates a new Bio::Matrix::PSM::SiteMatrix object from memory
 Throws :  If inconsistent data for all vectors (A,C,G and T) is
           provided, if you mix input types (string vs array) or if a
           position freq is 0.
 Returns :  Bio::Matrix::PSM::SiteMatrix object
 Args    :  -pA    => vector with the frequencies or counts of A
            -pC    => vector for C
            -pG    => vector for G
            -pt    => vector for T
            -lA    => vector for the log of A
            -lC    => vector for the log of C
            -lG    => vector for the log of G
            -lT    => vector for the log of T
            -IC    => real number, the information content of this matrix
            -e_val => real number, the expect value
            -id    => string, an identifier
            -width => int, width of the matrix in nucleotides
            -sites => int, the number of sites that went into this matrix
            -model => hash ref, background frequencies for A, C, G and T
            -correction => number, the number to add to all positions to achieve
                           psuedo count correction (default 0: no correction)
                           NB: do not use correction when your input is
                           frequences!
            -accession_number => string, an accession number

            Vectors can be strings of the frequencies where the frequencies are
            multiplied by 10 and rounded to the nearest whole number, and where
            'a' is used to denote the maximal frequency 10. There should be no
            punctuation (spaces etc.) in the string. For example, 'a0501'.
            Alternatively frequencies or counts can be represented by an array
            ref containing the counts, frequencies or logs as any kind of
            number.

=cut

sub new {
    my ($class, @args) = @_;
    my $self = $class->SUPER::new(@args);
    my $consensus;
    # Too many things to rearrange, and I am creating simultanuously >500 
    # such objects routinely, so this becomes performance issue
    my %input;
    while (@args) {
        (my $key = shift @args) =~ s/-//g; #deletes all dashes (only dashes)!
        $input{$key} = shift @args;
    }
    $self->{_position}   = 0;
    $self->{IC}     = $input{IC};
    $self->{e_val}  = $input{e_val};
    $self->{width}  = $input{width};
	$self->{logA}   = $input{lA};
	$self->{logC}   = $input{lC};
	$self->{logG}   = $input{lG};
	$self->{logT}   = $input{lT};
    $self->{sites}  = $input{sites};
    $self->{id}     = $input{id} || 'null';
    $self->{correction} = $input{correction} || 0;
    $self->{accession_number} = $input{accession_number};
	return $self unless (defined($input{pA}) && defined($input{pC}) && defined($input{pG}) && defined($input{pT}));
    
    # This should go to _initialize?
    # Check for input type- no mixing alllowed, throw ex
    if (ref($input{pA}) =~ /ARRAY/i ) {
        $self->throw("Mixing matrix data types not allowed: C is not reference") unless(ref($input{pC}));
        $self->throw("Mixing matrix data types not allowed: G is not reference") unless (ref($input{pG}));
        $self->throw("Mixing matrix data types not allowed: T is not reference") unless (ref($input{pT}));
        $self->{probA} = $input{pA};
        $self->{probC} = $input{pC};
        $self->{probG} = $input{pG};
        $self->{probT} = $input{pT};
    }
    else {
        $self->throw("Mixing matrix data types not allowed: C is reference") if (ref($input{pC}));
        $self->throw("Mixing matrix data types not allowed: G is reference") if (ref($input{pG}));
        $self->throw("Mixing matrix data types not allowed: T is reference") if (ref($input{pT}));
        $self->{probA} = [split(//,$input{pA})];
        $self->{probC} = [split(//,$input{pC})];
        $self->{probG} = [split(//,$input{pG})];
        $self->{probT} = [split(//,$input{pT})];
        for (my $i=0; $i<= @{$self->{probA}}+1; $i++) {
            # we implictely assume these are MEME-style frequencies x 10, so
            # 'a' represents the 'maximum', 10. Other positions can actually
            # add up to over 10 due to rounding, but I don't think that is a
            # problem?
            if (${$self->{probA}}[$i] and ${$self->{probA}}[$i] eq 'a') {
                ${$self->{probA}}[$i]='10';
            }
            if (${$self->{probC}}[$i] and ${$self->{probC}}[$i] eq 'a') {
                ${$self->{probC}}[$i]='10';
            }
            if (${$self->{probG}}[$i] and ${$self->{probG}}[$i] eq 'a') {
                ${$self->{probG}}[$i]='10';
            }
            if (${$self->{probT}}[$i] and ${$self->{probT}}[$i] eq 'a') {
                ${$self->{probT}}[$i]='10';
            }
        }
    }
    
