Pattern match with 2 conditions

[Andrew Perrin]

Stephan Bour <sbour@niaid.nih.gov> writes:

> #! usr/bin/perl -w

use strict; # is your friend

>
> $/="\n>";
>
> while(<>){
>
> $seq=$_;

what's the point of this when you just set it back to "" below?

>
> @fields=split(/\n/,$_);
> $i=-1;
> $seq="";
>
> foreach $line (@fields) {

why assign $line when you never use it?

>
> $i++;
> if($i==0) {
> $def=$fields[$i];
> }
> else {
> $seq.=$fields[$i];
> }
> }

This is not in answer to your question, but you could clean this up
and get rid of that nasty $i stuff with something along the lines of
(untested):
$def = shift(@fields);
foreach my $line (@fields) {
$seq .= $line
}

> if ($seq=~ /(.....)(LY.L..L..L)(.....)/ and $def=~ /[Homo sapiens]/) {

The problems you asked about are in this line. They are:
1.) [Homo sapiens] is a character class; perldoc perlre for more
information. Drop the square brackets if (as I assume) you just
want to check for the string "Homo sapiens" somewhere in the
string. Or, if you need it at the beginning, use /^Homo sapiens/
or, for that matter, ( ...and substr($def, 0, 12) eq 'Homo sapiens')
2.) The second regex resets the $1..$3 variables to undef(), since
they refer to the last match. You could try switching the order
of the regexen, which might speed things up a little anyway.


Hope this helps.

--
----------------------------------------------------------------------
Andrew J Perrin - [url]http://www.unc.edu/~aperrin[/url]
Assistant Professor of Sociology, U of North Carolina, Chapel Hill
[email]clists@perrin.socsci.unc.edu[/email] * andrew_perrin (at) unc.edu

[Stephan Bour]

in article [email]847k77uthr.fsf@perrin.socsci.unc.edu[/email], Andrew Perrin CLists at
[email]clists@perrin.socsci.unc.edu[/email] wrote on 6/27/03 16:59:

> Stephan Bour <sbour@niaid.nih.gov> writes:
>

>> #! usr/bin/perl -w

>
> use strict; # is your friend
>

>>
>> $/="\n>";
>>
>> while(<>){
>>
>> $seq=$_;

>
> what's the point of this when you just set it back to "" below?
>

>>
>> @fields=split(/\n/,$_);
>> $i=-1;
>> $seq="";
>>
>> foreach $line (@fields) {

>
> why assign $line when you never use it?
>

>>
>> $i++;
>> if($i==0) {
>> $def=$fields[$i];
>> }
>> else {
>> $seq.=$fields[$i];
>> }
>> }

>
> This is not in answer to your question, but you could clean this up
> and get rid of that nasty $i stuff with something along the lines of
> (untested):
> $def = shift(@fields);
> foreach my $line (@fields) {
> $seq .= $line
> }
>

>> if ($seq=~ /(.....)(LY.L..L..L)(.....)/ and $def=~ /[Homo sapiens]/) {

>
> The problems you asked about are in this line. They are:
> 1.) [Homo sapiens] is a character class; perldoc perlre for more
> information. Drop the square brackets if (as I assume) you just
> want to check for the string "Homo sapiens" somewhere in the
> string. Or, if you need it at the beginning, use /^Homo sapiens/
> or, for that matter, ( ...and substr($def, 0, 12) eq 'Homo sapiens')
> 2.) The second regex resets the $1..$3 variables to undef(), since
> they refer to the last match. You could try switching the order
> of the regexen, which might speed things up a little anyway.
>
>
> Hope this helps.

You're right, you're right and you're right. It is still very slow (the
FASTA file is over 600 MB) but it works!
Thanks for your help,
Stephan.

[Jay Tilton]

Stephan Bour <sbour@niaid.nih.gov> wrote:

: I have a FASTA file containing alternated ID and sequence lines. I need to
: find a pattern in the sequence but only return it as a match if the
: corresponding ID line contains the string [Home sapiens]. The code below
: returns all matches when I omit the "and $def=~ /[Homo sapiens]/" string
^^^^^^^^^^^^^^
Square brackets have meaning in a regex. That will actually match any
string containg the letters H, a, e, i, m, n, o, p, or s. Escape the
brackets to match a literal '[Homo sapiens]' .

$def=~ /\[Homo sapiens\]/

Better still, use index(), which is more suited to finding literal
substrings.

: but gives me the following error when included:
:
: Use of uninitialized value in concatenation (.) at blast.pl line 26, <>
: chunk 22308.
: Use of uninitialized value in concatenation (.) at blast.pl line 26, <>
: chunk 22308.
: Use of uninitialized value in concatenation (.) at blast.pl line 26, <>
: chunk 22308.

Those are warnings, not errors.

: Does that mean that $def is null in the regex?

The warnings point at the print statement, not at the match.

[code trimmed]

: if ($seq=~ /(.....)(LY.L..L..L)(.....)/

At this point, $1, $2 and $3 are defined when that match succeeds.

: and $def=~ /[Homo sapiens]/

And when that match succeeds, the values in $1, $2 and $3 get
clobbered. Doh!

: ) {
: $j++;
: print ">$def\n\n$1\t$2\t$3\n\n\n";
: }

Use the list returned by m// to keep captured substrings around,

if (
if (
my @grabs = $seq=~ /(.....)(LY.L..L..L)(.....)/
and
$def=~ /\[Homo sapiens\]/
) {
$j++;
print ">$def\n\n$grabs[0]\t$grabs[1]\t$grabs[2]\n\n\n";
}

or use index() in place of the second m//,

if(
$seq=~ /(.....)(LY.L..L..L)(.....)/
and
index($def, '[Homo sapiens]') >= 0
) {
$j++;
print ">$def\n\n$1\t$2\t$3\n\n\n";
}

or just swap the order of the matches in the condition.