    # Check for position with 0 for all bases, throw exception if so
    for (my $i=0;$i <= $#{$self->{probA}}; $i++) {
        if ((${$self->{probA}}[$i] + ${$self->{probC}}[$i] + ${$self->{probG}}[$i] + ${$self->{probT}}[$i]) == 0) {
            $self->throw("Position meaningless-all frequencies are 0");
        }
        
        # apply psuedo-count correction to all values - this will result in
        # very bad frequencies if the input is already frequences and a
        # correction value as large as 1 is used!
        if ($self->{correction}) {
            ${$self->{probA}}[$i] += $self->{correction};
            ${$self->{probC}}[$i] += $self->{correction};
            ${$self->{probG}}[$i] += $self->{correction};
            ${$self->{probT}}[$i] += $self->{correction};
        }
        
        # (re)calculate frequencies
        my $div= ${$self->{probA}}[$i] + ${$self->{probC}}[$i] + ${$self->{probG}}[$i] + ${$self->{probT}}[$i];
        ${$self->{probA}}[$i]=${$self->{probA}}[$i]/$div;
        ${$self->{probC}}[$i]=${$self->{probC}}[$i]/$div;
        ${$self->{probG}}[$i]=${$self->{probG}}[$i]/$div;
        ${$self->{probT}}[$i]=${$self->{probT}}[$i]/$div;
    }
    
    # Calculate the logs
    if ((!defined($self->{logA})) && ($input{model})) {
        $self->calc_weight($input{model});
    }
    
    # Make consensus, throw if any one of the vectors is shorter
    $self->_calculate_consensus;
    return $self;
}

=head2 _calculate_consensus

 Title   : _calculate_consensus
 Function: Internal stuff

=cut

sub _calculate_consensus {
    my $self=shift;
    my ($lc,$lt,$lg)=($#{$self->{probC}},$#{$self->{probT}},$#{$self->{probG}});
    my $len=$#{$self->{probA}};
    $self->throw("Probability matrix is damaged for C: $len vs $lc") if ($len != $lc);
    $self->throw("Probability matrix is damaged for T: $len vs $lt") if ($len != $lt);
    $self->throw("Probability matrix is damaged for G: $len vs $lg") if ($len != $lg);
    for (my $i=0; $i<$len+1; $i++) {
        #*** IUPACp values not actually used (eg. by next_pos)
        (${$self->{IUPAC}}[$i],${$self->{IUPACp}}[$i])=_to_IUPAC(${$self->{probA}}[$i], ${$self->{probC}}[$i], ${$self->{probG}}[$i], ${$self->{probT}}[$i]);
        (${$self->{seq}}[$i], ${$self->{seqp}}[$i]) = _to_cons(${$self->{probA}}[$i], ${$self->{probC}}[$i], ${$self->{probG}}[$i], ${$self->{probT}}[$i]);
    }
    return $self;
}

=head2 calc_weight

 Title   : calc_weight
 Usage   : $obj->calc_weight({A=>0.2562, C=>0.2438, G=>0.2432, T=>0.2568});
 Function: Recalculates the PSM (or weights) based on the PFM (the frequency
           matrix) and user supplied background model.
 Throws  : if no model is supplied
 Returns : n/a
 Args    : reference to a hash with background frequencies for A,C,G and T

=cut

sub calc_weight {
    my ($self, $model) = @_;
    my %model;
    $model{probA}=$model->{A};
    $model{probC}=$model->{C};
    $model{probG}=$model->{G};
    $model{probT}=$model->{T};
    foreach my $let qw(probA probC probG probT) {
      my @str;
      $self->throw('You did not provide valid model\n') unless (($model{$let}>0) && ($model{$let}<1));
      foreach my $f (@{$self->{$let}}) {
        my $w=log($f)-log($model{$let});
        push @str,$w;
      }
      my $llet=$let;
      $llet=~s/prob/log/;
      $self->{$llet}=\@str;
    }
    return $self;
}

=head2 next_pos

 Title   : next_pos
 Usage   :
 Function: Retrives the next position features: frequencies for A,C,G,T, the
           main letter (as in consensus) and the probabilty for this letter to
           occur at this position and the current position
 Returns : hash (pA,pC,pG,pT,logA,logC,logG,logT,base,prob,rel)
 Args    : none

=cut

sub next_pos {
    my $self = shift;
    $self->throw("instance method called on class") unless ref $self;
    my $len=@{$self->{seq}};
    my $pos=$self->{_position};
    # End reached?
    if ($pos<$len) {
	my $pA=${$self->{probA}}[$pos];
	my $pC=${$self->{probC}}[$pos];
	my $pG=${$self->{probG}}[$pos];
	my $pT=${$self->{probT}}[$pos];
	my $lA=${$self->{logA}}[$pos];
	my $lC=${$self->{logC}}[$pos];
	my $lG=${$self->{logG}}[$pos];
	my $lT=${$self->{logT}}[$pos];
	my $base=${$self->{seq}}[$pos];
	my $prob=${$self->{seqp}}[$pos];
	$self->{_position}++;
	my %seq=(pA=>$pA,pT=>$pT,pC=>$pC,pG=>$pG, lA=>$lA,lT=>$lT,lC=>$lC,lG=>$lG,base=>$base,rel=>$pos, prob=>$prob);
	return %seq;
    }
    else {$self->{_position}=0; return;}
}

=head2 curpos

 Title   : curpos
 Usage   :
 Function: Gets/sets the current position. Converts to 0 if argument is minus
           and to width if greater than width
 Returns : integer
 Args    : integer

=cut

sub curpos {
    my $self = shift;
    my $prev = $self->{_position};
    if (@_) { $self->{_position} = shift; }
    return $prev;
}

=head2 e_val

 Title   : e_val
 Usage   : 
 Function: Gets/sets the e-value
 Returns : real number
 Args    : none to get, real number to set

=cut

sub e_val {
    my $self = shift;
    my $prev = $self->{e_val};
    if (@_) { $self->{e_val} = shift; }
    return $prev;
}

=head2 IC

 Title   : IC
 Usage   :
 Function: Get/set the Information Content
 Returns : real number
 Args    : none to get, real number to set

=cut

sub IC {
    my $self = shift;
    my $prev = $self->{IC};
    if (@_) { $self->{IC} = shift; }
    return $prev;
}

=head2 accession_number

 Title   : accession_number
 Function: Get/set the accession number, this will be unique id for the
           SiteMatrix object as well for any other object, inheriting from
           SiteMatrix
 Returns : string
 Args    : none to get, string to set

=cut

sub accession_number {
    my $self = shift;
    my $prev = $self->{accession_number};
    if (@_) { $self->{accession_number} = shift; }
    return $prev;
}

=head2 consensus

 Title   : consensus
 Usage   :
 Function: Returns the consensus
 Returns : string
 Args    : (optional) threshold value 1 to 10, default 5
           '5' means the returned characters had a 50% or higher presence at
           their position

=cut

sub consensus {
    my ($self, $thresh) = @_;
    if ($thresh) {
        my $len=$#{$self->{probA}};
        for (my $i=0; $i<$len+1; $i++) {
            (${$self->{seq}}[$i], ${$self->{seqp}}[$i]) = _to_cons(${$self->{probA}}[$i], ${$self->{probC}}[$i], ${$self->{probG}}[$i], ${$self->{probT}}[$i], $thresh);
        }
    }
    my $consensus='';
    foreach my $letter (@{$self->{seq}}) {
        $consensus .= $letter;
    }
    return $consensus;
}

=head2 width

 Title   : width
 Usage   :
 Function: Returns the length of the sites in used to make this matrix
 Returns : int
 Args    : none

=cut

sub width {
    my $self = shift;
    my $width=@{$self->{probA}};
    return $width;
}

=head2 sites

 Title   : sites
 Usage   :
 Function: Get/set the number of sites that were used to make this matrix
 Returns : int
 Args    : none to get, int to set

=cut

sub sites {
    my $self = shift;
    if (@_) { $self->{sites} = shift }
    return $self->{sites} || return;
}

=head2 IUPAC

 Title   : IUPAC
 Usage   :
 Function: Returns IUPAC compliant consensus
 Returns : string
 Args    : optionally, also supply a whole number (int) of 1 or higher to set
           the significance level when considering the frequencies. 1 (the
           default) means a 0.05 significance level: frequencies lower than
           0.05 will be ignored. 2 Means a 0.005 level, and so on.

=cut

sub IUPAC {
	my ($self, $thresh) = @_;
    if ($thresh) {
        my $len=$#{$self->{probA}};
        for (my $i=0; $i<$len+1; $i++) {
            (${$self->{IUPAC}}[$i],${$self->{IUPACp}}[$i])=_to_IUPAC(${$self->{probA}}[$i], ${$self->{probC}}[$i], ${$self->{probG}}[$i], ${$self->{probT}}[$i], $thresh);
        }
    }
	my $iu=$self->{IUPAC};
	my $iupac='';
	foreach my $let (@{$iu}) {
		$iupac .= $let;
	}
    return $iupac;
}

=head2 _to_IUPAC

 Title   : _to_IUPAC
 Usage   :
 Function: Converts a single position to IUPAC compliant symbol.
           For rules see the implementation
 Returns : char, real number
 Args    : real numbers for frequencies of A,C,G,T (positional)

           optionally, also supply a whole number (int) of 1 or higher to set
           the significance level when considering the frequencies. 1 (the
           default) means a 0.05 significance level: frequencies lower than
           0.05 will be ignored. 2 Means a 0.005 level, and so on.

=cut

sub _to_IUPAC {
	my ($a, $c, $g, $t, $thresh) = @_;
    $thresh ||= 1;
    $thresh = int($thresh);
    $a = sprintf ("%.${thresh}f", $a);
    $c = sprintf ("%.${thresh}f", $c);
    $g = sprintf ("%.${thresh}f", $g);
    $t = sprintf ("%.${thresh}f", $t);
    
    my $total = $a + $c + $g + $t;
    
	return 'A' if ($a == $total);
	return 'G' if ($g == $total);
	return 'C' if ($c == $total);
	return 'T' if ($t == $total);
	my $r=$g+$a;
	return 'R' if ($r == $total);
	my $y=$t+$c;
	return 'Y' if ($y == $total);
	my $m=$a+$c;
	return 'M' if ($m == $total);
	my $k=$g+$t;
	return 'K' if ($k == $total);
	my $s=$g+$c;
	return 'S' if ($s == $total);
	my $w=$a+$t;
	return 'W' if ($w == $total);
	my $d=$r+$t;
	return 'D' if ($d == $total);
	my $v=$r+$c;
	return 'V' if ($v == $total);
	my $b=$y+$g;
	return 'B' if ($b == $total);
	my $h=$y+$a;
	return 'H' if ($h == $total);
	return 'N';
}

=head2 _to_cons

 Title   : _to_cons
 Usage   :
 Function: Converts a single position to simple consensus character and returns
           its probability. For rules see the implementation
 Returns : char, real number
 Args    : real numbers for A,C,G,T (positional), and optional 5th argument of
           threshold (as a number between 1 and 10, where 5 is default and
           means the returned character had a 50% or higher presence at this
           position)

=cut

sub _to_cons {
	my ($A, $C, $G, $T, $thresh) = @_;
    $thresh ||= 5;
    
    # this multiplication by 10 is just to satisfy the thresh range of 1-10
	my $a = $A * 10;
	my $c = $C * 10;
	my $g = $G * 10;
	my $t = $T * 10;
    
    return 'N',10 if (($a<$thresh) && ($c<$thresh) && ($g<$thresh) && ($t<$thresh));
	return 'N',10 if (($a==$t) && ($a==$c) && ($a==$g));
    
    # threshold could be lower than 50%, so must check is not only over
    # threshold, but also the highest frequency
	return 'A',$a if (($a>=$thresh) && ($a>$t) && ($a>$c) && ($a>$g));
	return 'C',$c if (($c>=$thresh) && ($c>$t) && ($c>$a) && ($c>$g));
	return 'G',$g if (($g>=$thresh) && ($g>$t) && ($g>$c) && ($g>$a));
	return 'T',$t if (($t>=$thresh) && ($t>$g) && ($t>$c) && ($t>$a));
	
    return 'N',10;
}

=head2 get_string

 Title   : get_string
 Usage   :
 Function: Returns given probability vector as a string. Useful if you want to
           store things in a rel database, where arrays are not first choice
 Throws  : If the argument is outside {A,C,G,T}
 Returns : string
 Args    : character {A,C,G,T}

=cut

sub get_string {
	my $self=shift;
	my $base=shift;
	my $string='';
	my @prob;
    
	BASE: {
		if ($base eq 'A') {@prob= @{$self->{probA}}; last BASE; }
		if ($base eq 'C') {@prob= @{$self->{probC}}; last BASE; }
		if ($base eq 'G') {@prob= @{$self->{probG}}; last BASE; }
		if ($base eq 'T') {@prob= @{$self->{probT}}; last BASE; }
		$self->throw ("No such base: $base!\n");
	}
    
    foreach  my $prob (@prob) {
        my $corrected = $prob*10;
        my $next=sprintf("%.0f",$corrected);
        $next='a' if ($next eq '10');
        $string .= $next;
    }
    return $string;
}

=head2 get_array

 Title   : get_array
 Usage   :
 Function: Returns an array with frequencies for a specified base
 Returns : array
 Args    : char

=cut

sub get_array {
	my $self=shift;
	my $base=uc(shift);
	return  @{$self->{probA}} if ($base eq 'A');
	return  @{$self->{probC}} if ($base eq 'C');
	return  @{$self->{probG}} if ($base eq 'G');
	return  @{$self->{probT}} if ($base eq 'T');
	$self->throw("No such base: $base!\n");
}

=head2 get_logs_array

 Title   : get_logs_array
 Usage   :
 Function: Returns an array with log_odds for a specified base
 Returns : array
 Args    : char

=cut

sub get_logs_array {
	my $self=shift;
	my $base=uc(shift);
	return  @{$self->{logA}} if (($base eq 'A')  && ($self->{logA}));
	return  @{$self->{logC}} if (($base eq 'C')  && ($self->{logC}));
	return  @{$self->{logG}} if (($base eq 'G')  && ($self->{logG}));
	return  @{$self->{logT}} if (($base eq 'T')  && ($self->{logT}));
	$self->throw ("No such base: $base!\n") if (!grep(/$base/,qw(A C G T)));
    return;
}

=head2 id

 Title   : id
 Usage   :
 Function: Gets/sets the site id
 Returns : string
 Args    : string

=cut

sub id {
    my $self = shift;
    my $prev = $self->{id};
    if (@_) { $self->{id} = shift; }
    return $prev;
}

=head2 regexp

 Title   : regexp
 Usage   :
 Function: Returns a regular expression which matches the IUPAC convention.
           N will match X, N, - and .
 Returns : string
 Args    : none (works at the threshold last used for making the IUPAC string)

=cut

sub regexp {
	my $self=shift;
	my $regexp;
	foreach my $letter (@{$self->{IUPAC}}) {
		my $reg;
		LETTER: {
			if ($letter eq 'A') { $reg='[Aa]'; last LETTER; }
			if ($letter eq 'C') { $reg='[Cc]'; last LETTER; }
			if ($letter eq 'G') { $reg='[Gg]'; last LETTER; }
			if ($letter eq 'T') { $reg='[Tt]'; last LETTER; }
			if ($letter eq 'M') { $reg='[AaCcMm]'; last LETTER; }
			if ($letter eq 'R') { $reg='[AaGgRr]'; last LETTER; }
			if ($letter eq 'W') { $reg='[AaTtWw]'; last LETTER; }
			if ($letter eq 'S') { $reg='[CcGgSs]'; last LETTER; }
			if ($letter eq 'Y') { $reg='[CcTtYy]'; last LETTER; }
			if ($letter eq 'K') { $reg='[GgTtKk]'; last LETTER; }
			if ($letter eq 'V') { $reg='[AaCcGgVv]'; last LETTER; }
			if ($letter eq 'H') { $reg='[AaCcTtHh]'; last LETTER; }
			if ($letter eq 'D') { $reg='[AaGgTtDd]'; last LETTER; }
			if ($letter eq 'B') { $reg='[CcGgTtBb]'; last LETTER; }
			$reg='\S';
		}
		$regexp .= $reg;
	}
    return $regexp;
}

=head2 regexp_array

 Title   : regexp_array
 Usage   :
 Function: Returns a regular expression which matches the IUPAC convention.
           N will match X, N, - and .
 Returns : array
 Args    : none (works at the threshold last used for making the IUPAC string)
 To do   : I have separated regexp and regexp_array, but
           maybe they can be rewritten as one - just check what should be returned

=cut

sub regexp_array {
	my $self=shift;
	my @regexp;
	foreach my $letter (@{$self->{IUPAC}}) {
		my $reg;
		LETTER: {
			if ($letter eq 'A') { $reg='[Aa]'; last LETTER; }
			if ($letter eq 'C') { $reg='[Cc]'; last LETTER; }
			if ($letter eq 'G') { $reg='[Gg]'; last LETTER; }
			if ($letter eq 'T') { $reg='[Tt]'; last LETTER; }
			if ($letter eq 'M') { $reg='[AaCcMm]'; last LETTER; }
			if ($letter eq 'R') { $reg='[AaGgRr]'; last LETTER; }
			if ($letter eq 'W') { $reg='[AaTtWw]'; last LETTER; }
			if ($letter eq 'S') { $reg='[CcGgSs]'; last LETTER; }
			if ($letter eq 'Y') { $reg='[CcTtYy]'; last LETTER; }
			if ($letter eq 'K') { $reg='[GgTtKk]'; last LETTER; }
			if ($letter eq 'V') { $reg='[AaCcGgVv]'; last LETTER; }
			if ($letter eq 'H') { $reg='[AaCcTtHh]'; last LETTER; }
			if ($letter eq 'D') { $reg='[AaGgTtDd]'; last LETTER; }
			if ($letter eq 'B') { $reg='[CcGgTtBb]'; last LETTER; }
			$reg='\S';
		}
		push @regexp,$reg;
	}
    return @regexp;
}


=head2 _compress_array

 Title   : _compress_array
 Usage   :
 Function: Will compress an array of real signed numbers to a string (ie vector
           of bytes) -127 to +127 for bi-directional(signed) and 0..255 for
           unsigned
 Returns : String
 Args    : array reference, followed by an max value and direction (optional,
           default 1-unsigned),1 unsigned, any other is signed. 

=cut

sub _compress_array {
	my ($array,$lm,$direct)=@_;
	my $str;
	return  unless(($array) && ($lm));
	$direct=1 unless ($direct);
	my $k1= ($direct==1) ? (255/$lm) : (127/$lm);
	foreach my $c (@{$array}) {
		$c=$lm if ($c>$lm);
		$c=-$lm if (($c<-$lm) && ($direct !=1));
    $c=0 if (($c<0) && ($direct ==1));
		my $byte=int($k1*$c);
    $byte=127+$byte if ($direct !=1);#Clumsy, should be really shift the bits
    my $char=chr($byte);
		$str.=$char;
	}
	return $str;
}

=head2 _uncompress_string

 Title   : _uncompress_string
 Usage   :
 Function: Will uncompress a string (vector of bytes) to create an array of
           real signed numbers (opposite to_compress_array)
 Returns : string, followed by an max value and
 		   direction (optional, default 1-unsigned), 1 unsigned, any other is signed.
 Args    : array

=cut

sub _uncompress_string {
	my ($str,$lm,$direct)=@_;
	my @array;
	return unless(($str) && ($lm));
	$direct=1 unless ($direct);
	my $k1= ($direct==1) ? (255/$lm) : (127/$lm);
	foreach my $c (split(//,$str)) {
		my $byte=ord($c);
		$byte=$byte-127 if ($direct !=1);#Clumsy, should be really shift the bits
		my $num=$byte/$k1;
		push @array,$num;
	}
	return @array;
}

=head2 get_compressed_freq

 Title   : get_compressed_freq
 Usage   :
 Function: A method to provide a compressed frequency vector. It uses one byte
           to code the frequence for one of the probability vectors for one
           position. Useful for relational database. Improvment of the previous
           0..a coding.
 Example :  my $strA=$self->get_compressed_freq('A');
 Returns :  String
 Args    :  char 

=cut

sub get_compressed_freq {
	my $self=shift;
	my $base=shift;
	my $string='';
	my @prob;
	BASE: {
		if ($base eq 'A') {
      @prob= @{$self->{probA}} unless (!defined($self->{probA}));
      last BASE;
    }
  		if ($base eq 'G') {
      @prob= @{$self->{probG}} unless (!defined($self->{probG}));
      last BASE;
    }
  		if ($base eq 'C') {
      @prob= @{$self->{probC}} unless (!defined($self->{probC}));
      last BASE;
    }
  		if ($base eq 'T') {
      @prob= @{$self->{probT}} unless (!defined($self->{probT}));
      last BASE;
    }
		$self->throw ("No such base: $base!\n");
	}
	my $str= _compress_array(\@prob,1,1);
    return $str;
}

=head2 get_compressed_logs

 Title   : get_compressed_logs
 Usage   :
 Function: A method to provide a compressed log-odd vector. It uses one byte to
 		   code the log value for one of the log-odds vectors for one position.
 Example : my $strA=$self->get_compressed_logs('A');
 Returns : String
 Args    : char 

=cut

sub get_compressed_logs {
	my $self=shift;
	my $base=shift;
	my $string='';
	my @prob;
	BASE: {
		if ($base eq 'A') {@prob= @{$self->{logA}} unless (!defined($self->{logA})); last BASE; }
		if ($base eq 'C') {@prob= @{$self->{logC}} unless (!defined($self->{logC})); last BASE; }
		if ($base eq 'G') {@prob= @{$self->{logG}} unless (!defined($self->{logG})); last BASE; }
		if ($base eq 'T') {@prob= @{$self->{logT}} unless (!defined($self->{logT})); last BASE; }
		$self->throw ("No such base: $base!\n");
	}
	return _compress_array(\@prob,1000,2);
}

=head2 sequence_match_weight

 Title   : sequence_match_weight
 Usage   :
 Function: This method will calculate the score of a match, based on the PWM
           if such is associated with the matrix object. Returns undef if no
           PWM data is available.
 Throws  : if the length of the sequence is different from the matrix width
 Example : my $score=$matrix->sequence_match_weight('ACGGATAG');
 Returns : Floating point
 Args    : string

=cut

sub sequence_match_weight {
    my ($self,$seq)=@_;
    return unless ($self->{logA});
    my $width=$self->width;
    $self->throw ("I can calculate the score only for sequence which are exactly my size for $seq, my width is $width\n") unless (length($seq)==@{$self->{logA}});
    $seq = uc($seq);
    my @seq=split(//,$seq);
    my $score = 0;
    my $i=0;
    foreach my $pos (@seq) {
        my $tv = 'log'.$pos;
        $self->warn("Position ".($i+1)." of input sequence has unknown (ambiguity?) character '$pos': scores will be wrong") unless defined $self->{$tv};
        $score += defined $self->{$tv} ? $self->{$tv}->[$i] : 0;
        $i++;
    }
    return $score;
}

=head2 get_all_vectors

 Title   : get_all_vectors
 Usage   :
 Function: returns all possible sequence vectors to satisfy the PFM under
           a given threshold
 Throws  : If threshold outside of 0..1 (no sense to do that)
 Example : my @vectors=$self->get_all_vectors(4);
 Returns : Array of strings
 Args    : (optional) floating

=cut

sub get_all_vectors {
	my $self=shift;
	my $thresh=shift;
    $self->throw("Out of range. Threshold should be >0 and 1<.\n") if (($thresh<0) || ($thresh>1));
    my @seq=split(//,$self->consensus($thresh*10));
    my @perm;
    for my $i (0..@{$self->{probA}}) {
        push @{$perm[$i]},'A' if ($self->{probA}->[$i]>$thresh);
        push @{$perm[$i]},'C' if ($self->{probC}->[$i]>$thresh);
        push @{$perm[$i]},'G' if ($self->{probG}->[$i]>$thresh);
        push @{$perm[$i]},'T' if ($self->{probT}->[$i]>$thresh);
        push @{$perm[$i]},'N' if  ($seq[$i] eq 'N');
    }
    my $fpos=shift @perm;
    my @strings=@$fpos;
    foreach my $pos (@perm) {
        my @newstr;
        foreach my $let (@$pos) {
            foreach my $string (@strings) {
                my $newstring = $string . $let;
                push @newstr,$newstring;
            }
        }
        @strings=@newstr;
    }
	return @strings;
}

1;